Beyond Efficiency: Rethinking Human-entered AI for Cognitive Sustainability

Beyond Efficiency: Rethinking Human-entered AI for Cognitive Sustainability | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Beyond Efficiency: Rethinking Human-entered AI for Cognitive Sustainability Oshri Bar-Gil This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9380629/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract This study examines the impact of AI-powered recommendation systems. Using a netnographic approach, we analyzed user experiences with Google's recommendation systems ecosystem over five years (2016–2020). Integrating post-phenomenological analysis, actor-network theory, and extended cognition theory, we used thematic analysis of 525 blog posts, 46 books, and 25 academic studies to reveal how recommendation systems mediate cognitive processes. Our findings revealed potential influences on human autonomy, intention, rationality, and memory. Three patterns were identified. First, users experience decisional delegation, where algorithmic suggestions gradually replace autonomous decision-making. Second, intentionality becomes hybridized as users struggle to distinguish self-generated intentions from algorithmically mediated ones. Third, memory as a basis for decisions and sense-making shifts from individual and collective to algorithmically mediated connective memory, where personal history becomes inseparable from platform-mediated data aggregation. These mediating effects challenge prevailing human-centered AI paradigms that prioritize human control and transparency without addressing potential cognitive impacts. We propose reconceptualizing HCAI around cognitive sustainability: preserving autonomous decision-making, maintaining authentic intention formation, and protecting memory sovereignty. Practical implications include implementing reflexivity mechanisms, establishing temporal boundaries for cognitive preservation, and developing assessment frameworks to evaluate long-term cognitive impacts beyond efficiency gains. Human-centered AI Recommendation Systems Digital Autonomy Algorithmic Decision-Making Sustainable Innovation User Empowerment Netnography Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction In recent years, the available information for decision-making has grown enormously, fueled by advances in high-speed communication networks, increased data storage capabilities, powerful algorithms, and cloud computing (McAfee & Brynjolfsson, 2017 ). It put individuals in a continual influx of extensive information streams, vying for attention, and requiring rapid decision-making (Andrejevic, 2013 ). According to the German philosopher Hartmut Rosa ( 2013 ), humans enter a cycle of acceleration as the rate of technological development increases. Technological acceleration causes a hastening of lifestyle, which requires new technologies to deal with this acceleration. One technology for achieving this goal is AI-powered recommendation systems (RS). The integration of RS into digital platforms has become a crucial technological solution for alleviating the burden of information overload. Recommendation systems can be defined as “software tools and techniques that provide suggestions for available items that are most likely of interest to a particular user” (Ricci et al., 2015 ). The technical details of this area are continuously developing and are beyond the scope of this study. RS use algorithms to analyze user data and provide personalized suggestions or content "helping" individuals handle day-to-day decision-making from restaurant recommendations (Abbas & Dervin, 2009 ; Duong, 2017 ; Lardinois, 2018 ), navigation, route selection (Heater, 2018 ), and entertainment such as video and music recommendations (Airoldi et al., 2016 ; Karakayali et al., 2018 ). Individuals now rely upon RS embedded in digital platforms to filter daily information and make thousands of daily decisions, with average Google users contacting platform servers more than 100,000 times daily, as the search guidance in routine decision-making regarding navigation, entertainment, and wider needs based on personalized suggestions (Hill, 2019 ). In the context of Google's ecosystem, we specifically focused on the following types of recommendation systems: Search Recommendations: Google's autocomplete and related search suggestions. Content Recommendations: YouTube's video suggestions and autoplay features. Navigation Recommendations: Google Maps route suggestions and place recommendations. Temporal Recommendations: Google Calendar’s personalized time-management suggestions. Although these systems may not always be considered traditional RS in the narrowest sense, they all employ AI and machine learning to personalize and shape user decision-making based on collected personal and collective data. This broader definition allows us to examine the wider impact of RS mediated decision-making across various aspects of daily life. Recent scholarship has examined how recommendation systems affect user autonomy (del Valle & Lara, 2024 ; Fink et al., 2024 ; Varshney, 2020 ) and decision-making processes (Chen et al., 2013 ; Glickman & Sharot, 2025 ; Ricci et al., 2015 ), with existing work primarily focusing on immediate interactions and conscious choices (André et al., 2018 ; Wertenbroch et al., 2020 ). However, longitudinal changes in cognitive and memory processes through the adoption of RS remain underexplored, particularly how users' fundamental capacities for intentionality, rationality, and memory formation evolve through pervasive algorithmic mediation (Carr, 2011 , 2015 ). Our study addresses this gap by examining not whether autonomy is affected but how? What are the mediation mechanisms that are restructured through extended engagement with RS. We address this gap by integrating two complementary theoretical perspectives: post-phenomenological actor-network analysis of technological mediation and extended cognition theory to have a better perspective on distributed agency of humans using RS. This integrated framework helps us to explain how the tight coupling between individuals and intelligent recommendation systems produces shifts in autonomy, intention, rationality, and memory. Addressing this gap through empirical research is significant for evaluating alignment with the ideals of human-centered AI (HCAI), mainly augmenting human capabilities while maintaining control, freedom, and broader ethical considerations (Bar-Gil, 2024 ; Ozmen Garibay et al., 2023 ; Shneiderman, 2022 ). To this end, this study explores the following four key research questions: How does the use of RS affect users' autonomy, intentionality, and agency? How does RS affect users' decision-making processes and rationality? How does reliance on RS influence personal and collective memory patterns? What are the implications of these findings for developing human-centered AI systems? By addressing these questions, we contribute to a more comprehensive understanding of HCAI to encompass psychological, social, and ethical dimensions of cognitive sustainability. This article begins by contextualizing the acceleration of information flows and the rise of RS as a societal response to information overload. The literature review critically examines how current HCAI definitions neglect the long-term effects on human cognitive capacities, proposing cognitive sustainability as an organizing principle for HCAI. It then develops a conceptual framework grounding autonomy, intentionality, and agency as distributed properties shaped by RS algorithmic mediation. Drawing on post-phenomenological theory (2009), actor-network theory (Latour, 2005 ), and extended cognition (Clark & Chalmers, 1998 ), it synthesizes an integrated theoretical framework for understanding how recommendation systems transform human cognition. The methodology section describes the qualitative netnography method selected and explains the selection of Google’s products and services, data collection, analysis, and research ethical considerations. The findings section is divided into three sections, each corresponding with a research question, and detailing the ways RS may affect 1. intentionality, autonomy, agency, 2. rationality, and 3. Individuals and collective memory patterns. The discussion section considers the limitations of the current research and highlights the need to redefine HCAI more comprehensively, including future research suggestions and practical implications. The contributions of this study are twofold. Theoretically, this study advances the current understanding of human-algorithm interactions. First, whereas existing research examines autonomy, whether it is preserved or diminished, we examine the mechanisms through which autonomy is mediated during sustained engagement with RS. Specifically, we trace how users progressively externalize cognitive processing by delegating decision-making functions to RS, detailing a more nuanced process than preservation or erosion, including the gradual restructuring of the rationale for autonomous choice itself. Grounding this analysis in Searle's (2010) framework of intentionality, we demonstrate how algorithmic mediation operates not simply as an external influence but as a mediator of intentions that constitute autonomous agency (del Valle & Lara, 2024 ; Fink et al., 2024 ; Varshney, 2020 ). Second, we extend Hoskins's (2011, 2018, 2024) connective memory concept to the domain of platformized RS that mediate memory formation and usage through aggregated behavioral data of the individual and the collective through RS human-algorithmic mediated entanglement, restructuring memory as a scaffolding for future decision-making processes. Practically, this study contributes to the reconceptualization of HCAI design as a means of cognitive sustainability by expanding its scope. Existing HCAI paradigms emphasize human control perspectives (Bar-Gil, 2024 ; Ozmen Garibay et al., 2023 ; Riedl, 2019 ; Shneiderman, 2022 ; D. Wang et al., 2021 ). Our empirical findings suggest that sustainable HCAI should extend to encompass the preservation and even enhancement of underlying human capacities. Specifically, our findings highlight that human flourishing in increasingly RS mediated contexts requires attention to the potential long-term effects on human autonomy, intentionality, and the autonomous capacity for sense-making. This reconceptualization challenges current design trajectories and calls for the integration of cognitive sustainability as a central principle, alongside the existing emphasis on transparency and user empowerment. Literature review Human-Centered AI as a Design Program for Cognitive Sustainability? HCAI emphasizes the design of AI systems that prioritize values such as fairness, accountability, interpretability, and transparency, as well as the primacy of human needs and well-being (Bar-Gil, 2024 ; Martini et al., 2024 ; Ozmen Garibay et al., 2023 ; Riedl, 2019 ; Shneiderman, 2022 ). In the context of HCAI, sustainability is not limited to environmental or economic outcomes but is redefined to encompass the ethical, social, and psychological dimensions of human-AI interaction (Cinar & Bilodeau, 2024 ; Martini et al., 2024 ; Mhlanga, 2022 ). Specifically, we adopt the concept of cognitive sustainability (Yadav, 2025 ) as the preservation and enhancement of human cognitive capabilities, decision-making autonomy, and skill development through AI usage. While AI systems may optimize immediate task performance, sustainable design should prioritize the long-term preservation of human functioning and flourishing, ensuring that users maintain their capacity for independent judgment, even as they benefit from RS assistance (Bar-Gil, 2024 ; Mhlanga, 2022 ). Agency as Distributed and Emergent Considering cognitive sustainability, HCAI frameworks underscore the preservation of human agency and autonomy in decision-making, which are central to self-actualization. Following Wertenbroch et al. ( 2020 , p. 430), we adopt a definition of autonomy grounded in both user behavior research and psychological theory, conceptualizing autonomy as an individual’s ability to make and enact decisions independently, free from external influences imposed by other agents, such as RS. This conceptualization aligns with other philosophical notions of self-determination and free will (André et al., 2018 ; Pham et al., 2021 ). Fink et al. ( 2024 ) point to the psychological basis of this definition in self-determination theory, which considers actions autonomous when they are characterized by feeling volitional or self-endorsed, where individuals feel choiceful and integrated in their behavior, fully standing behind their own actions (Ryan & Deci, 2011 ). In the context of RS autonomy becomes particularly salient, as their suggestions constitute external influences that shape decision-making. When users interact with RS, the question arises as to whether their subsequent choices remain truly autonomous and volitional or whether algorithmic mediation compromises their capacity for autonomous decision-making. Recent scholarship has established that RS affects user autonomy (del Valle & Lara, 2024 ). Varshney ( 2020 ) proposed operationalization principles for respecting human autonomy, emphasizing the need for systems that preserve user agency, and Fink et al. ( 2024 ) empirically demonstrated that increasing user autonomy enhances recommendation acceptance. This conceptualization provides the foundation for examining the mediated interaction with RS as influencing users' autonomy from conscious self-determination towards various degrees of diminished agency in decision-making. Intentionality and the Mediation of Intentions and Actions Thus, understanding the role of human agency in mediated decision-making is crucial. Pickering ( 2010 ) contends that agency can emerge from interactions between humans and non-human entities and is not solely the prerogative of human actors. He argues that agency is a relational and emergent property that arises from the entanglement of human and non-human elements and is not a fixed attribute of individual actors. His concept of “the dance of agency” captures the ongoing interplay and negotiation between different actors, including users, platforms, and algorithms, and recognizes that agency is distributed and enacted through these interactions. Another key concept in understanding and analyzing agency is intentionality. It has been defined and interpreted in various ways over the years (Anscombe, 1957 ; Searle, 2010 ; Smith, 2017 ). According to Searle ( 1983 , p. 1), intentionality is a mental state that involve being directed towards, about, or representing specific entities, events and situations. Searle ( 1983 ) differentiated between two types of intentions. The first type of intention is aimed at triggering action, such as forming the intention to raise one’s hand in 30 seconds. Searle refers to this as ‘prior intention.’ The second type of intention relates to the action itself, which occurs when an individual raises their hand. Searle calls this ‘intention in action.’ He argues that in every action we perform, we take for granted the social context in which we are embedded, which consists of beliefs, abilities, and possibilities, as manifested in prior intentions, which constitute an inner plan of action that directs the actions of the user. This entails visualizing the desired result and plotting the method to obtain it. The fulfillment of prior intention exemplifies the user’s logical evaluation of practicality and the actualization of potentialities. RS produce new affordances for decision-making, neither exclusively human nor algorithmic, but include technological and social interactions (Gibson, 1979 ). Google’s vast databases of users’ intentions and actions, combined with networking and computing power, allow for delegating one’s decision intentions to the RS so it can produce an ‘intention in action’ complementing or sometimes enhancing one’s own intention and decision making, thus diminishing the user’s autonomy. Rationality and Algorithmic Decision-Making ​RS mediate the intention-to-action pathway from user decision-making intention to the users’ adoption or rejection of RS suggestions. This generates a disparity between the prior intention of the user and the suggestion generated by the RS as a suggested new prior intention. This disparity, described by Searle ( 2010 ) as the "causality gap," contributes to the perceived choice of the user, as it maintains one’s free will. Fisher ( 2020a ) explains that algorithmic decision-making rationality is rooted in optimization, efficiency, and data-driven decision-making. This represents a departure from human decision-making rationality to computational and statistical algorithmic models, as RS’s unique logic is shaped by patterns and biases extracted from training data. Thus, reliance on RS algorithmic rationality can affect human notions of rationality (Doneson, 2019 ; Fisher & Mehozay, 2019 ). Connective Memory as Platformized Algorithmic Connectivity ​Hoskins ( 2011 , 2018 ) introduced the concept of connective memory as a framework that expands our understanding beyond personal or collective memory. His concept of connective memory has evolved from its initial focus on interpersonal digital networks to encompass broader forms of technologically mediated memory. While originally emphasizing human-to-human connections through digital platforms, recent developments in this concept support the claim that connective memory can increasingly emerge through complex assemblages of human and algorithmic actors (Bar-Gil, 2025 ; Hoskins, 2018 , 2024 ).​ ​In the context of RS, we extend his concept to examine how algorithmic mediation creates new patterns of individual and collective memory connectivity and their influence on ​decision making. For example, Google Maps aggregates real-time location data from millions of users to generate traffic predictions and route recommendations (Lau, 2020 ). It creates a form of collective memory that emerges from the aggregation of human behavior traces. Each user's navigational intentions (i.e., planning a route) and behavior (i.e., moving) contribute to a constantly updating database, as a memory that shapes future recommendations for all users and their future selves, as a form of algorithmically mediated connectivity. Recommendation Systems as Extended Cognition The idea of extended cognition proposes that the human mind and cognitive processes are not solely confined to the brain or body but can encompass interactions with external elements, including technologies (Clark & Chalmers, 1998 ). Clark and Chalmers crucial criteria center not on the location of the cognitive process but on the functional role played in enabling an integrated cognition. They suggested that even portable or transient external artifacts, such as notebooks accessed to retrieve pivotal stored details, can become tightly coupled with cognitive processes. Building on this conceptual foundation, over two decades of subsequent inquiry by neuroscientists, psychologists, and philosophers supports the extensive entanglement of the brain, body, and world into an ensemble that produces cognition, behavior, and intelligence while avoiding commitment to strict biological localization (Clark, 2016 ; Menary, 2010 ). This distributed ecological perspective contrasts with the brain as a predominantly inward data processor by highlighting the porous boundaries across humans and artifacts contingent on active environmental scaffolding (Heersmink, 2017 ). Clark and Chalmers ( 1998 ) seminal thought experiment contrasts two characters, Otto and Inga, who both wish to visit the Art Museum. Inga relied on her biological memory to recall the museum’s location and successfully navigated her planned route. In contrast, Otto has early stage Alzheimer’s disease; therefore, he writes down directions in a notebook to serve as an external memory aid. Otto arrives at the same destination despite his cognitive impairment by scaffolded interaction with the notebook as an external portable artifact preserving navigation details. This example delineates the core premise of their hypotheses, that cognition manifests not just ‘within the skull’ in the brain but rather emerges from coordinated engagement with environmental resources. Building on Otto and Inga, we can introduce a third character, Alex, who relies on a smartphone-based RS, Google Maps, to navigate to the museum. Alex verbally states her intention as desired destination and is presented with optimized driving, walking and public transit routes based on real-time traffic data, personal location history and stored points of interest. Turn-by-turn narration guides momentary micro-actions and navigational decision-making, while the application passively tracks progress, recalibrates suggestions if unexpected delays emerge en route, and logs a ‘memory’ of the route traveled. As Alex navigates to the museum via Google Maps, her set route is customized by saving details from previous searches and paths traveled, stored seamlessly in her Google account profile. Specific waypoints along the path are prominently shown if they are related to previous searches. Other museums can be suggested in her search results based on Google’s database or paid ad partnerships with Google, even though visiting them may divert Alex from her original intention. By leveraging Alex's prior intentions, Google aims to provide a "personalized" experience that caters recommendations to her perceived preferences, challenging Alex with each decision along the way. This additional scenario highlights how extensive entanglement with platformized RS, as sociotechnical artifacts, restructure not only navigation activities but also propensities, decisions, and recall capabilities distributed across the Google map service as an extended cognitive aid. The museum navigation example reveals several key distinctions in autonomy, intention, rationality, and memory patterns attributable to reliance on RS versus biological or other external cognitive resources. Most saliently, Alex delegates aspects of wayfinding to route-planning application rather than directing her internal efforts. Likewise, intention alignment may shift from self-selected intentions to those suggested algorithmically. Rationality shifts from individual bounded judgments to efficiency optimizations enacted through exhaustive, real-time data processing. Memory, as a basis for decision-making, transforms from individual constructive retention of navigation waypoints to reliance on a massive geographical database that includes users’ movement patterns to be invoked on demand. While Alex may arrive at the museum aided by her smartphone, the layers of digital mediation affect the richness of encoding spatial memories, attentional engagement, and sense of agency over her journey. Further sections delve deeper into the changes propagated across each of these dimensions from the theoretical perspective of technical mediation theory. Post-Phenomenology and Technical Mediation Over the last three decades, Contemporary philosophers of technology including Latour ( 1999 , 2005 ), Ihde ( 1990 , 2009 ) and Mitcham ( 1994 ), have that their predecessors, thinkers like Ellul (1954/2011), Heidegger (1953/2008), and Jaspers ( 1957 ), imposed a uniform model of dystopian attitudes toward ‘Technology.’ Their own perspective emphasize the function of technology as a mediator, constructing and shaping the relations between users and their environment (Coeckelbergh, 2013 , p. 41; Feenberg, 2005 ; Ihde, 2009 ). The French philosopher Bruno Latour focused on the role of technology in shaping human experience through networks of human and non-human actors (or actants) interactions (Latour, 1999 , 2005 ). Applied to our context of AI-based, platformized RSs of Google, Latour’s framework might facilitate the description and interpretation of the mediation networks involving the user as a human actor and the platformized, databased, algorithmic RSs as non-human actors. Ihde's post-phenomenological approach emphasizes the human experience of technological mediation and how it influences human actions, perceptions, and interpretations of the world (Ihde, 1990 , 2009 ). Ihde differentiates between micro and macro perceptions to explore how human perception is shaped by technology and the environment. Micro-perception refers to the immediate sensory experience of the world, where our senses are attuned to specific stimuli. For example, microscopes extend our micro-perception by allowing us to focus on and magnify specific aspects of our environment, thus contributing to changes in macro perception. Macro-perceptions refer to a broader contextual beliefs about the world, considering larger systems, structures, and meanings. Ihde emphasizes the exploration of both micro and macro perceptions to gain a comprehensive understanding of our lived experiences and the ways in which technology mediates our perception of the world. Applied to the context of RS, Ihde's framework enables an exploration of how users interact with RS mediated decision-making, as micro-perceptions, impacting their macro- perception of intentionality, agency, and rationality as a result of those interactions. Both theories might offer a valuable lens for interpretation, but they cannot be considered methodologies in the empirical sense (Latour, 2005 ; Nimmo, 2011 ). To enable empirical examination, this study adopts a netnographic method to analyze publicly posted narratives by individuals discussing encounters with Google’s sociotechnical RS ecosystem. Post-phenomenological concepts guide the theoretical interpretations of the findings on agency, autonomy, intentionality, rationality, and memory, as detailed in the ensuing methodology overview and described in the following section. Integrated Theoretical Framework The theoretical perspectives presented above converge into an analytical framework that mutually reinforces and supports each theory. Technological mediation creates the conditions and offers descriptive and interpretive constructs for distributed agency, enabling the use of actor-network analysis of RS as cognitive extensions. Thus, the framework provides a comprehensive lens for analyzing how AI recommendation systems transform autonomy, agency, intentionality, rationality, and memory, which are the key dimensions we have set for examination. Methodology This study employed netnography, a qualitative research method designed to study online communities and cultures (Kozinets & Gambetti, 2021 ). Netnography is a specific set of research practices related to data collection, analysis, research ethics, and presentation, rooted in participant observation (Kozinets, 2015 , 2019 ; Kozinets & Gambetti, 2021 ). We chose netnography for its ability to capture authentic user narratives about their evolving relationships with RS through naturally occurring digital discourses (Addeo et al., 2019 ; Bartl et al., 2016 ). Google's ecosystem was selected as the primary case study for several reasons: First, its comprehensive range of integrated services (Search, Maps, YouTube, Calendar, Assistant) enables systematic observation of how recommendation algorithms shape user behavior across different life domains. Second, the sophisticated AI algorithms of the platform for personalized RS create an ideal environment for research. Third, Google's ubiquity means that users have sustained longitudinal interactions that allow for the examination of longitudinal transformation patterns. Our research design examines both individual Google services and their platform-level integration, analyzing how interconnected recommendation systems collectively transform user autonomy, rationality, and memory patterns. This holistic approach is crucial for understanding the cumulative effects of algorithmic mediation on multiple life activities. Data Collection Data collection focused on capturing discourse about user experiences with Google's recommendation features across the 2016–2020 period. We selected this timeframe to capture RS after achieving presence but before the emergence of generative AI, providing a focused window into a specific phase of human-algorithm interaction. We collected data from five primary sources. Four technology review blogs (The Verge, Wired, Engadget, and Ars Technica) were selected based on their comprehensive Google product coverage, significant user engagement (minimum 500,000 monthly readers), and diverse readership, which spanned casual users to technology professionals. These sources provided organic user discourse on experiences with algorithmic recommendations. The Google Keyword Blog: While acknowledging its nature as corporate communication rather than organic user review blog, Keyword provided two analytical values: (1) unmediated access to Google's design intentions and feature framing and (2) early adopter responses from power users who engaged with features as designed. Content selection prioritized posts discussing user experiences with recommendation features, algorithmic decision-making, personalization systems, and behavioral adaptations to RS suggestions. This focused approach enabled the examination of how users negotiate with RS. We employed methodological triangulation (Køster & Fernandez, 2021 ; Levitt et al., 2017 ) by integrating additional sources: 46 books analyzing Google's ecosystem provided longitudinal context, and 25 academic studies offered theoretical frameworks and comparative baselines. Overall, our dataset comprised 525 blog posts, 46 books, and 25 academic studies (detailed in the Supplementary Materials). Researcher Immersion and Reflexive Practice Following the netnographic principles of researcher engagement and immersion (Kozinets, 2024 ), the research process included immersive and participant observation within the community and with the products themselves. The lead researcher integrated Google products into his daily life, experiencing firsthand the features discussed in user discourse while maintaining a detailed reflexive journal documenting his responses to algorithmic mediation. Data Analysis We employed computer-assisted qualitative data analysis software ATLAS.Ti (Version 8) to process and analyze the large volume of collected data. This software facilitated the systematic organization and coding of textual data while maintaining analytical transparency and rigor (Krippendorff, 2018 ; Woolf & Silver, 2017 ). Using Braun and Clarke's (2006) approach, we employed a three-stage analytical process that progressed from empirical observation to theoretical interpretation: First, we conducted open coding using an inductive approach to identify emerging themes and patterns in user discourse about Google services. Second, we organized the codes into higher-order themes capturing key patterns (e.g., "hybrid agency," "algorithmic rationality," "connective memory"). This involved organizing the codes into meaningful themes and exploring the relationships between them (the full codebook is available as supplemental material). Third, as theoretical interpretation, we applied our integrated framework of post-phenomenology, actor-network theory, and extended cognition to interpret how recommendation systems transform user experience. For example, Google Calendar's promise to 'help you find the time and stick to it' (Ramnath, 2016 ) was coded in the first stage as: calendar; Algorithmic scheduling; Delegation of decision-making; and Temporal agency . In the second stage, thematic development, it was grouped with similar codes from across the dataset into the higher-order theme: "dance of agencies" - capturing the ambiguous state where users' goals become entangled with algorithmic suggestions, creating uncertainty about the locus of agency in decision-making. In the third stage, actor-network theory analysis (Latour, 2005 ) revealed that the Calendar functions as an actant that actively shapes user intentions rather than merely saving time slots for the user, as earlier calendar services did (Lord, 2008 ). This systematic analytical progression enabled both a detailed examination of specific mediations by product and a broader theoretical understanding of how Google’s RS transforms user experiences. Research Ethics The research protocol was approved by [Institution anonymized] Ethics Committee. Following digital research ethics conventions (Hookway, 2008 ; Kozinets, 2019 ), we collected only publicly accessible data. While recognizing that users might not anticipate research use, we implemented strict anonymization by removing all identifiable information from non-author commentators. To preserve authentic discourse, we maintained a non-interventionist approach, avoiding any participation that might influence natural discussions. Findings The finding presented below were organized to addresses each our research questions through a systematic presentation of user discourse and it’s interpretation in each domain. Intentionality, agency, and autonomy Our analysis of user discourse reveals specific patterns in how autonomy mediation manifests, directly addressing the research question: How does prolonged use of RS affect users' autonomy, intentionality, and agency? Autonomy can be thought of as the capacity to be one’s own person, to live one’s life according to reasons and motives that are taken as one’s own and not the product of manipulative or distorting external forces, to be independent (Christman, 2020 ; del Valle & Lara, 2024 ). Intention plays a key role in autonomous human cognition and behavior by regulating deliberative processes tied to fulfilling purposes that give meaning to activities over time (Bratman 1987 ; Searle 1983 ). When considering John Searle’s and Bruno Latour’s ( 2005 ) theoretical perspectives, the implications of technological mediation through the delegation of intentionality are noticeable. By delegating some of their intention of decision-making to RS, users allow the algorithm to shape not only their choices or their resulting behavioral actions but also to conform their intention itself to algorithmic rationality. The Google Maps navigation tool continually optimizes routes as RS based on external data analysis for over 10 billion devices (J. Wang, 2021 ), including various features aimed to assist the user in decision making, such as “ popular times to avoid” graph (Duong, 2017 ), waiting time estimates that can influence the intention to select and visit a particular place in a particular time, when seeing that the place is busier than usual (see Fig. 1 ). YouTube's recommendation algorithm exemplifies the profound transformation of user agency and intentionality through algorithmic mediation. The platform's autoplay feature doesn't merely suggest content, but it actively reconstructs viewing intentions by creating seamless transitions between videos that users never consciously selected. Beyond mere mediation of intentionality, this represents what Karakayali et al. ( 2018 ) identify as the platform's prioritization of engagement metrics and advertiser interests over users' organic viewing intentions. The algorithm's influence on intentionality operates through multiple mechanisms. First, it exploits the prior-intention to intention-in-action gap (Searle, 1983 ) by automatically loading new content before users can exercise a deliberate choice. Second, it progressively learns and shapes user preferences through a feedback loop in which each viewed video (whether actively chosen or passively accepted) influences future recommendations, gradually steering users toward content that maximizes platform engagement rather than satisfying their original viewing intention. Third, the system creates what users describe as 'rabbit holes', trajectories of increasingly specific or extreme content that led far from the original intentions (Ledwich et al., 2022 ). Ledwich and Zaitsev ( 2019 ) found that YouTube's recommendation system can lead to harmful consequences such as online radicalization. This represents not only a failure of content moderation but also a fundamental challenge to user autonomy: the algorithm does not simply respond to existing intentions and preferences but actively constructs new ones, potentially transforming users' political orientations and worldviews without their awareness or consent. Users' agency becomes entangled with algorithmic processes designed to maximize watch time, mediating intention, where the distinction between what users want to watch and what the algorithm wants them to watch becomes increasingly blurred. Across contexts, RS leveraging behavioral data and predictive analytics for personalization to subtly align intentions more with platform priorities than support users’ agency in their sense-making, thereby influencing their decisions (Clowes, 2015 ; Kitchin & Dodge, 2011 ). Google advertises the hyperpersonalization of map services as follows: “What if we told you that during your lifetime, Google could create millions of custom maps...each one just for you? In the past, such a notion would have been unbelievable: a map was just a map, and you got the same one for New York City, whether you were searching for the Empire State Building or a coffee shop on the street. What if, instead, you had a map that is unique to you, always adapting to the task you want to perform right this minute?“ (Pichai, 2013 ). Another example is Google Calendar's 'Goals' feature, which automatically schedules time for user-defined objectives (exercise, reading, learning, etc.). Users specify what they want to achieve and their preferred frequency; the algorithm then analyzes their existing calendar, identifies free time slots, and automatically schedules goal sessions (see Figs. 2 , 3 ). The system adapts when users defer or complete sessions and learns optimal timing patterns: “Whether it is reading more books, learning a new language, or working out regularly, achieving your goals can be difficult. One day it's "I got called into a last-minute meeting." The next day it's "I have a friend in town." Before you know it, your goals are delayed or forgotten. In fact, with all the things you need to do in a given week, it is probably harder than ever to find the time—even when your goal really matters to you. That’s why starting today, we are introducing Goals in Google Calendar. Just add a personal goal—like “run 3 times a week”—and Calendar will help you find the time and stick to it.” (Ramnath, 2016 ). An additional feature in Google’s calendar even nudges you to complete your tasks: “You might already create calendar entries to remind you to call the doctor or pick up groceries on the way home. But while those entries come and go, Reminders stick with you over time so you can track them until they are actually done. If a Reminder isn't completed, it will appear at the top of your Calendar the next day. And the next. When you do finally call the doctor or pick up those necessities” (Umapathy, 2015 ). Google Calendar represents a shift from a passive calendar to an active temporal mediation agent, shaping user behavior through algorithmic scheduling and execution through nudges and reminders. Generalizing for different Google RS, we can see that users increasingly delegate parts of their decision-making process to algorithmic based RS: Temporal aspects such as automated meeting scheduling (Asara, 2016 ; Schieffer, 2016 ) and optimization of time slots for achieving personal goals (Goerisch, 2017 ) to Google Calendar, spatial aspects of travel planning that use machine learning and connective memory of traffic patterns in real time to Google Maps (Gartenberg, 2017 ; Lau, 2020 ; Maier, 2016 ). Seeking frictionless, optimal time and space experiences from users can lead to dependence on external recommendation algorithms that might hinder the capacities for autonomous decision-making. One way of understanding these user choices, which diminish their agency, is to consider a potential change in the rationality of the perceived decision-making using RS. Intentionality, agency, and autonomy Our analysis of user discourse reveals specific patterns in how autonomy mediation manifests, directly addressing the research question: How does prolonged use of RS affect users' autonomy, intentionality, and agency? Autonomy can be thought of as the capacity to be one’s own person, to live one’s life according to reasons and motives that are taken as one’s own and not the product of manipulative or distorting external forces, to be independent (Christman, 2020 ; del Valle & Lara, 2024 ). Intention plays a key role in autonomous human cognition and behavior by regulating deliberative processes tied to fulfilling purposes that give meaning to activities over time (Bratman 1987 ; Searle 1983 ). When considering John Searle’s and Bruno Latour’s ( 2005 ) theoretical perspectives, the implications of technological mediation through the delegation of intentionality are noticeable. By delegating some of their intention of decision-making to RS, users allow the algorithm to shape not only their choices or their resulting behavioral actions but also to conform their intention itself to algorithmic rationality. The Google Maps navigation tool continually optimizes routes as RS based on external data analysis for over 10 billion devices (J. Wang, 2021 ), including various features aimed to assist the user in decision making, such as “ popular times to avoid” graph (Duong, 2017 ), waiting time estimates that can influence the intention to select and visit a particular place in a particular time, when seeing that the place is busier than usual (see Fig. 1 ). YouTube's recommendation algorithm exemplifies the profound transformation of user agency and intentionality through algorithmic mediation. The platform's autoplay feature doesn't merely suggest content, but it actively reconstructs viewing intentions by creating seamless transitions between videos that users never consciously selected. Beyond mere mediation of intentionality, this represents what Karakayali et al. ( 2018 ) identify as the platform's prioritization of engagement metrics and advertiser interests over users' organic viewing intentions. The algorithm's influence on intentionality operates through multiple mechanisms. First, it exploits the prior-intention to intention-in-action gap (Searle, 1983 ) by automatically loading new content before users can exercise a deliberate choice. Second, it progressively learns and shapes user preferences through a feedback loop in which each viewed video (whether actively chosen or passively accepted) influences future recommendations, gradually steering users toward content that maximizes platform engagement rather than satisfying their original viewing intention. Third, the system creates what users describe as 'rabbit holes', trajectories of increasingly specific or extreme content that led far from the original intentions (Ledwich et al., 2022 ). Ledwich and Zaitsev ( 2019 ) found that YouTube's recommendation system can lead to harmful consequences such as online radicalization. This represents not only a failure of content moderation but also a fundamental challenge to user autonomy: the algorithm does not simply respond to existing intentions and preferences but actively constructs new ones, potentially transforming users' political orientations and worldviews without their awareness or consent. Users' agency becomes entangled with algorithmic processes designed to maximize watch time, mediating intention, where the distinction between what users want to watch and what the algorithm wants them to watch becomes increasingly blurred. Across contexts, RS leveraging behavioral data and predictive analytics for personalization to subtly align intentions more with platform priorities than support users’ agency in their sense-making, thereby influencing their decisions (Clowes, 2015 ; Kitchin & Dodge, 2011 ). Google advertises the hyperpersonalization of map services as follows: “What if we told you that during your lifetime, Google could create millions of custom maps...each one just for you? In the past, such a notion would have been unbelievable: a map was just a map, and you got the same one for New York City, whether you were searching for the Empire State Building or a coffee shop on the street. What if, instead, you had a map that is unique to you, always adapting to the task you want to perform right this minute?“ (Pichai, 2013 ). Another example is Google Calendar's 'Goals' feature, which automatically schedules time for user-defined objectives (exercise, reading, learning, etc.). Users specify what they want to achieve and their preferred frequency; the algorithm then analyzes their existing calendar, identifies free time slots, and automatically schedules goal sessions (see Figs. 2 , 3 ). The system adapts when users defer or complete sessions and learns optimal timing patterns: “Whether it is reading more books, learning a new language, or working out regularly, achieving your goals can be difficult. One day it's "I got called into a last-minute meeting." The next day it's "I have a friend in town." Before you know it, your goals are delayed or forgotten. In fact, with all the things you need to do in a given week, it is probably harder than ever to find the time—even when your goal really matters to you. That’s why starting today, we are introducing Goals in Google Calendar. Just add a personal goal—like “run 3 times a week”—and Calendar will help you find the time and stick to it.” (Ramnath, 2016 ). An additional feature in Google’s calendar even nudges you to complete your tasks: “You might already create calendar entries to remind you to call the doctor or pick up groceries on the way home. But while those entries come and go, Reminders stick with you over time so you can track them until they are actually done. If a Reminder isn't completed, it will appear at the top of your Calendar the next day. And the next. When you do finally call the doctor or pick up those necessities” (Umapathy, 2015 ). Google Calendar represents a shift from a passive calendar to an active temporal mediation agent, shaping user behavior through algorithmic scheduling and execution through nudges and reminders. Generalizing for different Google RS, we can see that users increasingly delegate parts of their decision-making process to algorithmic based RS: Temporal aspects such as automated meeting scheduling (Asara, 2016 ; Schieffer, 2016 ) and optimization of time slots for achieving personal goals (Goerisch, 2017 ) to Google Calendar, spatial aspects of travel planning that use machine learning and connective memory of traffic patterns in real time to Google Maps (Gartenberg, 2017 ; Lau, 2020 ; Maier, 2016 ). Seeking frictionless, optimal time and space experiences from users can lead to dependence on external recommendation algorithms that might hinder the capacities for autonomous decision-making. One way of understanding these user choices, which diminish their agency, is to consider a potential change in the rationality of the perceived decision-making using RS. Rationality Rational decision-making in human cognition operates within natural constraints—limitations in memory, attention, and cognitive resources, termed by Herbert Simon ( 1971 ) as "bounded rationality." These inherent restrictions make algorithmic RS particularly appealing, as they promise to transcend human cognitive limitations through superior processing speed, vast data integration and computational optimization (Earl, 2016 ). However, the promise of algorithmic rationality represents more than simple augmentation; it fundamentally restructures how intentions are transformed into actions and how users conceptualize rational decision-making itself. Following Searle's (1983) framework of intentionality, with its division of prior intentions and intentions in actions, and Bratman's (1987) elaboration of prior intention as an internal plan with bounded flexibility, we can understand how algorithms mediate the intention-to-action pathway. When cognitive resources are limited, as Simon noted, the success of prior intentions depends on efficient resource allocation for planning and execution processes. From this efficiency perspective, delegating larger parts of decision-making to external algorithmic systems appears not just convenient but rational choice. When actualizing intentions becomes a computational problem, Google's RS algorithms provide optimal pathways from intention to action. The Calendar's "Find a time" feature exemplifies this (see Fig. 4 ): “With a single tap, ‘Find a time’ helps you find meeting times that work for everyone—even if they're in different time zones—based on their availability and the times they usually have meetings. If there are no times that work, Calendar will look at which conflicting meetings can most easily be rescheduled.” (Schieffer, 2016 ). Users perceive this algorithmic logic as inherently more rational because it operates through systematic and consistent rules, which is the foundation of how algorithms function. By incorporating this algorithmic logic into their decision-making processes, users feel more rational, efficiently bridging the gap between abstract intentions and completed actions. Hence, we can notice that users do not merely use RS for efficiency; they internalize algorithmic reasoning patterns. This represents what Fisher and Mehozay ( 2019 ) identified as algorithmic rationality colonizing human judgment; however, our data hint that users actively embrace it, viewing it as a helpful enhancement rather than a logic replacement. More profoundly, these RS act as if they know users' "true" intentions better than users know themselves. This algorithmic epistemology suggests that by analyzing vast behavioral data on what users actually watch, where they actually go, and what they actually search for, algorithms can identify "authentic" preferences that transcend conscious self-reports. If Google's aggregated data about a user provides superior self-knowledge, what could be more rational than allowing this algorithmic representation to mediate intentions, determine navigation paths, schedule time slots, and more for the actual self? This transformation is evident in Google Calendar's approach to time management, where the system does not just execute user commands but actively shapes temporal decisions, as discussed earlier. The calendar’s algorithmic system determines not only when meetings occur, but also which commitments take priority. Significantly, Google maintains the illusion of user control even as it shapes decisions: "'Find a time' makes suggestions, but you're still in control" (Schieffer, 2016 ). This rhetoric echoes the findings of Varshney ( 2020 ) and Fink et al. ( 2024 ) that perceived autonomy increases recommendation acceptance; paradoxically, it should be predicted to lead to a higher acceptance rate of its recommendations. These implications extend beyond specific decisions. These optimized routes, viewing sequences, and scheduling patterns reflect platform priorities – maximizing engagement, advertising exposure, and data collection rather than supporting users' organic self-development or contextual meaning-making (Fisher, 2020a ; Zuboff, 2018 ). Through this creeping delegation process, users transfer increasing portions of their intentional apparatus to RS, despite the risk of eroding the boundaries between their self and technological others. It seems that personalized RS, operated by Google and aligned with corporate intentions, progressively colonize the space of human intention formation. Users increasingly evaluate their decision-making against algorithmic standards, judge their choices using optimization metrics, and doubt their intuitions when they conflict with data-driven recommendations. The question is not whether this is rational, but what definitions of rationality govern human action when interacting with RS. Impact on Memory To address RQ3, which explores how reliance on RS influences personal and collective memory patterns, we examined how users use Google RS as a mediating interface for external connective memory. Recent research examining shifts in memory patterns with technological mediation found the potential for collective memory to emerge through users’ digital behaviors. The increased use of digitally mediated social connections has given rise to a “connective” memory paradigm, facilitating collaborative memory storage, recall, and sensemaking (Hoskins, 2018 ). Our analysis shows that RS create a variation of Hoskins’ connective memory concept, where connectivity emerges not just through platformized human interaction but through algorithmic aggregation, analysis, and reinterpretations of user behaviors and behavior intentions. Consider Google Maps: users’ intention to arrive at a particular destination is manifested by asking for recommended routes and timing, the recommendation stems from their own and collective travel times, and the chosen paths themselves become part of the system's collective dataset, influencing future recommendations for other users. A traffic slowdown experienced by one user immediately affects route suggestions for others, creating an indirect form of collective memory mediated by algorithms. Rather than individual recollections, pieces of fragmentary traces, together with cloud-based crowd-sourced scaffolds, form the connective database for personalized decisions for each user. While Hoskins's (2011) original formulation of connective memory emphasized direct interpersonal connections through digital platforms, our findings follow his later conceptualization of collective algorithmic memory mediation beyond the realm of individual human influence and control (Hoskins, 2018 , 2024 ). We see that RS create de facto data connections, even if they are indirect and mediated by platform databases. When YouTube's algorithm recommends videos based on viewing patterns of similar users, it creates a form of collective influence where users shape each other's experiences without direct interaction, as “connectivity” in RS operates through aggregation and pattern recognition across users. Google's RS does not just respond to individual preferences but analyzes collective behavioral patterns, making each user's actions part of a larger social dataset. Thus, our findings support the concept of connective memory as a pattern of memory emerging from increased connectivity and entanglements with others in real time through an assortment of digital apps, platforms, and networks. Personal contributions to these databases concurrently extract costs by diminishing the motivation for dedicated encoding, storage, and retrieval, which are practiced through cognitive processes traditionally linked to human development. Offloading personal memory onto networks of memorial intermediaries for recall in future decision-making, available on demand across domains such as navigation and entertainment, allows promises of superior memory performance by the platforms but can come with personal and individual costs (Fisher, 2020b ; Risko & Gilbert, 2016 ; Ward, 2013 ; Ward et al., 2017 ). Personalization-induced automation and a lack of transparency around inferential affordances risk misdirecting memory and decision-making (Pasquale, 2015 , 2016 ). Delegating personal memory might also threaten complicity rather than the experience and intuition gained by consciously situating intentions within narratives (Hoskins, 2018 ; van Dijck, 2007 ). Discussion Deeply Human-Centered AI for Sustainability While industry roadmaps currently position progress largely in terms of surpassing functionality milestones around accuracy, speed, scale, and robustness (Metz, 2016 ), this study indicates that such trajectories might have unintended consequences, such as undermining human well-being by disrupting cognitive sustainability. Analysis across the interrelated dimensions of intentionality, agency, rationality, and memory suggests that the use of RS triggers effects on users that progressively diminish facets of decision-making autonomy over time as the dependency and complexity of decision-making increase (Earl, 2016 ; Pham et al., 2021 ). Allowing behavioral data accumulation and predictive nudging to shape intentions, constraining cognitive overload coping strategies to prescribed routines, and restructuring memory patterns around connective databases optimized for algorithmic usability rather than personal ownership risks changing the weights of agencies in decision-making over time. Further research is vital for detailing the boundaries of what sorts of augmentation remain compatible with the notions of human thriving tied to the directed development of self-knowledge, context-dependent rationality, and memory anchoring humans as individuals and humanity as a whole, as can be expected from aligning RS to be more “human-centric.” Revisiting widely cited definitions of HCAI concentrate predominantly on preserving overt control, transparency, and human values validation (Bar-Gil, 2024 ; Mhlanga, 2022 ; Shneiderman, 2022 ). Recentering progress on multifaceted, situated human capabilities, in the context of pervasive RS usage compels expanding frameworks to continuously monitor the broader psychological, political, and social dimensions that are at risk of unintended erosion under increased technological immersion mediating ever growing life aspects (Coeckelbergh, 2013 ; Díaz-Rodríguez et al., 2023 ). Design approaches must be grounded in recognizing humans as growing, intentional actors recursively shaping digital environments that likewise profoundly shape emerging skills, relationships, and self-concept in a reciprocal fashion, rather than one-way efficiency improving delivery mechanisms (Li et al., 2023 ; Robert et al., 2020 ; Wellner & Mykhailov, 2023 ). Limitations and future research This study faced several constraints, some of which were inherent to netnographic research on rapidly evolving technological systems. The decision to focus exclusively on Google's ecosystem provided analytical depth but limited generalizability. Google's specific design and monetization logics, particularly its emphasis on predictive automation and comprehensive data integration across platform services, may produce distinct patterns of cognitive delegation that are not replicated on other platforms. For example, Facebook's social graph-based recommendations or Amazon's commerce-driven algorithmic recommendations likely generate different forms of user behavior, autonomy disruption, and memory externalization. Reliance on publicly posted user discourse introduces selection bias towards users motivated to articulate their experience online. Users who seamlessly integrate with algorithmic systems or even those who resist them but without public commentary remain invisible in our data. Moreover, technology blog comments attract technology reflective users who may not represent typical usage patterns. The inclusion of Google's Keyword blog introduced a specific methodological tension, unlike independent technology blogs, where users freely critique and discuss their experiences. The Keyword represents Google's curated narrative from developers, product managers, and designers as early adopters and power users whose experiences may anticipate but not reflect mainstream adoption. Another limitation of the data collection is the selection of only English-language data, as it limits cultural diversity and might miss how different linguistic structures and cultural concepts of self, memory, and agency shape human-algorithm relations. For example, Mandarin-speaking users navigating localized Google services may experience fundamentally different techno-cultural affordances. Reflecting on the temporal scope of our data collection, spanned 2016–2020, we can see that it represents a critical period when platformized RS became ubiquitous in daily life. While our data predates recent AI developments such as ChatGPT and other generative models, the patterns we identified–delegation of cognitive functions, shifts in personal agency, and transformation of memory practices–appear to be intensifying rather than diminishing with newer technologies. The theoretical framework we developed for understanding human-AI interaction through the lenses of autonomy, intentionality, and rationality provides enduring analytical value. These psychological and philosophical dimensions of human experience remain constant, even as the technical sophistication of AI systems evolves, zs Shneiderman ( 2022 ) notes, the human-centered challenges of AI transcend specific, technical implementations. Nevertheless, this temporal boundary constitutes a significant limitation. Recent developments, particularly in conversational AI and more sophisticated personalization algorithms, may introduce new forms of human-AI interaction that are not captured in our analysis. Future research should examine whether the patterns of cognitive delegation, intentionality shifts, and memory externalization that we document have intensified or evolved with newer AI systems. Theoretical choices shape what becomes visible or invisible in our analysis. Our theoretical framework, which integrates post-phenomenology, ANT, and extended cognition may obscure alternative interpretations. For example a critical theory lens might reveal the power dynamics we choose to leave out of the scope of this article, and a neuroscience perspective could challenge our assumptions about cognitive restructuring. Following these limitations, several research directions emerge: longitudinal studies tracking individual users over time could reveal whether the patterns we identified stabilize, intensify, or even reverse over time with changes in adoption patterns and the emergence of new technological affordances. It will be necessary to determine whether users develop resistance strategies or whether cognitive adaptation reaches equilibrium. Methodologically, combining netnography with cognitive testing could validate whether self-reported experiences of memory externalization correspond to measurable changes in recall or cognitive function. Similarly, experimental interventions temporarily removing algorithmic recommendations could test the dependency effects we theorized as limiting the autonomy of individual users. Cross-cultural analysis should examine not only different user populations but also how varying concepts of autonomy, self, and cognition across cultures shape human-algorithm coupling. For example, does Western emphasis on individual autonomy make users more vulnerable to certain algorithmic influences? Implications for Human-Centered AI Development: Prioritizing Cognitive Empowerment Our findings reveal tensions between current AI development trajectories that prioritize functional expansion, as adopted by users, and the preservation of human cognitive sustainability (Yadav, 2025 ). While prevailing approaches emphasize efficiency gains and frictionless user experiences, our analysis demonstrates how these design choices might erode fundamental human capacities for autonomous decision-making, intentional action, and personal memory formation. These implications necessitate reconceptualizing HCAI development to include cognitive empowerment (Bar-Gil, 2024 ; Ozmen Garibay et al., 2023 ). Preserving Autonomy Through Design Transparency The progressive delegation of decision-making to RS requires systematic interventions to maintain user awareness of mediation processes. RS should implement reflexivity mechanisms that make algorithmic influence visible rather than seamless. One potential suggestion would be to develop transparency utilities that track users' reliance patterns on algorithmic recommendations across temporal scales, enabling the recognition of creeping dependency before it becomes constitutive. Such metrics would illuminate when external recommendations begin to colonize internal judgment processes, for example, by considering the recommendation adoption rate or the reaction time for recommendations allowing transparent presentation for reflection. Increasing the transparency of recommendation patterns and adoption was found to facilitate a shared mental model (Srivastava et al., 2023 ), in itself an intervention that supports cognitive sustainability. ​Design implementations should include clear indicators distinguishing algorithmic suggestions from organic content discovery, with options to experience platforms without personalization filters. This transparency offer extends beyond simple labeling to include explanatory interfaces that reveal how user data shapes recommendations and makes visible the feedback loops between past behavior and future suggestions. Users need comprehension not just of what is recommended but why specific suggestions emerge from their past behavioral or connective patterns.​ Implementing Autonomy Controls Rather than resisting recommendations by users, systems can provide graduated controls, allowing users who would like a higher degree of control to modulate algorithmic influence across different decision domains. Buçinca et al. ( 2021 ) called these interventions as Cognitive Forcing Functions and found that they can reduce overreliance on AI suggestions in decision support systems. Additionally, high-stakes decisions requiring contextual judgment could maintain human primacy, whereas routine tasks employ greater automation. This selective delegation control preserves cognitive resources for meaningful choices while preventing cognitive overloading of decisional capacity. At their best, control mechanisms should enable users to observe how parameter adjustments affect recommendations, fostering algorithmic literacy through experiential learning. By manipulating recommendation weights and observing outcomes, users develop shared mental models and an intuitive understanding of algorithmic logic, transforming from passive recipients to active collaborators in system behavior (Srivastava et al., 2023 ). This pedagogical approach to interface design cultivates critical engagement, rather than unconscious acceptance of algorithmic authority. Establishing Boundaries for Cognitive Preservation Just as physical fitness requires resistance training, cognitive autonomy demands regular engagement in resisting algorithmic offloading. Systems should incorporate "algorithm-free" modes for aspired users, similar to the idea of “digital detox” (Gaju, 2025 ), temporarily disabling recommendations and requiring users to navigate their internal judgment themselves. Implementation could include scheduled periods where users experience platforms through chronological or other presentation rather than personalized recommended curation. These boundaries create comparative experiences that highlight the extent of typical algorithmic influence and foster critical awareness of mediation effects. Developing Assessment Frameworks Beyond Productivity ​Current evaluation metrics focusing on engagement optimization and task efficiency fail to capture the impa​ct on human flourishing. Assessment frameworks must be expanded to encompass cognitive diversity, autonomous capability preservation, and authentic self-development. This requires longitudinal studies tracking not only immediate user satisfaction but also long-term changes in decision-making patterns as a proxy metric for intentional action capacity. Conclusion: Toward Sustainable Cognitive Futures The path forward requires a fundamental reorientation from maximizing functional capabilities for efficient decision-making to sustainable human cognitive sovereignty. This means experimenting with design constraints that maintain space for human judgment, even when full automation might prove more efficient, as a goal to shift from frictionless user experience to sustainable human-algorithm collaboration that enhances, rather than replaces, core cognitive capacities. These implications challenge industry assumptions about progress defined by ever-increasing automation and personalization in RS. Addressing these challenges requires collaboration beyond the fields of computer science and engineering. Development teams should integrate expertise from cognitive psychology, philosophy, sociology, anthropology, and other fields to assess collective behavioral shifts and develop ethics for long-term implications for human flourishing. This interdisciplinary approach ensures that technical capabilities align with human values, rather than inadvertently undermining them (Bar-Gil, 2024 ; Gigerenzer, 2024 ; Newman-Griffis, 2024 ). Declarations Author Contribution OB is the sole author of the manuscript References Abbas Y, Dervin F (eds) (2009) Digital technologies of the self. Cambridge Scholars Publ Addeo F, Delli Paoli A, Esposito M, Ylenia Bolcato M (2019) Doing Social Research on Online Communities: The Benefits of Netnography. Athens J Social Sci 7(1):9–38. https://doi.org/10.30958/ajss.7-1-1 Airoldi M, Beraldo D, Gandini A (2016) Follow the algorithm: An exploratory investigation of music on YouTube. Poetics 57:1–13. https://doi.org/10.1016/j.poetic.2016.05.001 André Q, Carmon Z, Wertenbroch K, Crum A, Frank D, Goldstein W, Huber J, van Boven L, Weber B, Yang H (2018) Consumer Choice and Autonomy in the Age of Artificial Intelligence and Big Data. Customer Needs Solutions 5(1):28–37. https://doi.org/10.1007/s40547-017-0085-8 Andrejevic M (2013) Infoglut: How Too Much Information Is Changing the Way We Think and Know. Routledge Anscombe GEM (1957) Intention. Harvard University Press Asara F (2016), September 29 Save time with smart scheduling in Google Calendar . Google. https://blog.google/products/calendar/save-time-with-smart-scheduling-in-google-calendar/ Bar-Gil O (2024) Redefining Human-Centered AI: The Human Impact of AI-based Recommendation Engines. In: Régis C, Denis J-L, Axente ML, Kishimoto A (eds) Human-Centered AI A Multidisciplinary Perspective for Policy-Makers, Auditors, and Users. Taylor & Francis, pp 34–45. https://doi.org/10.1201/9781003320791-5 Bar-Gil O (2025) Holocaust remembrance in the digital age: The transformative influence of technology, digital archives, and connective memory. Memory Stud 17506980241312341. https://doi.org/10.1177/17506980241312341 Bartl M, Kannan VK, Stockinger H (2016) A review and analysis of literature on netnography research. Int J Technol Mark 11(2):165–196. https://doi.org/10.1504/IJTMKT.2016.075687 Bratman M (1987) Intention, Plans, and Practical Reason. Center for the Study of Language and Information Braun V, Clarke V (2006) Using thematic analysis in psychology. Qualitative Res Psychol 3(2):77–101. https://doi.org/10.1191/1478088706qp063oa Buçinca Z, Malaya MB, Gajos KZ (2021) To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proceedings of the ACM on Human-Computer Interaction , 5 (CSCW1), 1–21. https://doi.org/10.1145/3449287 Carr N (2011) The Shallows: What the Internet Is Doing to Our Brains. W. W. Norton & Company Carr N (2015) The Glass Cage: How Our Computers Are Changing Us. W. W. Norton & Company Chen L, de Gemmis M, Felfernig A, Lops P, Ricci F, Semeraro G (2013) Human Decision Making and Recommender Systems. ACM Trans Interact Intell Syst 3(3):17:1–177. https://doi.org/10.1145/2533670.2533675 Christman J (2020) Autonomy in Moral and Political Philosophy. In Stanford Encyclopedia of Philosophy Archive (Fall 2020 Edition). https://plato.stanford.edu/archives/fall2020/entries/autonomy-moral/ Cinar AB, Bilodeau S (2024) Incorporating AI into the Inner Circle of Emotional Intelligence for Sustainability. Sustainability 16(15). https://doi.org/10.3390/su16156648 Clark A (2016) Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press Clark A, Chalmers DJ (1998) Ext Mind Analysis 58(1):7–19 Clowes R (2015) Thinking in the Cloud: The Cognitive Incorporation of Cloud-Based Technology. Philos Technol 28(2):261–296. https://doi.org/10.1007/s13347-014-0153-z Coeckelbergh M (2013) Human being risk: Enhancement, technology and the evaluation of vulnerability transformations. Springer del Valle JI, Lara F (2024) AI-powered recommender systems and the preservation of personal autonomy. AI Soc 39(5):2479–2491. https://doi.org/10.1007/s00146-023-01720-2 Díaz-Rodríguez N, Ser D, Coeckelbergh J, Prado M, Herrera-Viedma M, E., Herrera F (2023) Connecting the Dots in Trustworthy Artificial Intelligence: From AI Principles, Ethics, and Key Requirements to Responsible AI Systems and Regulation Doneson D (2019) The Conquest of Fortune: On the Machiavellian Character of Algorithmic Judgment. Social Research: Int Q 86(4):871–883 Duong Q (2017), November 7 Skip the line: Restaurant wait times on Search and Maps. Google . https://www.blog.google/products/maps/skip-line-restaurant-wait-times-search-and-maps/ Earl PE (2016) Bounded Rationality in the Digital Age. In R. Frantz & L. Marsh (Eds.), Minds, Models and Milieux: Commemorating the Centennial of the Birth of Herbert Simon (pp. 253–271). Palgrave Macmillan UK. https://doi.org/10.1057/9781137442505_15 Ellul J (2011) The technological society (J. Wilkinson, Trans.). Vintage books. (Original work published 1954) Feenberg A (2005) Heidegger and Marcuse: The catastrophe and redemption of history. Routledge Fink L, Newman L, Haran U (2024) Let me decide: Increasing user autonomy increases recommendation acceptance. Comput Hum Behav 156:108244. https://doi.org/10.1016/j.chb.2024.108244 Fisher E (2020a) Can Algorithmic Knowledge about the Self Be Critical? In M. Stocchetti (Ed.), The digital age and its discontents: Critical reflections in education (pp. 111–122). Helsinki University Press. https://doi.org/10.33134/HUP-4-6 Fisher E (2020b) The ledger and the diary: Algorithmic knowledge and subjectivity. Continuum 1–20. https://doi.org/10.1080/10304312.2020.1717445 Fisher E, Mehozay Y (2019) How algorithms see their audience: Media epistemes and the changing conception of the individual. Media Cult Soc. https://doi.org/10.1177/0163443719831598 Floridi L (2023) Present: AI as a New Form of Agency, Not Intelligence. In: Floridi L (ed) The Ethics of Artificial Intelligence: Principles, Challenges, and Opportunities. Oxford University Press, pp 14–30. https://doi.org/10.1093/oso/9780198883098.003.0002 Gaju A (2025) Digital Detox in the Age of Hyperconnectivity: Organizational Strategies for Mitigating Technology Overload and Enhancing Employee Flourishing. Proceedings of The International Conference on Research in Human Resource Management , 2 (1), Article 1. https://doi.org/10.33422/icrhrm.v2i1.1119 Gartenberg C (2017), July 14 Google Maps now can tell you the best time of day to travel to your destination. The Verge . https://www.theverge.com/2017/7/14/15973384/google-maps-update-travel-time-when-to-leave-directions-android Gibson JJ (1979) The Ecological Approach To Visual Perception. Psychology Gigerenzer G (2024) Psychological AI: Designing Algorithms Informed by Human Psychology. Perspect Psychol Sci 19(5):839–848. https://doi.org/10.1177/17456916231180597 Glickman M, Sharot T (2025) How human–AI feedback loops alter human perceptual, emotional and social judgements. Nat Hum Behav 9(2):345–359. https://doi.org/10.1038/s41562-024-02077-2 Goerisch F (2017), January 5 Track your New Year’s fitness goals with Google Calendar . Google. https://blog.google/products/calendar/track-your-new-years-fitness-goals-with-google-calendar/ Heater B (2018), May 9 Maps walking navigation is Google’s most compelling use for AR yet. TechCrunch . http://social.techcrunch.com/2018/05/08/maps-walking-navigation-is-googles-most-compelling-use-for-ar-yet/ Heersmink R (2017) Distributed selves: Personal identity and extended memory systems. Synthese 194(8):3135–3151. https://doi.org/10.1007/s11229-016-1102-4 Heidegger M (2008) The Question Concerning Technology (W. Lovitt, Trans.). In Basic Writings (Revised, Expanded ed. edition). Harper Perennial Modern Classics. (Original work published 1953) Hill K (2019), January 29 I Cut Google Out Of My Life. It Screwed Up Everything. Gizmodo . https://gizmodo.com/i-cut-google-out-of-my-life-it-screwed-up-everything-1830565500 Hookway N (2008) `Entering the blogosphere’: Some strategies for using blogs in social research. Qualitative Res 8(1):91–113. https://doi.org/10.1177/1468794107085298 Hoskins A (2011) 7/7 and connective memory: Interactional trajectories of remembering in post-scarcity culture. Memory Stud 4(3):269–280. https://doi.org/10.1177/1750698011402570 Hoskins A (2018) Memory of the Multitude: The End of Collective Memory. In: Hoskins A (ed) Digital Memory Studies: Media Pasts in Transition. Routledge, pp 85–109 Hoskins A (2024) AI and memory. Memory Mind & Media , 3 . https://doi.org/10.1017/mem.2024.16 Ihde D (1990) Technology and the lifeworld: From garden to earth. Indiana University Press Ihde D (2009) Postphenomenology and technoscience: The Peking University lectures. SUNY Jaspers K (1957) Man in the Modern Age. Anchor Books Karakayali N, Kostem B, Galip I (2018) Recommendation Systems as Technologies of the Self: Algorithmic Control and the Formation of Music Taste. Theory Cult Soc 35:3–24. https://doi.org/10.1177/0263276417722391 Kitchin R, Dodge M (2011) Code/space: Software and everyday life. MIT Press Køster A, Fernandez AV (2021) Investigating modes of being in the world: An introduction to Phenomenologically grounded qualitative research. Phenomenology Cogn Sci. https://doi.org/10.1007/s11097-020-09723-w Kozinets R (2015) Netnography: Redefined. SAGE Kozinets R (2019) Netnography: The Essential Guide to Qualitative Social Media Research. SAGE Kozinets R, Gambetti R (eds) (2021) Netnography unlimited: Understanding technoculture using qualitative social media research. Routledge/Taylor & Francis Group Kozinets R (2024) Researching AI chatbots, platforms and the metaverse: Understanding today’s netnography. In R. W. Belk & C. Otnes (Eds.), Handbook of Qualitative Research Methods in Marketing (pp. 185–196). Routledge. https://doi.org/10.4337/9781035302727.00025 Krippendorff K (2018) Content analysis: An introduction to its methodology. Sage Lardinois F (2018), May 8 Google Maps will soon give you better recommendations. TechCrunch . https://techcrunch.com/2018/05/08/google-maps-will-soon-give-you-better-recommendations/ Latour B (1999) Pandora’s Hope: Essays on the Reality of Science Studies. Harvard University Press Latour B (2005) Reassembling the social: An introduction to actor-network-theory. Oxford University Press Lau J (2020), September 3 Google Maps 101: How AI helps predict traffic and determine routes . Google. https://blog.google/products/maps/google-maps-101-how-ai-helps-predict-traffic-and-determine-routes/ Ledwich M, Zaitsev A (2019) Algorithmic Extremism: Examining YouTube’s Rabbit Hole of Radicalization (No. arXiv:1912.11211). arXiv. https://doi.org/10.48550/arXiv.1912.11211 Ledwich M, Zaitsev A, Laukemper A (2022) Radical bubbles on YouTube? Revisiting algorithmic extremism with personalised recommendations. First Monday. https://firstmonday.org/ojs/index.php/fm/article/view/12552 Levitt HM, Motulsky SL, Wertz FJ, Morrow SL, Ponterotto JG (2017) Recommendations for designing and reviewing qualitative research in psychology: Promoting methodological integrity. Qualitative Psychol 4(1):2–22. https://doi.org/10.1037/qup0000082 Li F, Ruijs N, Lu Y (2023) Ethics & AI: A Systematic Review on Ethical Concerns and Related Strategies for Designing with AI in Healthcare. AI 4(1). Article 1. https://doi.org/10.3390/ai4010003 Lord C (2008) Evolution of the electronic calendar: Introducing social calendaring . Citeseer Maier Z (2016), September 30 Now you can see your Google Calendar events in Google Maps. Google . https://www.blog.google/products/maps/now-you-can-see-your-google-calendar-events-google-maps/ Martini B, Bellisario D, Coletti P (2024) Human-Centered and Sustainable Artificial Intelligence in Industry 5.0: Challenges and Perspectives. Sustainability 16(13) Article 13. https://doi.org/10.3390/su16135448 McAfee A, Brynjolfsson E (2017) Machine, Platform, Crowd: Harnessing Our Digital Future. W. W. Norton & Company Menary R (ed) (2010) The extended mind. MIT Press Metz C (2016), September 27 An Infusion of AI Makes Google Translate More Powerful Than Ever. Wired . https://www.wired.com/2016/09/google-claims-ai-breakthrough-machine-translation/ Mhlanga D (2022) Human-Centered Artificial Intelligence: The Superlative Approach to Achieve Sustainable Development Goals in the Fourth Industrial Revolution. Sustainability 14(13) Article 13. https://doi.org/10.3390/su14137804 Mitcham C (1994) Thinking Through Technology: The Path Between Engineering and Philosophy. University of Chicago Press Newman-Griffis D (2024) AI Thinking: A framework for rethinking artificial intelligence in practice (No. arXiv:2409.12922). arXiv. https://doi.org/10.48550/arXiv.2409.12922 Nimmo R (2011) Actor-Network Theory and Methodology: Social Research in a More-Than-Human World. Methodological Innovations Online 6(3):108–119. https://doi.org/10.4256/mio.2011.010 Ozmen Garibay O, Winslow B, Andolina S, Antona M, Bodenschatz A, Grieman K, Jirotka M, Ten Holter CA (2023) Six human-centered artificial intelligence grand challenges. Int J Hum Comput Interact 39(3). https://doi.org/10.1080/10447318.2022.2153320 Pasquale F (2015) The Algorithmic Self. Hedgehog Rev, 30–45 Pasquale F (2016) The Black Box Society: The Secret Algorithms That Control Money and Information (Reprint edition). Harvard University Press Pham A, Rubel A, Castro C (eds) (2021) Autonomy, Agency, and Responsibility. Algorithms and Autonomy: The Ethics of Automated Decision Systems. Cambridge University Press, pp 21–42. https://doi.org/10.1017/9781108895057.002 Pichai S (2013), May 15 Live from Google I/O: Mo’ screens, mo’ goodness . The Keyword. https://www.blog.google/products/android/live-from-google-io-mo-screens-mo/ Pickering A (2010) Material Culture and the Dance of Agency. The Oxford Handbook of Material Culture Studies . https://doi.org/10.1093/oxfordhb/9780199218714.013.0007 Ramnath J (2016), April 12 Find time for your goals with Google Calendar. The Keyword . https://www.blog.google/products/calendar/find-time-goals-google-calendar/ Ricci F, Rokach L, Shapira B (2015) Recommender Systems: Introduction and Challenges. In: Ricci F, Rokach L, Shapira B (eds) Recommender Systems Handbook. Springer US, pp 1–34. https://doi.org/10.1007/978-1-4899-7637-6_1 Riedl MO (2019) Human-centered artificial intelligence and machine learning. Hum Behav Emerg Technol 1(1):33–36. https://doi.org/10.1002/hbe2.117 Risko EF, Gilbert SJ (2016) Cognitive Offloading. Trends Cogn Sci 20(9):676–688. https://doi.org/10.1016/j.tics.2016.07.002 Robert LP, Pierce C, Marquis L, Kim S, Alahmad R (2020) Designing fair AI for managing employees in organizations: A review, critique, and design agenda. Human–Computer Interact 35(5–6):545–575. https://doi.org/10.1080/07370024.2020.1735391 Rosa H (2013) Social acceleration: A new theory of modernity. Columbia University, Trans, J. Trejo-Mathys Ryan RM, Deci EL (2011) A Self-Determination Theory Perspective on Social, Institutional, Cultural, and Economic Supports for Autonomy and Their Importance for Well-Being. In V. I. Chirkov, R. M. Ryan, & K. M. Sheldon (Eds.), Human Autonomy in Cross-Cultural Context: Perspectives on the Psychology of Agency, Freedom, and Well-Being (pp. 45–64). Springer Netherlands. https://doi.org/10.1007/978-90-481-9667-8_3 Schieffer S (2016) April 27). Google Calendar for Android: Find a time for my meeting. Google Blog Searle JR (1983) Intentionality, an essay in the philosophy of mind. Cambridge University Press Searle JR (2010) Making the social world: The structure of human civilization. Oxford University Press Shneiderman B (2022) Human-centered AI. Oxford University Press Simon HA (1971) Designing organizations for an information-rich world. In M. Greenberger (Ed.), Computers, Communication, and the Public Interest (pp. 37–72). The Johns Hopkins Press. https://digitalcollections.library.cmu.edu/awweb/awarchive?type=file&item=33748 Smith MN (2017) Intentions: Past, present, future. Philosophical Explorations 20(sup2):1–12. https://doi.org/10.1080/13869795.2017.1356360 Srivastava DK, Lilly JM, Feigh KM (2023) Improving Operator Situation Awareness when Working with AI Recommender Systems (No. arXiv:2310.11370). arXiv. https://doi.org/10.48550/arXiv.2310.11370 Umapathy V (2015), December 7 Add to-dos to your Google Calendar using Reminders . The Keyword. https://blog.google/products/calendar/add-to-dos-to-your-google-calendar/ van Dijck J (2007) Mediated Memories in the Digital Age. Stanford University Press Varshney LR (2020) Respect for Human Autonomy in Recommender Systems (No. arXiv:2009.02603). arXiv. https://doi.org/10.48550/arXiv.2009.02603 Wang D, Ma X, Wang AY (2021) Human-Centered AI for Data Science: A Systematic Approach (No. arXiv:2110.01108). arXiv. https://doi.org/10.48550/arXiv.2110.01108 Wang J (2021), November 4 Google Maps navigates its way to 10 billion installs . Android Police. https://www.androidpolice.com/google-maps-navigates-its-way-to-10-billion-installs/ Ward AF (2013) One with the Cloud: Why People Mistake the Internet’s Knowledge for Their Own . https://dash.harvard.edu/handle/1/11004901 Ward AF, Duke K, Gneezy A, Bos MW (2017) Brain Drain: The Mere Presence of One’s Own Smartphone Reduces Available Cognitive Capacity. J Association Consumer Res 2(2):140–154. https://doi.org/10.1086/691462 Wellner G, Mykhailov D (2023) Caring in an Algorithmic World: Ethical Perspectives for Designers and Developers in Building AI Algorithms to Fight Fake News. Science and Engineering Ethics , 29 . https://doi.org/10.1007/s11948-023-00450-4 Wertenbroch K, Schrift RY, Alba JW, Barasch A, Bhattacharjee A, Giesler M, Knobe J, Lehmann DR, Matz S, Nave G, Parker JR, Puntoni S, Zheng Y, Zwebner Y (2020) Autonomy in consumer choice. Mark Lett 31(4):429–439. https://doi.org/10.1007/s11002-020-09521-z Woolf NH, Silver C (2017) Qualitative analysis using ATLAS.ti: The five-level QDA method. Routledge Yadav PS (2025) Cognitive Sustainability in the Age of AI: A Philosophical Framework for Understanding Competency Erosion and Cognitive Stratification in Human-AI Systems (SSRN Scholarly Paper No. 5317987). Social Science Research Network. https://doi.org/10.2139/ssrn.5317987 Zuboff S (2018) The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power . PublicAffairs Additional Declarations No competing interests reported. Supplementary Files Codebook20062023.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 19 May, 2026 Reviewers agreed at journal 18 May, 2026 Reviewers agreed at journal 08 May, 2026 Reviewers invited by journal 15 Apr, 2026 Editor assigned by journal 15 Apr, 2026 Submission checks completed at journal 15 Apr, 2026 First submitted to journal 10 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9380629","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":626681745,"identity":"4f62fd98-f144-4a2a-8151-90b9d71dade4","order_by":0,"name":"Oshri Bar-Gil","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+klEQVRIiWNgGAWjYBACfgYeBoYEhgNgzoEHFQd4IOIGFji1SDYga0k4A9cigVOLAUQNRAtDYtsBmAQeLefPHvvwgOGOvHkD+8UDifPuyBgcYH74gaEAj5YbeckzEhieGc45wFNwIHHbMx6DA2zGEngddoPHGOiXw4wz5N8kALUcBmphMMPvl/NnwFrsZzDwALXMAWlh/0bA+zlgLYkzGNgPHEhsAGnhwW+L5AyQFoNnyUBbgIF87BmP5GGeYokEPFr4+c8YM/6ouGMLtOXxhw81d+z5jrdv/PDhjw1OLVDngQgeAwiHmQEUuUQB9gfEqRsFo2AUjIIRBwA2W1YHiSMXMQAAAABJRU5ErkJggg==","orcid":"","institution":"Ruppin Academic Center","correspondingAuthor":true,"prefix":"","firstName":"Oshri","middleName":"","lastName":"Bar-Gil","suffix":""}],"badges":[],"createdAt":"2026-04-10 14:24:45","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9380629/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9380629/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107652608,"identity":"6672efc3-79f9-465b-a41a-1ef76b08d87c","added_by":"auto","created_at":"2026-04-23 15:18:19","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":89494,"visible":true,"origin":"","legend":"\u003cp\u003eBusy times (Google KeyFigurelog)\u003cstrong\u003e[1]\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[1]\u003c/strong\u003e https://blog.google/products/maps/maps101-popular-times-and-live-busyness-information/)\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/dcec51f34ccfcfa02ac3f41c.png"},{"id":107652605,"identity":"266bb113-2850-4511-a661-2f41aa8d49d2","added_by":"auto","created_at":"2026-04-23 15:18:19","extension":"gif","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":184074,"visible":true,"origin":"","legend":"\u003cp\u003eTime and scheduling recommendations (Google Keyword Blog)\u003c/p\u003e","description":"","filename":"image2.gif","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/ad9f686a4593bfc8f129f431.gif"},{"id":107707490,"identity":"9a677a45-5207-4a3c-bae2-d64337f83643","added_by":"auto","created_at":"2026-04-24 09:20:26","extension":"gif","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":707449,"visible":true,"origin":"","legend":"\u003cp\u003eGoal setting feature (Google Keyword blog)\u003c/p\u003e","description":"","filename":"image3.gif","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/13f0220f4cd194491335186c.gif"},{"id":107652607,"identity":"1a7cc7b6-beb5-4c5b-9cfe-8f0b41a09fa9","added_by":"auto","created_at":"2026-04-23 15:18:19","extension":"gif","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":818017,"visible":true,"origin":"","legend":"\u003cp\u003eGoogle \"find a time\" feature (Google Keyword Blog)\u003cstrong\u003e[2]\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[2]\u003c/strong\u003e \u0026nbsp;https://blog.google/products/calendar/google-calendar-for-android-find-time/\u003c/p\u003e","description":"","filename":"image4.gif","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/89bec161030248099ffc96b3.gif"},{"id":107709216,"identity":"aafd0e38-b670-40ab-ba15-827d2e75db97","added_by":"auto","created_at":"2026-04-24 09:35:02","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":589828,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/e2dfd6ea-7745-4594-8e47-87840d45a9e3.pdf"},{"id":107652604,"identity":"6fa0661d-3f89-4fa7-9154-298987881d54","added_by":"auto","created_at":"2026-04-23 15:18:19","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":11650,"visible":true,"origin":"","legend":"","description":"","filename":"Codebook20062023.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9380629/v1/7106a4d3f278d89eb21446bd.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Beyond Efficiency: Rethinking Human-entered AI for Cognitive Sustainability","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn recent years, the available information for decision-making has grown enormously, fueled by advances in high-speed communication networks, increased data storage capabilities, powerful algorithms, and cloud computing (McAfee \u0026amp; Brynjolfsson, \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). It put individuals in a continual influx of extensive information streams, vying for attention, and requiring rapid decision-making (Andrejevic, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). According to the German philosopher Hartmut Rosa (\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), humans enter a cycle of acceleration as the rate of technological development increases. Technological acceleration causes a hastening of lifestyle, which requires new technologies to deal with this acceleration. One technology for achieving this goal is AI-powered recommendation systems (RS). The integration of RS into digital platforms has become a crucial technological solution for alleviating the burden of information overload.\u003c/p\u003e \u003cp\u003eRecommendation systems can be defined as \u0026ldquo;software tools and techniques that provide suggestions for available items that are most likely of interest to a particular user\u0026rdquo; (Ricci et al., \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). The technical details of this area are continuously developing and are beyond the scope of this study. RS use algorithms to analyze user data and provide personalized suggestions or content \"helping\" individuals handle day-to-day decision-making from restaurant recommendations (Abbas \u0026amp; Dervin, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Duong, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lardinois, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), navigation, route selection (Heater, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), and entertainment such as video and music recommendations (Airoldi et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Karakayali et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Individuals now rely upon RS embedded in digital platforms to filter daily information and make thousands of daily decisions, with average Google users contacting platform servers more than 100,000 times daily, as the search guidance in routine decision-making regarding navigation, entertainment, and wider needs based on personalized suggestions (Hill, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In the context of Google's ecosystem, we specifically focused on the following types of recommendation systems:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eSearch Recommendations: Google's autocomplete and related search suggestions.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eContent Recommendations: YouTube's video suggestions and autoplay features.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eNavigation Recommendations: Google Maps route suggestions and place recommendations.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTemporal Recommendations: Google Calendar\u0026rsquo;s personalized time-management suggestions.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eAlthough these systems may not always be considered traditional RS in the narrowest sense, they all employ AI and machine learning to personalize and shape user decision-making based on collected personal and collective data. This broader definition allows us to examine the wider impact of RS mediated decision-making across various aspects of daily life.\u003c/p\u003e \u003cp\u003eRecent scholarship has examined how recommendation systems affect user autonomy (del Valle \u0026amp; Lara, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Fink et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Varshney, \u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) and decision-making processes (Chen et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Glickman \u0026amp; Sharot, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ricci et al., \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), with existing work primarily focusing on immediate interactions and conscious choices (Andr\u0026eacute; et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Wertenbroch et al., \u003cspan citationid=\"CR106\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). However, longitudinal changes in cognitive and memory processes through the adoption of RS remain underexplored, particularly how users' fundamental capacities for intentionality, rationality, and memory formation evolve through pervasive algorithmic mediation (Carr, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Our study addresses this gap by examining not whether autonomy is affected but how? What are the mediation mechanisms that are restructured through extended engagement with RS. We address this gap by integrating two complementary theoretical perspectives: post-phenomenological actor-network analysis of technological mediation and extended cognition theory to have a better perspective on distributed agency of humans using RS. This integrated framework helps us to explain how the tight coupling between individuals and intelligent recommendation systems produces shifts in autonomy, intention, rationality, and memory. Addressing this gap through empirical research is significant for evaluating alignment with the ideals of human-centered AI (HCAI), mainly augmenting human capabilities while maintaining control, freedom, and broader ethical considerations (Bar-Gil, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ozmen Garibay et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Shneiderman, \u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). To this end, this study explores the following four key research questions:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eHow does the use of RS affect users' autonomy, intentionality, and agency?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eHow does RS affect users' decision-making processes and rationality?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eHow does reliance on RS influence personal and collective memory patterns?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eWhat are the implications of these findings for developing human-centered AI systems?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eBy addressing these questions, we contribute to a more comprehensive understanding of HCAI to encompass psychological, social, and ethical dimensions of cognitive sustainability.\u003c/p\u003e \u003cp\u003eThis article begins by contextualizing the acceleration of information flows and the rise of RS as a societal response to information overload. The literature review critically examines how current HCAI definitions neglect the long-term effects on human cognitive capacities, proposing cognitive sustainability as an organizing principle for HCAI. It then develops a conceptual framework grounding autonomy, intentionality, and agency as distributed properties shaped by RS algorithmic mediation. Drawing on post-phenomenological theory (2009), actor-network theory (Latour, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), and extended cognition (Clark \u0026amp; Chalmers, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1998\u003c/span\u003e), it synthesizes an integrated theoretical framework for understanding how recommendation systems transform human cognition. The methodology section describes the qualitative netnography method selected and explains the selection of Google\u0026rsquo;s products and services, data collection, analysis, and research ethical considerations. The findings section is divided into three sections, each corresponding with a research question, and detailing the ways RS may affect 1. intentionality, autonomy, agency, 2. rationality, and 3. Individuals and collective memory patterns. The discussion section considers the limitations of the current research and highlights the need to redefine HCAI more comprehensively, including future research suggestions and practical implications.\u003c/p\u003e \u003cp\u003eThe contributions of this study are twofold. Theoretically, this study advances the current understanding of human-algorithm interactions. First, whereas existing research examines autonomy, whether it is preserved or diminished, we examine the mechanisms through which autonomy is mediated during sustained engagement with RS. Specifically, we trace how users progressively externalize cognitive processing by delegating decision-making functions to RS, detailing a more nuanced process than preservation or erosion, including the gradual restructuring of the rationale for autonomous choice itself. Grounding this analysis in Searle's (2010) framework of intentionality, we demonstrate how algorithmic mediation operates not simply as an external influence but as a mediator of intentions that constitute autonomous agency (del Valle \u0026amp; Lara, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Fink et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Varshney, \u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Second, we extend Hoskins's (2011, 2018, 2024) connective memory concept to the domain of platformized RS that mediate memory formation and usage through aggregated behavioral data of the individual and the collective through RS human-algorithmic mediated entanglement, restructuring memory as a scaffolding for future decision-making processes.\u003c/p\u003e \u003cp\u003ePractically, this study contributes to the reconceptualization of HCAI design as a means of cognitive sustainability by expanding its scope. Existing HCAI paradigms emphasize human control perspectives (Bar-Gil, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ozmen Garibay et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Riedl, \u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Shneiderman, \u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; D. Wang et al., \u003cspan citationid=\"CR101\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Our empirical findings suggest that sustainable HCAI should extend to encompass the preservation and even enhancement of underlying human capacities. Specifically, our findings highlight that human flourishing in increasingly RS mediated contexts requires attention to the potential long-term effects on human autonomy, intentionality, and the autonomous capacity for sense-making. This reconceptualization challenges current design trajectories and calls for the integration of cognitive sustainability as a central principle, alongside the existing emphasis on transparency and user empowerment.\u003c/p\u003e"},{"header":"Literature review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eHuman-Centered AI as a Design Program for Cognitive Sustainability?\u003c/h2\u003e \u003cp\u003eHCAI emphasizes the design of AI systems that prioritize values such as fairness, accountability, interpretability, and transparency, as well as the primacy of human needs and well-being (Bar-Gil, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Martini et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ozmen Garibay et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Riedl, \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Shneiderman, \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). In the context of HCAI, sustainability is not limited to environmental or economic outcomes but is redefined to encompass the ethical, social, and psychological dimensions of human-AI interaction (Cinar \u0026amp; Bilodeau, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Martini et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mhlanga, \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). Specifically, we adopt the concept of cognitive sustainability (Yadav, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e) as the preservation and enhancement of human cognitive capabilities, decision-making autonomy, and skill development through AI usage. While AI systems may optimize immediate task performance, sustainable design should prioritize the long-term preservation of human functioning and flourishing, ensuring that users maintain their capacity for independent judgment, even as they benefit from RS assistance (Bar-Gil, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mhlanga, \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAgency as Distributed and Emergent\u003c/h3\u003e\n\u003cp\u003eConsidering cognitive sustainability, HCAI frameworks underscore the preservation of human agency and autonomy in decision-making, which are central to self-actualization. Following Wertenbroch et al. (\u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e, p. 430), we adopt a definition of autonomy grounded in both user behavior research and psychological theory, conceptualizing autonomy as an individual’s ability to make and enact decisions independently, free from external influences imposed by other agents, such as RS. This conceptualization aligns with other philosophical notions of self-determination and free will (André et al., \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e; Pham et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e). Fink et al. (\u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e) point to the psychological basis of this definition in self-determination theory, which considers actions autonomous when they are characterized by feeling volitional or self-endorsed, where individuals feel choiceful and integrated in their behavior, fully standing behind their own actions (Ryan \u0026amp; Deci, \u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e). In the context of RS autonomy becomes particularly salient, as their suggestions constitute external influences that shape decision-making. When users interact with RS, the question arises as to whether their subsequent choices remain truly autonomous and volitional or whether algorithmic mediation compromises their capacity for autonomous decision-making. Recent scholarship has established that RS affects user autonomy (del Valle \u0026amp; Lara, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Varshney (\u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e) proposed operationalization principles for respecting human autonomy, emphasizing the need for systems that preserve user agency, and Fink et al. (\u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e) empirically demonstrated that increasing user autonomy enhances recommendation acceptance. This conceptualization provides the foundation for examining the mediated interaction with RS as influencing users' autonomy from conscious self-determination towards various degrees of diminished agency in decision-making.\u003c/p\u003e\n\u003ch3\u003eIntentionality and the Mediation of Intentions and Actions\u003c/h3\u003e\n\u003cp\u003eThus, understanding the role of human agency in mediated decision-making is crucial. Pickering (\u003cspan class=\"CitationRef\"\u003e2010\u003c/span\u003e) contends that agency can emerge from interactions between humans and non-human entities and is not solely the prerogative of human actors. He argues that agency is a relational and emergent property that arises from the entanglement of human and non-human elements and is not a fixed attribute of individual actors. His concept of “the dance of agency” captures the ongoing interplay and negotiation between different actors, including users, platforms, and algorithms, and recognizes that agency is distributed and enacted through these interactions.\u003c/p\u003e \u003cp\u003eAnother key concept in understanding and analyzing agency is intentionality. It has been defined and interpreted in various ways over the years (Anscombe, \u003cspan class=\"CitationRef\"\u003e1957\u003c/span\u003e; Searle, \u003cspan class=\"CitationRef\"\u003e2010\u003c/span\u003e; Smith, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e). According to Searle (\u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e, p. 1), intentionality is a mental state that involve being directed towards, about, or representing specific entities, events and situations. Searle (\u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e) differentiated between two types of intentions. The first type of intention is aimed at triggering action, such as forming the intention to raise one’s hand in 30 seconds. Searle refers to this as ‘prior intention.’ The second type of intention relates to the action itself, which occurs when an individual raises their hand. Searle calls this ‘intention in action.’ He argues that in every action we perform, we take for granted the social context in which we are embedded, which consists of beliefs, abilities, and possibilities, as manifested in prior intentions, which constitute an inner plan of action that directs the actions of the user. This entails visualizing the desired result and plotting the method to obtain it. The fulfillment of prior intention exemplifies the user’s logical evaluation of practicality and the actualization of potentialities.\u003c/p\u003e \u003cp\u003eRS produce new affordances for decision-making, neither exclusively human nor algorithmic, but include technological and social interactions (Gibson, \u003cspan class=\"CitationRef\"\u003e1979\u003c/span\u003e). Google’s vast databases of users’ intentions and actions, combined with networking and computing power, allow for delegating one’s decision intentions to the RS so it can produce an ‘intention in action’ complementing or sometimes enhancing one’s own intention and decision making, thus diminishing the user’s autonomy.\u003c/p\u003e\n\u003ch3\u003eRationality and Algorithmic Decision-Making\u003c/h3\u003e\n\u003cp\u003e​RS mediate the intention-to-action pathway from user decision-making intention to the users’ adoption or rejection of RS suggestions. This generates a disparity between the prior intention of the user and the suggestion generated by the RS as a suggested new prior intention. This disparity, described by Searle (\u003cspan class=\"CitationRef\"\u003e2010\u003c/span\u003e) as the \"causality gap,\" contributes to the perceived choice of the user, as it maintains one’s free will. Fisher (\u003cspan class=\"CitationRef\"\u003e2020a\u003c/span\u003e) explains that algorithmic decision-making rationality is rooted in optimization, efficiency, and data-driven decision-making. This represents a departure from human decision-making rationality to computational and statistical algorithmic models, as RS’s unique logic is shaped by patterns and biases extracted from training data. Thus, reliance on RS algorithmic rationality can affect human notions of rationality (Doneson, \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Fisher \u0026amp; Mehozay, \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eConnective Memory as Platformized Algorithmic Connectivity\u003c/h3\u003e\n\u003cp\u003e​Hoskins (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e) introduced the concept of connective memory as a framework that expands our understanding beyond personal or collective memory. His concept of connective memory has evolved from its initial focus on interpersonal digital networks to encompass broader forms of technologically mediated memory. While originally emphasizing human-to-human connections through digital platforms, recent developments in this concept support the claim that connective memory can increasingly emerge through complex assemblages of human and algorithmic actors (Bar-Gil, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e; Hoskins, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e).​\u003c/p\u003e \u003cp\u003e​In the context of RS, we extend his concept to examine how algorithmic mediation creates new patterns of individual and collective memory connectivity and their influence on ​decision making. For example, Google Maps aggregates real-time location data from millions of users to generate traffic predictions and route recommendations (Lau, \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e). It creates a form of collective memory that emerges from the aggregation of human behavior traces. Each user's navigational intentions (i.e., planning a route) and behavior (i.e., moving) contribute to a constantly updating database, as a memory that shapes future recommendations for all users and their future selves, as a form of algorithmically mediated connectivity.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eRecommendation Systems as Extended Cognition\u003c/h2\u003e \u003cp\u003eThe idea of extended cognition proposes that the human mind and cognitive processes are not solely confined to the brain or body but can encompass interactions with external elements, including technologies (Clark \u0026amp; Chalmers, \u003cspan class=\"CitationRef\"\u003e1998\u003c/span\u003e). Clark and Chalmers crucial criteria center not on the location of the cognitive process but on the functional role played in enabling an integrated cognition. They suggested that even portable or transient external artifacts, such as notebooks accessed to retrieve pivotal stored details, can become tightly coupled with cognitive processes. Building on this conceptual foundation, over two decades of subsequent inquiry by neuroscientists, psychologists, and philosophers supports the extensive entanglement of the brain, body, and world into an ensemble that produces cognition, behavior, and intelligence while avoiding commitment to strict biological localization (Clark, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Menary, \u003cspan class=\"CitationRef\"\u003e2010\u003c/span\u003e). This distributed ecological perspective contrasts with the brain as a predominantly inward data processor by highlighting the porous boundaries across humans and artifacts contingent on active environmental scaffolding (Heersmink, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eClark and Chalmers (\u003cspan class=\"CitationRef\"\u003e1998\u003c/span\u003e) seminal thought experiment contrasts two characters, Otto and Inga, who both wish to visit the Art Museum. Inga relied on her biological memory to recall the museum’s location and successfully navigated her planned route. In contrast, Otto has early stage Alzheimer’s disease; therefore, he writes down directions in a notebook to serve as an external memory aid. Otto arrives at the same destination despite his cognitive impairment by scaffolded interaction with the notebook as an external portable artifact preserving navigation details.\u003c/p\u003e \u003cp\u003eThis example delineates the core premise of their hypotheses, that cognition manifests not just ‘within the skull’ in the brain but rather emerges from coordinated engagement with environmental resources. Building on Otto and Inga, we can introduce a third character, Alex, who relies on a smartphone-based RS, Google Maps, to navigate to the museum. Alex verbally states her intention as desired destination and is presented with optimized driving, walking and public transit routes based on real-time traffic data, personal location history and stored points of interest. Turn-by-turn narration guides momentary micro-actions and navigational decision-making, while the application passively tracks progress, recalibrates suggestions if unexpected delays emerge en route, and logs a ‘memory’ of the route traveled.\u003c/p\u003e \u003cp\u003eAs Alex navigates to the museum via Google Maps, her set route is customized by saving details from previous searches and paths traveled, stored seamlessly in her Google account profile. Specific waypoints along the path are prominently shown if they are related to previous searches. Other museums can be suggested in her search results based on Google’s database or paid ad partnerships with Google, even though visiting them may divert Alex from her original intention. By leveraging Alex's prior intentions, Google aims to provide a \"personalized\" experience that caters recommendations to her perceived preferences, challenging Alex with each decision along the way.\u003c/p\u003e \u003cp\u003eThis additional scenario highlights how extensive entanglement with platformized RS, as sociotechnical artifacts, restructure not only navigation activities but also propensities, decisions, and recall capabilities distributed across the Google map service as an extended cognitive aid. The museum navigation example reveals several key distinctions in autonomy, intention, rationality, and memory patterns attributable to reliance on RS versus biological or other external cognitive resources. Most saliently, Alex delegates aspects of wayfinding to route-planning application rather than directing her internal efforts. Likewise, intention alignment may shift from self-selected intentions to those suggested algorithmically. Rationality shifts from individual bounded judgments to efficiency optimizations enacted through exhaustive, real-time data processing. Memory, as a basis for decision-making, transforms from individual constructive retention of navigation waypoints to reliance on a massive geographical database that includes users’ movement patterns to be invoked on demand.\u003c/p\u003e \u003cp\u003eWhile Alex may arrive at the museum aided by her smartphone, the layers of digital mediation affect the richness of encoding spatial memories, attentional engagement, and sense of agency over her journey. Further sections delve deeper into the changes propagated across each of these dimensions from the theoretical perspective of technical mediation theory.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePost-Phenomenology and Technical Mediation\u003c/h3\u003e\n\u003cp\u003eOver the last three decades, Contemporary philosophers of technology including Latour (\u003cspan class=\"CitationRef\"\u003e1999\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e), Ihde (\u003cspan class=\"CitationRef\"\u003e1990\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e) and Mitcham (\u003cspan class=\"CitationRef\"\u003e1994\u003c/span\u003e), have that their predecessors, thinkers like Ellul (1954/2011), Heidegger (1953/2008), and Jaspers (\u003cspan class=\"CitationRef\"\u003e1957\u003c/span\u003e), imposed a uniform model of dystopian attitudes toward ‘Technology.’ Their own perspective emphasize the function of technology as a mediator, constructing and shaping the relations between users and their environment (Coeckelbergh, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e, p. 41; Feenberg, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e; Ihde, \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe French philosopher Bruno Latour focused on the role of technology in shaping human experience through networks of human and non-human actors (or actants) interactions (Latour, \u003cspan class=\"CitationRef\"\u003e1999\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e). Applied to our context of AI-based, platformized RSs of Google, Latour’s framework might facilitate the description and interpretation of the mediation networks involving the user as a human actor and the platformized, databased, algorithmic RSs as non-human actors.\u003c/p\u003e \u003cp\u003eIhde's post-phenomenological approach emphasizes the human experience of technological mediation and how it influences human actions, perceptions, and interpretations of the world (Ihde, \u003cspan class=\"CitationRef\"\u003e1990\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e). Ihde differentiates between micro and macro perceptions to explore how human perception is shaped by technology and the environment. Micro-perception refers to the immediate sensory experience of the world, where our senses are attuned to specific stimuli. For example, microscopes extend our micro-perception by allowing us to focus on and magnify specific aspects of our environment, thus contributing to changes in macro perception. Macro-perceptions refer to a broader contextual beliefs about the world, considering larger systems, structures, and meanings. Ihde emphasizes the exploration of both micro and macro perceptions to gain a comprehensive understanding of our lived experiences and the ways in which technology mediates our perception of the world. Applied to the context of RS, Ihde's framework enables an exploration of how users interact with RS mediated decision-making, as micro-perceptions, impacting their macro- perception of intentionality, agency, and rationality as a result of those interactions.\u003c/p\u003e \u003cp\u003eBoth theories might offer a valuable lens for interpretation, but they cannot be considered methodologies in the empirical sense (Latour, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e; Nimmo, \u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e). To enable empirical examination, this study adopts a netnographic method to analyze publicly posted narratives by individuals discussing encounters with Google’s sociotechnical RS ecosystem. Post-phenomenological concepts guide the theoretical interpretations of the findings on agency, autonomy, intentionality, rationality, and memory, as detailed in the ensuing methodology overview and described in the following section.\u003c/p\u003e\n\u003ch3\u003eIntegrated Theoretical Framework\u003c/h3\u003e\n\u003cp\u003eThe theoretical perspectives presented above converge into an analytical framework that mutually reinforces and supports each theory. Technological mediation creates the conditions and offers descriptive and interpretive constructs for distributed agency, enabling the use of actor-network analysis of RS as cognitive extensions. Thus, the framework provides a comprehensive lens for analyzing how AI recommendation systems transform autonomy, agency, intentionality, rationality, and memory, which are the key dimensions we have set for examination.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003cdiv class=\"BlockQuote\"\u003e\u003c/div\u003e \u003c/div\u003e "},{"header":"Methodology","content":"\u003cp\u003eThis study employed netnography, a qualitative research method designed to study online communities and cultures (Kozinets \u0026amp; Gambetti, \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e). Netnography is a specific set of research practices related to data collection, analysis, research ethics, and presentation, rooted in participant observation (Kozinets, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Kozinets \u0026amp; Gambetti, \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e). We chose netnography for its ability to capture authentic user narratives about their evolving relationships with RS through naturally occurring digital discourses (Addeo et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Bartl et al., \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eGoogle's ecosystem was selected as the primary case study for several reasons: First, its comprehensive range of integrated services (Search, Maps, YouTube, Calendar, Assistant) enables systematic observation of how recommendation algorithms shape user behavior across different life domains. Second, the sophisticated AI algorithms of the platform for personalized RS create an ideal environment for research. Third, Google's ubiquity means that users have sustained longitudinal interactions that allow for the examination of longitudinal transformation patterns.\u003c/p\u003e\u003cp\u003eOur research design examines both individual Google services and their platform-level integration, analyzing how interconnected recommendation systems collectively transform user autonomy, rationality, and memory patterns. This holistic approach is crucial for understanding the cumulative effects of algorithmic mediation on multiple life activities.\u003c/p\u003e\u003ch2\u003eData Collection\u003c/h2\u003e\u003cp\u003eData collection focused on capturing discourse about user experiences with Google's recommendation features across the 2016–2020 period. We selected this timeframe to capture RS after achieving presence but before the emergence of generative AI, providing a focused window into a specific phase of human-algorithm interaction. We collected data from five primary sources. Four technology review blogs (The Verge, Wired, Engadget, and Ars Technica) were selected based on their comprehensive Google product coverage, significant user engagement (minimum 500,000 monthly readers), and diverse readership, which spanned casual users to technology professionals. These sources provided organic user discourse on experiences with algorithmic recommendations.\u003c/p\u003e\u003cp\u003eThe Google Keyword Blog: While acknowledging its nature as corporate communication rather than organic user review blog, Keyword provided two analytical values: (1) unmediated access to Google's design intentions and feature framing and (2) early adopter responses from power users who engaged with features as designed.\u003c/p\u003e\u003cp\u003eContent selection prioritized posts discussing user experiences with recommendation features, algorithmic decision-making, personalization systems, and behavioral adaptations to RS suggestions. This focused approach enabled the examination of how users negotiate with RS.\u003c/p\u003e\u003cp\u003eWe employed methodological triangulation (Køster \u0026amp; Fernandez, \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e; Levitt et al., \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e) by integrating additional sources: 46 books analyzing Google's ecosystem provided longitudinal context, and 25 academic studies offered theoretical frameworks and comparative baselines. Overall, our dataset comprised 525 blog posts, 46 books, and 25 academic studies (detailed in the Supplementary Materials).\u003c/p\u003e\u003ch2\u003eResearcher Immersion and Reflexive Practice\u003c/h2\u003e\u003cp\u003eFollowing the netnographic principles of researcher engagement and immersion (Kozinets, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e), the research process included immersive and participant observation within the community and with the products themselves. The lead researcher integrated Google products into his daily life, experiencing firsthand the features discussed in user discourse while maintaining a detailed reflexive journal documenting his responses to algorithmic mediation.\u003c/p\u003e\u003ch2\u003eData Analysis\u003c/h2\u003e\u003cp\u003eWe employed computer-assisted qualitative data analysis software ATLAS.Ti (Version 8) to process and analyze the large volume of collected data. This software facilitated the systematic organization and coding of textual data while maintaining analytical transparency and rigor (Krippendorff, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e; Woolf \u0026amp; Silver, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUsing Braun and Clarke's (2006) approach, we employed a three-stage analytical process that progressed from empirical observation to theoretical interpretation: First, we conducted open coding using an inductive approach to identify emerging themes and patterns in user discourse about Google services. Second, we organized the codes into higher-order themes capturing key patterns (e.g., \"hybrid agency,\" \"algorithmic rationality,\" \"connective memory\"). This involved organizing the codes into meaningful themes and exploring the relationships between them (the full codebook is available as supplemental material). Third, as theoretical interpretation, we applied our integrated framework of post-phenomenology, actor-network theory, and extended cognition to interpret how recommendation systems transform user experience.\u003c/p\u003e\u003cp\u003eFor example, Google Calendar's promise to 'help you find the time and stick to it' (Ramnath, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e) was coded in the first stage as: \u003cem\u003ecalendar; Algorithmic scheduling; Delegation of decision-making; and Temporal agency\u003c/em\u003e. In the second stage, thematic development, it was grouped with similar codes from across the dataset into the higher-order theme: \u003cem\u003e\"dance of agencies\"\u003c/em\u003e - capturing the ambiguous state where users' goals become entangled with algorithmic suggestions, creating uncertainty about the locus of agency in decision-making. In the third stage, actor-network theory analysis (Latour, \u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e) revealed that the Calendar functions as an actant that actively shapes user intentions rather than merely saving time slots for the user, as earlier calendar services did (Lord, \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e). This systematic analytical progression enabled both a detailed examination of specific mediations by product and a broader theoretical understanding of how Google’s RS transforms user experiences.\u003c/p\u003e\u003ch2\u003eResearch Ethics\u003c/h2\u003e\u003cp\u003eThe research protocol was approved by [Institution anonymized] Ethics Committee. Following digital research ethics conventions (Hookway, \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e; Kozinets, \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e), we collected only publicly accessible data. While recognizing that users might not anticipate research use, we implemented strict anonymization by removing all identifiable information from non-author commentators. To preserve authentic discourse, we maintained a non-interventionist approach, avoiding any participation that might influence natural discussions.\u003c/p\u003e"},{"header":"Findings","content":"\u003cp\u003eThe finding presented below were organized to addresses each our research questions through a systematic presentation of user discourse and it’s interpretation in each domain.\u003c/p\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eIntentionality, agency, and autonomy\u003c/h2\u003e \u003cp\u003eOur analysis of user discourse reveals specific patterns in how autonomy mediation manifests, directly addressing the research question: How does prolonged use of RS affect users' autonomy, intentionality, and agency?\u003c/p\u003e \u003cp\u003eAutonomy can be thought of as the capacity to be one’s own person, to live one’s life according to reasons and motives that are taken as one’s own and not the product of manipulative or distorting external forces, to be independent (Christman, \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; del Valle \u0026amp; Lara, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Intention plays a key role in autonomous human cognition and behavior by regulating deliberative processes tied to fulfilling purposes that give meaning to activities over time (Bratman \u003cspan class=\"CitationRef\"\u003e1987\u003c/span\u003e; Searle \u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e). When considering John Searle’s and Bruno Latour’s (\u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e) theoretical perspectives, the implications of technological mediation through the delegation of intentionality are noticeable. By delegating some of their intention of decision-making to RS, users allow the algorithm to shape not only their choices or their resulting behavioral actions but also to conform their intention itself to algorithmic rationality.\u003c/p\u003e \u003cp\u003eThe Google Maps navigation tool continually optimizes routes as RS based on external data analysis for over 10\u0026nbsp;billion devices (J. Wang, \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e), including various features aimed to assist the user in decision making, such as “ popular times to avoid” graph (Duong, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e), waiting time estimates that can influence the intention to select and visit a particular place in a particular time, when seeing that the place is busier than usual (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eYouTube's recommendation algorithm exemplifies the profound transformation of user agency and intentionality through algorithmic mediation. The platform's autoplay feature doesn't merely suggest content, but it actively reconstructs viewing intentions by creating seamless transitions between videos that users never consciously selected. Beyond mere mediation of intentionality, this represents what Karakayali et al. (\u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e) identify as the platform's prioritization of engagement metrics and advertiser interests over users' organic viewing intentions.\u003c/p\u003e \u003cp\u003eThe algorithm's influence on intentionality operates through multiple mechanisms. First, it exploits the prior-intention to intention-in-action gap (Searle, \u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e) by automatically loading new content before users can exercise a deliberate choice. Second, it progressively learns and shapes user preferences through a feedback loop in which each viewed video (whether actively chosen or passively accepted) influences future recommendations, gradually steering users toward content that maximizes platform engagement rather than satisfying their original viewing intention. Third, the system creates what users describe as 'rabbit holes', trajectories of increasingly specific or extreme content that led far from the original intentions (Ledwich et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLedwich and Zaitsev (\u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e) found that YouTube's recommendation system can lead to harmful consequences such as online radicalization. This represents not only a failure of content moderation but also a fundamental challenge to user autonomy: the algorithm does not simply respond to existing intentions and preferences but actively constructs new ones, potentially transforming users' political orientations and worldviews without their awareness or consent. Users' agency becomes entangled with algorithmic processes designed to maximize watch time, mediating intention, where the distinction between what users want to watch and what the algorithm wants them to watch becomes increasingly blurred.\u003c/p\u003e \u003cp\u003eAcross contexts, RS leveraging behavioral data and predictive analytics for personalization to subtly align intentions more with platform priorities than support users’ agency in their sense-making, thereby influencing their decisions (Clowes, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e; Kitchin \u0026amp; Dodge, \u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e). Google advertises the hyperpersonalization of map services as follows:\u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cem\u003e“What if we told you that during your lifetime, Google could create millions of custom maps...each one just for you?\u003c/em\u003e \u003c/p\u003e\u003cp\u003e \u003cem\u003eIn the past, such a notion would have been unbelievable: a map was just a map, and you got the same one for New York City, whether you were searching for the Empire State Building or a coffee shop on the street. What if, instead, you had a map that is unique to you, always adapting to the task you want to perform right this minute?“\u003c/em\u003e (Pichai, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e \u003cp\u003eAnother example is Google Calendar's 'Goals' feature, which automatically schedules time for user-defined objectives (exercise, reading, learning, etc.). Users specify what they want to achieve and their preferred frequency; the algorithm then analyzes their existing calendar, identifies free time slots, and automatically schedules goal sessions (see Figs.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e,\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). The system adapts when users defer or complete sessions and learns optimal timing patterns:\u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cem\u003e“Whether it is reading more books, learning a new language, or working out regularly, achieving your goals can be difficult. One day it's \"I got called into a last-minute meeting.\" The next day it's \"I have a friend in town.\" Before you know it, your goals are delayed or forgotten. In fact, with all the things you need to do in a given week, it is probably harder than ever to find the time—even when your goal really matters to you.\u003c/em\u003e \u003c/p\u003e\u003cp\u003e \u003cem\u003eThat’s why starting today, we are introducing Goals in Google Calendar. Just add a personal goal—like “run 3 times a week”—and Calendar will help you find the time and stick to it.”\u003c/em\u003e (Ramnath, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAn additional feature in Google’s calendar even nudges you to complete your tasks:\u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cem\u003e“You might already create calendar entries to remind you to call the doctor or pick up groceries on the way home. But while those entries come and go, Reminders stick with you over time so you can track them until they are actually done. If a Reminder isn't completed, it will appear at the top of your Calendar the next day. And the next. When you do finally call the doctor or pick up those necessities”\u003c/em\u003e (Umapathy, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e \u003cp\u003eGoogle Calendar represents a shift from a passive calendar to an active temporal mediation agent, shaping user behavior through algorithmic scheduling and execution through nudges and reminders. Generalizing for different Google RS, we can see that users increasingly delegate parts of their decision-making process to algorithmic based RS: Temporal aspects such as automated meeting scheduling (Asara, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Schieffer, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e) and optimization of time slots for achieving personal goals (Goerisch, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e) to Google Calendar, spatial aspects of travel planning that use machine learning and connective memory of traffic patterns in real time to Google Maps (Gartenberg, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lau, \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Maier, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). Seeking frictionless, optimal time and space experiences from users can lead to dependence on external recommendation algorithms that might hinder the capacities for autonomous decision-making. One way of understanding these user choices, which diminish their agency, is to consider a potential change in the rationality of the perceived decision-making using RS.\u003c/p\u003e \u003c/div\u003e\u003ch2\u003eIntentionality, agency, and autonomy\u003c/h2\u003e\u003cp\u003eOur analysis of user discourse reveals specific patterns in how autonomy mediation manifests, directly addressing the research question: How does prolonged use of RS affect users' autonomy, intentionality, and agency?\u003c/p\u003e\u003cp\u003eAutonomy can be thought of as the capacity to be one’s own person, to live one’s life according to reasons and motives that are taken as one’s own and not the product of manipulative or distorting external forces, to be independent (Christman, \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; del Valle \u0026amp; Lara, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Intention plays a key role in autonomous human cognition and behavior by regulating deliberative processes tied to fulfilling purposes that give meaning to activities over time (Bratman \u003cspan class=\"CitationRef\"\u003e1987\u003c/span\u003e; Searle \u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e). When considering John Searle’s and Bruno Latour’s (\u003cspan class=\"CitationRef\"\u003e2005\u003c/span\u003e) theoretical perspectives, the implications of technological mediation through the delegation of intentionality are noticeable. By delegating some of their intention of decision-making to RS, users allow the algorithm to shape not only their choices or their resulting behavioral actions but also to conform their intention itself to algorithmic rationality.\u003c/p\u003e\u003cp\u003eThe Google Maps navigation tool continually optimizes routes as RS based on external data analysis for over 10\u0026nbsp;billion devices (J. Wang, \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e), including various features aimed to assist the user in decision making, such as “ popular times to avoid” graph (Duong, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e), waiting time estimates that can influence the intention to select and visit a particular place in a particular time, when seeing that the place is busier than usual (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cp\u003eYouTube's recommendation algorithm exemplifies the profound transformation of user agency and intentionality through algorithmic mediation. The platform's autoplay feature doesn't merely suggest content, but it actively reconstructs viewing intentions by creating seamless transitions between videos that users never consciously selected. Beyond mere mediation of intentionality, this represents what Karakayali et al. (\u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e) identify as the platform's prioritization of engagement metrics and advertiser interests over users' organic viewing intentions.\u003c/p\u003e\u003cp\u003eThe algorithm's influence on intentionality operates through multiple mechanisms. First, it exploits the prior-intention to intention-in-action gap (Searle, \u003cspan class=\"CitationRef\"\u003e1983\u003c/span\u003e) by automatically loading new content before users can exercise a deliberate choice. Second, it progressively learns and shapes user preferences through a feedback loop in which each viewed video (whether actively chosen or passively accepted) influences future recommendations, gradually steering users toward content that maximizes platform engagement rather than satisfying their original viewing intention. Third, the system creates what users describe as 'rabbit holes', trajectories of increasingly specific or extreme content that led far from the original intentions (Ledwich et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eLedwich and Zaitsev (\u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e) found that YouTube's recommendation system can lead to harmful consequences such as online radicalization. This represents not only a failure of content moderation but also a fundamental challenge to user autonomy: the algorithm does not simply respond to existing intentions and preferences but actively constructs new ones, potentially transforming users' political orientations and worldviews without their awareness or consent. Users' agency becomes entangled with algorithmic processes designed to maximize watch time, mediating intention, where the distinction between what users want to watch and what the algorithm wants them to watch becomes increasingly blurred.\u003c/p\u003e\u003cp\u003eAcross contexts, RS leveraging behavioral data and predictive analytics for personalization to subtly align intentions more with platform priorities than support users’ agency in their sense-making, thereby influencing their decisions (Clowes, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e; Kitchin \u0026amp; Dodge, \u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e). Google advertises the hyperpersonalization of map services as follows:\u003c/p\u003e\u003cp\u003e \u003cem\u003e“What if we told you that during your lifetime, Google could create millions of custom maps...each one just for you?\u003c/em\u003e \u003c/p\u003e\u003cp\u003e \u003cem\u003eIn the past, such a notion would have been unbelievable: a map was just a map, and you got the same one for New York City, whether you were searching for the Empire State Building or a coffee shop on the street. What if, instead, you had a map that is unique to you, always adapting to the task you want to perform right this minute?“\u003c/em\u003e (Pichai, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAnother example is Google Calendar's 'Goals' feature, which automatically schedules time for user-defined objectives (exercise, reading, learning, etc.). Users specify what they want to achieve and their preferred frequency; the algorithm then analyzes their existing calendar, identifies free time slots, and automatically schedules goal sessions (see Figs.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e,\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). The system adapts when users defer or complete sessions and learns optimal timing patterns:\u003c/p\u003e\u003cp\u003e \u003cem\u003e“Whether it is reading more books, learning a new language, or working out regularly, achieving your goals can be difficult. One day it's \"I got called into a last-minute meeting.\" The next day it's \"I have a friend in town.\" Before you know it, your goals are delayed or forgotten. In fact, with all the things you need to do in a given week, it is probably harder than ever to find the time—even when your goal really matters to you.\u003c/em\u003e \u003c/p\u003e\u003cp\u003e \u003cem\u003eThat’s why starting today, we are introducing Goals in Google Calendar. Just add a personal goal—like “run 3 times a week”—and Calendar will help you find the time and stick to it.”\u003c/em\u003e (Ramnath, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cp\u003eAn additional feature in Google’s calendar even nudges you to complete your tasks:\u003c/p\u003e\u003cp\u003e \u003cem\u003e“You might already create calendar entries to remind you to call the doctor or pick up groceries on the way home. But while those entries come and go, Reminders stick with you over time so you can track them until they are actually done. If a Reminder isn't completed, it will appear at the top of your Calendar the next day. And the next. When you do finally call the doctor or pick up those necessities”\u003c/em\u003e (Umapathy, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGoogle Calendar represents a shift from a passive calendar to an active temporal mediation agent, shaping user behavior through algorithmic scheduling and execution through nudges and reminders. Generalizing for different Google RS, we can see that users increasingly delegate parts of their decision-making process to algorithmic based RS: Temporal aspects such as automated meeting scheduling (Asara, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Schieffer, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e) and optimization of time slots for achieving personal goals (Goerisch, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e) to Google Calendar, spatial aspects of travel planning that use machine learning and connective memory of traffic patterns in real time to Google Maps (Gartenberg, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lau, \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Maier, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). Seeking frictionless, optimal time and space experiences from users can lead to dependence on external recommendation algorithms that might hinder the capacities for autonomous decision-making. One way of understanding these user choices, which diminish their agency, is to consider a potential change in the rationality of the perceived decision-making using RS.\u003c/p\u003e\u003ch2\u003eRationality\u003c/h2\u003e\u003cp\u003eRational decision-making in human cognition operates within natural constraints—limitations in memory, attention, and cognitive resources, termed by Herbert Simon (\u003cspan class=\"CitationRef\"\u003e1971\u003c/span\u003e) as \"bounded rationality.\" These inherent restrictions make algorithmic RS particularly appealing, as they promise to transcend human cognitive limitations through superior processing speed, vast data integration and computational optimization (Earl, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). However, the promise of algorithmic rationality represents more than simple augmentation; it fundamentally restructures how intentions are transformed into actions and how users conceptualize rational decision-making itself.\u003c/p\u003e\u003cp\u003eFollowing Searle's (1983) framework of intentionality, with its division of prior intentions and intentions in actions, and Bratman's (1987) elaboration of prior intention as an internal plan with bounded flexibility, we can understand how algorithms mediate the intention-to-action pathway. When cognitive resources are limited, as Simon noted, the success of prior intentions depends on efficient resource allocation for planning and execution processes. From this efficiency perspective, delegating larger parts of decision-making to external algorithmic systems appears not just convenient but rational choice.\u003c/p\u003e\u003cp\u003eWhen actualizing intentions becomes a computational problem, Google's RS algorithms provide optimal pathways from intention to action. The Calendar's \"Find a time\" feature exemplifies this (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e):\u003c/p\u003e\u003cp\u003e \u003cem\u003e“With a single tap, ‘Find a time’ helps you find meeting times that work for everyone—even if they're in different time zones—based on their availability and the times they usually have meetings. If there are no times that work, Calendar will look at which conflicting meetings can most easily be rescheduled.”\u003c/em\u003e (Schieffer, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cp\u003eUsers perceive this algorithmic logic as inherently more rational because it operates through systematic and consistent rules, which is the foundation of how algorithms function. By incorporating this algorithmic logic into their decision-making processes, users feel more rational, efficiently bridging the gap between abstract intentions and completed actions. Hence, we can notice that users do not merely use RS for efficiency; they internalize algorithmic reasoning patterns. This represents what Fisher and Mehozay (\u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e) identified as algorithmic rationality colonizing human judgment; however, our data hint that users actively embrace it, viewing it as a helpful enhancement rather than a logic replacement.\u003c/p\u003e\u003cp\u003eMore profoundly, these RS act as if they know users' \"true\" intentions better than users know themselves. This algorithmic epistemology suggests that by analyzing vast behavioral data on what users actually watch, where they actually go, and what they actually search for, algorithms can identify \"authentic\" preferences that transcend conscious self-reports. If Google's aggregated data about a user provides superior self-knowledge, what could be more rational than allowing this algorithmic representation to mediate intentions, determine navigation paths, schedule time slots, and more for the actual self?\u003c/p\u003e\u003cp\u003eThis transformation is evident in Google Calendar's approach to time management, where the system does not just execute user commands but actively shapes temporal decisions, as discussed earlier. The calendar’s algorithmic system determines not only when meetings occur, but also which commitments take priority.\u003c/p\u003e\u003cp\u003eSignificantly, Google maintains the illusion of user control even as it shapes decisions: \"'Find a time' makes suggestions, but you're still in control\" (Schieffer, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). This rhetoric echoes the findings of Varshney (\u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e) and Fink et al. (\u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e) that perceived autonomy increases recommendation acceptance; paradoxically, it should be predicted to lead to a higher acceptance rate of its recommendations.\u003c/p\u003e\u003cp\u003eThese implications extend beyond specific decisions. These optimized routes, viewing sequences, and scheduling patterns reflect platform priorities – maximizing engagement, advertising exposure, and data collection rather than supporting users' organic self-development or contextual meaning-making (Fisher, \u003cspan class=\"CitationRef\"\u003e2020a\u003c/span\u003e; Zuboff, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e). Through this creeping delegation process, users transfer increasing portions of their intentional apparatus to RS, despite the risk of eroding the boundaries between their self and technological others. It seems that personalized RS, operated by Google and aligned with corporate intentions, progressively colonize the space of human intention formation. Users increasingly evaluate their decision-making against algorithmic standards, judge their choices using optimization metrics, and doubt their intuitions when they conflict with data-driven recommendations. The question is not whether this is rational, but what definitions of rationality govern human action when interacting with RS.\u003c/p\u003e\u003ch2\u003eImpact on Memory\u003c/h2\u003e\u003cp\u003eTo address RQ3, which explores how reliance on RS influences personal and collective memory patterns, we examined how users use Google RS as a mediating interface for external connective memory. Recent research examining shifts in memory patterns with technological mediation found the potential for collective memory to emerge through users’ digital behaviors. The increased use of digitally mediated social connections has given rise to a “connective” memory paradigm, facilitating collaborative memory storage, recall, and sensemaking (Hoskins, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eOur analysis shows that RS create a variation of Hoskins’ connective memory concept, where connectivity emerges not just through platformized human interaction but through algorithmic aggregation, analysis, and reinterpretations of user behaviors and behavior intentions. Consider Google Maps: users’ intention to arrive at a particular destination is manifested by asking for recommended routes and timing, the recommendation stems from their own and collective travel times, and the chosen paths themselves become part of the system's collective dataset, influencing future recommendations for other users. A traffic slowdown experienced by one user immediately affects route suggestions for others, creating an indirect form of collective memory mediated by algorithms. Rather than individual recollections, pieces of fragmentary traces, together with cloud-based crowd-sourced scaffolds, form the connective database for personalized decisions for each user. While Hoskins's (2011) original formulation of connective memory emphasized direct interpersonal connections through digital platforms, our findings follow his later conceptualization of collective algorithmic memory mediation beyond the realm of individual human influence and control (Hoskins, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). We see that RS create de facto data connections, even if they are indirect and mediated by platform databases. When YouTube's algorithm recommends videos based on viewing patterns of similar users, it creates a form of collective influence where users shape each other's experiences without direct interaction, as “connectivity” in RS operates through aggregation and pattern recognition across users. Google's RS does not just respond to individual preferences but analyzes collective behavioral patterns, making each user's actions part of a larger social dataset. Thus, our findings support the concept of connective memory as a pattern of memory emerging from increased connectivity and entanglements with others in real time through an assortment of digital apps, platforms, and networks. Personal contributions to these databases concurrently extract costs by diminishing the motivation for dedicated encoding, storage, and retrieval, which are practiced through cognitive processes traditionally linked to human development.\u003c/p\u003e\u003cp\u003eOffloading personal memory onto networks of memorial intermediaries for recall in future decision-making, available on demand across domains such as navigation and entertainment, allows promises of superior memory performance by the platforms but can come with personal and individual costs (Fisher, \u003cspan class=\"CitationRef\"\u003e2020b\u003c/span\u003e; Risko \u0026amp; Gilbert, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Ward, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Ward et al., \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e). Personalization-induced automation and a lack of transparency around inferential affordances risk misdirecting memory and decision-making (Pasquale, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). Delegating personal memory might also threaten complicity rather than the experience and intuition gained by consciously situating intentions within narratives (Hoskins, \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e; van Dijck, \u003cspan class=\"CitationRef\"\u003e2007\u003c/span\u003e).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eDeeply Human-Centered AI for Sustainability\u003c/h2\u003e \u003cp\u003eWhile industry roadmaps currently position progress largely in terms of surpassing functionality milestones around accuracy, speed, scale, and robustness (Metz, \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), this study indicates that such trajectories might have unintended consequences, such as undermining human well-being by disrupting cognitive sustainability. Analysis across the interrelated dimensions of intentionality, agency, rationality, and memory suggests that the use of RS triggers effects on users that progressively diminish facets of decision-making autonomy over time as the dependency and complexity of decision-making increase (Earl, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Pham et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Allowing behavioral data accumulation and predictive nudging to shape intentions, constraining cognitive overload coping strategies to prescribed routines, and restructuring memory patterns around connective databases optimized for algorithmic usability rather than personal ownership risks changing the weights of agencies in decision-making over time. Further research is vital for detailing the boundaries of what sorts of augmentation remain compatible with the notions of human thriving tied to the directed development of self-knowledge, context-dependent rationality, and memory anchoring humans as individuals and humanity as a whole, as can be expected from aligning RS to be more \u0026ldquo;human-centric.\u0026rdquo;\u003c/p\u003e \u003cp\u003eRevisiting widely cited definitions of HCAI concentrate predominantly on preserving overt control, transparency, and human values validation (Bar-Gil, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mhlanga, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Shneiderman, \u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Recentering progress on multifaceted, situated human capabilities, in the context of pervasive RS usage compels expanding frameworks to continuously monitor the broader psychological, political, and social dimensions that are at risk of unintended erosion under increased technological immersion mediating ever growing life aspects (Coeckelbergh, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; D\u0026iacute;az-Rodr\u0026iacute;guez et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Design approaches must be grounded in recognizing humans as growing, intentional actors recursively shaping digital environments that likewise profoundly shape emerging skills, relationships, and self-concept in a reciprocal fashion, rather than one-way efficiency improving delivery mechanisms (Li et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Robert et al., \u003cspan citationid=\"CR88\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Wellner \u0026amp; Mykhailov, \u003cspan citationid=\"CR105\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eLimitations and future research\u003c/h2\u003e \u003cp\u003eThis study faced several constraints, some of which were inherent to netnographic research on rapidly evolving technological systems. The decision to focus exclusively on Google's ecosystem provided analytical depth but limited generalizability. Google's specific design and monetization logics, particularly its emphasis on predictive automation and comprehensive data integration across platform services, may produce distinct patterns of cognitive delegation that are not replicated on other platforms. For example, Facebook's social graph-based recommendations or Amazon's commerce-driven algorithmic recommendations likely generate different forms of user behavior, autonomy disruption, and memory externalization.\u003c/p\u003e \u003cp\u003eReliance on publicly posted user discourse introduces selection bias towards users motivated to articulate their experience online. Users who seamlessly integrate with algorithmic systems or even those who resist them but without public commentary remain invisible in our data. Moreover, technology blog comments attract technology reflective users who may not represent typical usage patterns. The inclusion of Google's Keyword blog introduced a specific methodological tension, unlike independent technology blogs, where users freely critique and discuss their experiences. The Keyword represents Google's curated narrative from developers, product managers, and designers as early adopters and power users whose experiences may anticipate but not reflect mainstream adoption. Another limitation of the data collection is the selection of only English-language data, as it limits cultural diversity and might miss how different linguistic structures and cultural concepts of self, memory, and agency shape human-algorithm relations. For example, Mandarin-speaking users navigating localized Google services may experience fundamentally different techno-cultural affordances.\u003c/p\u003e \u003cp\u003eReflecting on the temporal scope of our data collection, spanned 2016\u0026ndash;2020, we can see that it represents a critical period when platformized RS became ubiquitous in daily life. While our data predates recent AI developments such as ChatGPT and other generative models, the patterns we identified\u0026ndash;delegation of cognitive functions, shifts in personal agency, and transformation of memory practices\u0026ndash;appear to be intensifying rather than diminishing with newer technologies. The theoretical framework we developed for understanding human-AI interaction through the lenses of autonomy, intentionality, and rationality provides enduring analytical value. These psychological and philosophical dimensions of human experience remain constant, even as the technical sophistication of AI systems evolves, zs Shneiderman (\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) notes, the human-centered challenges of AI transcend specific, technical implementations. Nevertheless, this temporal boundary constitutes a significant limitation. Recent developments, particularly in conversational AI and more sophisticated personalization algorithms, may introduce new forms of human-AI interaction that are not captured in our analysis. Future research should examine whether the patterns of cognitive delegation, intentionality shifts, and memory externalization that we document have intensified or evolved with newer AI systems.\u003c/p\u003e \u003cp\u003eTheoretical choices shape what becomes visible or invisible in our analysis. Our theoretical framework, which integrates post-phenomenology, ANT, and extended cognition may obscure alternative interpretations. For example a critical theory lens might reveal the power dynamics we choose to leave out of the scope of this article, and a neuroscience perspective could challenge our assumptions about cognitive restructuring.\u003c/p\u003e \u003cp\u003eFollowing these limitations, several research directions emerge: longitudinal studies tracking individual users over time could reveal whether the patterns we identified stabilize, intensify, or even reverse over time with changes in adoption patterns and the emergence of new technological affordances. It will be necessary to determine whether users develop resistance strategies or whether cognitive adaptation reaches equilibrium. Methodologically, combining netnography with cognitive testing could validate whether self-reported experiences of memory externalization correspond to measurable changes in recall or cognitive function. Similarly, experimental interventions temporarily removing algorithmic recommendations could test the dependency effects we theorized as limiting the autonomy of individual users. Cross-cultural analysis should examine not only different user populations but also how varying concepts of autonomy, self, and cognition across cultures shape human-algorithm coupling. For example, does Western emphasis on individual autonomy make users more vulnerable to certain algorithmic influences?\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eImplications for Human-Centered AI Development: Prioritizing Cognitive Empowerment\u003c/h2\u003e \u003cp\u003eOur findings reveal tensions between current AI development trajectories that prioritize functional expansion, as adopted by users, and the preservation of human cognitive sustainability (Yadav, \u003cspan citationid=\"CR108\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). While prevailing approaches emphasize efficiency gains and frictionless user experiences, our analysis demonstrates how these design choices might erode fundamental human capacities for autonomous decision-making, intentional action, and personal memory formation. These implications necessitate reconceptualizing HCAI development to include cognitive empowerment (Bar-Gil, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ozmen Garibay et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003ePreserving Autonomy Through Design Transparency\u003c/h2\u003e \u003cp\u003eThe progressive delegation of decision-making to RS requires systematic interventions to maintain user awareness of mediation processes. RS should implement reflexivity mechanisms that make algorithmic influence visible rather than seamless. One potential suggestion would be to develop transparency utilities that track users' reliance patterns on algorithmic recommendations across temporal scales, enabling the recognition of creeping dependency before it becomes constitutive. Such metrics would illuminate when external recommendations begin to colonize internal judgment processes, for example, by considering the recommendation adoption rate or the reaction time for recommendations allowing transparent presentation for reflection. Increasing the transparency of recommendation patterns and adoption was found to facilitate a shared mental model (Srivastava et al., \u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), in itself an intervention that supports cognitive sustainability.\u003c/p\u003e \u003cp\u003e​Design implementations should include clear indicators distinguishing algorithmic suggestions from organic content discovery, with options to experience platforms without personalization filters. This transparency offer extends beyond simple labeling to include explanatory interfaces that reveal how user data shapes recommendations and makes visible the feedback loops between past behavior and future suggestions. Users need comprehension not just of what is recommended but why specific suggestions emerge from their past behavioral or connective patterns.​\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eImplementing Autonomy Controls\u003c/h2\u003e \u003cp\u003eRather than resisting recommendations by users, systems can provide graduated controls, allowing users who would like a higher degree of control to modulate algorithmic influence across different decision domains. Bu\u0026ccedil;inca et al. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) called these interventions as Cognitive Forcing Functions and found that they can reduce overreliance on AI suggestions in decision support systems. Additionally, high-stakes decisions requiring contextual judgment could maintain human primacy, whereas routine tasks employ greater automation. This selective delegation control preserves cognitive resources for meaningful choices while preventing cognitive overloading of decisional capacity.\u003c/p\u003e \u003cp\u003eAt their best, control mechanisms should enable users to observe how parameter adjustments affect recommendations, fostering algorithmic literacy through experiential learning. By manipulating recommendation weights and observing outcomes, users develop shared mental models and an intuitive understanding of algorithmic logic, transforming from passive recipients to active collaborators in system behavior (Srivastava et al., \u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This pedagogical approach to interface design cultivates critical engagement, rather than unconscious acceptance of algorithmic authority.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eEstablishing Boundaries for Cognitive Preservation\u003c/h2\u003e \u003cp\u003eJust as physical fitness requires resistance training, cognitive autonomy demands regular engagement in resisting algorithmic offloading. Systems should incorporate \"algorithm-free\" modes for aspired users, similar to the idea of \u0026ldquo;digital detox\u0026rdquo; (Gaju, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), temporarily disabling recommendations and requiring users to navigate their internal judgment themselves. Implementation could include scheduled periods where users experience platforms through chronological or other presentation rather than personalized recommended curation. These boundaries create comparative experiences that highlight the extent of typical algorithmic influence and foster critical awareness of mediation effects.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eDeveloping Assessment Frameworks Beyond Productivity\u003c/h2\u003e \u003cp\u003e​Current evaluation metrics focusing on engagement optimization and task efficiency fail to capture the impa​ct on human flourishing. Assessment frameworks must be expanded to encompass cognitive diversity, autonomous capability preservation, and authentic self-development. This requires longitudinal studies tracking not only immediate user satisfaction but also long-term changes in decision-making patterns as a proxy metric for intentional action capacity.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eConclusion: Toward Sustainable Cognitive Futures\u003c/h2\u003e \u003cp\u003eThe path forward requires a fundamental reorientation from maximizing functional capabilities for efficient decision-making to sustainable human cognitive sovereignty. This means experimenting with design constraints that maintain space for human judgment, even when full automation might prove more efficient, as a goal to shift from frictionless user experience to sustainable human-algorithm collaboration that enhances, rather than replaces, core cognitive capacities. These implications challenge industry assumptions about progress defined by ever-increasing automation and personalization in RS.\u003c/p\u003e \u003cp\u003eAddressing these challenges requires collaboration beyond the fields of computer science and engineering. Development teams should integrate expertise from cognitive psychology, philosophy, sociology, anthropology, and other fields to assess collective behavioral shifts and develop ethics for long-term implications for human flourishing. This interdisciplinary approach ensures that technical capabilities align with human values, rather than inadvertently undermining them (Bar-Gil, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Gigerenzer, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Newman-Griffis, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eOB is the sole author of the manuscript\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbbas Y, Dervin F (eds) (2009) Digital technologies of the self. Cambridge Scholars Publ\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAddeo F, Delli Paoli A, Esposito M, Ylenia Bolcato M (2019) Doing Social Research on Online Communities: The Benefits of Netnography. Athens J Social Sci 7(1):9\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.30958/ajss.7-1-1\u003c/span\u003e\u003cspan address=\"10.30958/ajss.7-1-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAiroldi M, Beraldo D, Gandini A (2016) Follow the algorithm: An exploratory investigation of music on YouTube. Poetics 57:1\u0026ndash;13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.poetic.2016.05.001\u003c/span\u003e\u003cspan address=\"10.1016/j.poetic.2016.05.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndr\u0026eacute; Q, Carmon Z, Wertenbroch K, Crum A, Frank D, Goldstein W, Huber J, van Boven L, Weber B, Yang H (2018) Consumer Choice and Autonomy in the Age of Artificial Intelligence and Big Data. Customer Needs Solutions 5(1):28\u0026ndash;37. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40547-017-0085-8\u003c/span\u003e\u003cspan address=\"10.1007/s40547-017-0085-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndrejevic M (2013) Infoglut: How Too Much Information Is Changing the Way We Think and Know. Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnscombe GEM (1957) Intention. Harvard University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsara F (2016), September 29 \u003cem\u003eSave time with smart scheduling in Google Calendar\u003c/em\u003e. Google. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://blog.google/products/calendar/save-time-with-smart-scheduling-in-google-calendar/\u003c/span\u003e\u003cspan address=\"https://blog.google/products/calendar/save-time-with-smart-scheduling-in-google-calendar/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBar-Gil O (2024) Redefining Human-Centered AI: The Human Impact of AI-based Recommendation Engines. In: R\u0026eacute;gis C, Denis J-L, Axente ML, Kishimoto A (eds) Human-Centered AI A Multidisciplinary Perspective for Policy-Makers, Auditors, and Users. Taylor \u0026amp; Francis, pp 34\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1201/9781003320791-5\u003c/span\u003e\u003cspan address=\"10.1201/9781003320791-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBar-Gil O (2025) Holocaust remembrance in the digital age: The transformative influence of technology, digital archives, and connective memory. Memory Stud 17506980241312341. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/17506980241312341\u003c/span\u003e\u003cspan address=\"10.1177/17506980241312341\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBartl M, Kannan VK, Stockinger H (2016) A review and analysis of literature on netnography research. Int J Technol Mark 11(2):165\u0026ndash;196. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1504/IJTMKT.2016.075687\u003c/span\u003e\u003cspan address=\"10.1504/IJTMKT.2016.075687\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBratman M (1987) Intention, Plans, and Practical Reason. Center for the Study of Language and Information\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraun V, Clarke V (2006) Using thematic analysis in psychology. Qualitative Res Psychol 3(2):77\u0026ndash;101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1191/1478088706qp063oa\u003c/span\u003e\u003cspan address=\"10.1191/1478088706qp063oa\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBu\u0026ccedil;inca Z, Malaya MB, Gajos KZ (2021) To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. \u003cem\u003eProceedings of the ACM on Human-Computer Interaction\u003c/em\u003e, \u003cem\u003e5\u003c/em\u003e(CSCW1), 1\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3449287\u003c/span\u003e\u003cspan address=\"10.1145/3449287\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarr N (2011) The Shallows: What the Internet Is Doing to Our Brains. W. W. Norton \u0026amp; Company\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarr N (2015) The Glass Cage: How Our Computers Are Changing Us. W. W. Norton \u0026amp; Company\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen L, de Gemmis M, Felfernig A, Lops P, Ricci F, Semeraro G (2013) Human Decision Making and Recommender Systems. ACM Trans Interact Intell Syst 3(3):17:1\u0026ndash;177. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/2533670.2533675\u003c/span\u003e\u003cspan address=\"10.1145/2533670.2533675\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChristman J (2020) Autonomy in Moral and Political Philosophy. In \u003cem\u003eStanford Encyclopedia of Philosophy Archive\u003c/em\u003e (Fall 2020 Edition). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://plato.stanford.edu/archives/fall2020/entries/autonomy-moral/\u003c/span\u003e\u003cspan address=\"https://plato.stanford.edu/archives/fall2020/entries/autonomy-moral/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCinar AB, Bilodeau S (2024) Incorporating AI into the Inner Circle of Emotional Intelligence for Sustainability. Sustainability 16(15). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su16156648\u003c/span\u003e\u003cspan address=\"10.3390/su16156648\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClark A (2016) Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClark A, Chalmers DJ (1998) Ext Mind Analysis 58(1):7\u0026ndash;19\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClowes R (2015) Thinking in the Cloud: The Cognitive Incorporation of Cloud-Based Technology. Philos Technol 28(2):261\u0026ndash;296. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s13347-014-0153-z\u003c/span\u003e\u003cspan address=\"10.1007/s13347-014-0153-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoeckelbergh M (2013) Human being risk: Enhancement, technology and the evaluation of vulnerability transformations. Springer\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003edel Valle JI, Lara F (2024) AI-powered recommender systems and the preservation of personal autonomy. AI Soc 39(5):2479\u0026ndash;2491. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-023-01720-2\u003c/span\u003e\u003cspan address=\"10.1007/s00146-023-01720-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD\u0026iacute;az-Rodr\u0026iacute;guez N, Ser D, Coeckelbergh J, Prado M, Herrera-Viedma M, E., Herrera F (2023) \u003cem\u003eConnecting the Dots in Trustworthy Artificial Intelligence: From AI Principles, Ethics, and Key Requirements to Responsible AI Systems and Regulation\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoneson D (2019) The Conquest of Fortune: On the Machiavellian Character of Algorithmic Judgment. Social Research: Int Q 86(4):871\u0026ndash;883\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuong Q (2017), November 7 Skip the line: Restaurant wait times on Search and Maps. \u003cem\u003eGoogle\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.blog.google/products/maps/skip-line-restaurant-wait-times-search-and-maps/\u003c/span\u003e\u003cspan address=\"https://www.blog.google/products/maps/skip-line-restaurant-wait-times-search-and-maps/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEarl PE (2016) Bounded Rationality in the Digital Age. In R. Frantz \u0026amp; L. Marsh (Eds.), \u003cem\u003eMinds, Models and Milieux: Commemorating the Centennial of the Birth of Herbert Simon\u003c/em\u003e (pp. 253\u0026ndash;271). Palgrave Macmillan UK. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/9781137442505_15\u003c/span\u003e\u003cspan address=\"10.1057/9781137442505_15\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEllul J (2011) \u003cem\u003eThe technological society\u003c/em\u003e (J. Wilkinson, Trans.). Vintage books. (Original work published 1954)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeenberg A (2005) Heidegger and Marcuse: The catastrophe and redemption of history. Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFink L, Newman L, Haran U (2024) Let me decide: Increasing user autonomy increases recommendation acceptance. Comput Hum Behav 156:108244. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chb.2024.108244\u003c/span\u003e\u003cspan address=\"10.1016/j.chb.2024.108244\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFisher E (2020a) Can Algorithmic Knowledge about the Self Be Critical? In M. Stocchetti (Ed.), \u003cem\u003eThe digital age and its discontents: Critical reflections in education\u003c/em\u003e (pp. 111\u0026ndash;122). Helsinki University Press. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.33134/HUP-4-6\u003c/span\u003e\u003cspan address=\"10.33134/HUP-4-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFisher E (2020b) The ledger and the diary: Algorithmic knowledge and subjectivity. Continuum 1\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10304312.2020.1717445\u003c/span\u003e\u003cspan address=\"10.1080/10304312.2020.1717445\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFisher E, Mehozay Y (2019) How algorithms see their audience: Media epistemes and the changing conception of the individual. Media Cult Soc. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0163443719831598\u003c/span\u003e\u003cspan address=\"10.1177/0163443719831598\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFloridi L (2023) Present: AI as a New Form of Agency, Not Intelligence. In: Floridi L (ed) The Ethics of Artificial Intelligence: Principles, Challenges, and Opportunities. Oxford University Press, pp 14\u0026ndash;30. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/oso/9780198883098.003.0002\u003c/span\u003e\u003cspan address=\"10.1093/oso/9780198883098.003.0002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaju A (2025) Digital Detox in the Age of Hyperconnectivity: Organizational Strategies for Mitigating Technology Overload and Enhancing Employee Flourishing. \u003cem\u003eProceedings of The International Conference on Research in Human Resource Management\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), Article 1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.33422/icrhrm.v2i1.1119\u003c/span\u003e\u003cspan address=\"10.33422/icrhrm.v2i1.1119\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGartenberg C (2017), July 14 Google Maps now can tell you the best time of day to travel to your destination. \u003cem\u003eThe Verge\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.theverge.com/2017/7/14/15973384/google-maps-update-travel-time-when-to-leave-directions-android\u003c/span\u003e\u003cspan address=\"https://www.theverge.com/2017/7/14/15973384/google-maps-update-travel-time-when-to-leave-directions-android\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGibson JJ (1979) The Ecological Approach To Visual Perception. Psychology\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGigerenzer G (2024) Psychological AI: Designing Algorithms Informed by Human Psychology. Perspect Psychol Sci 19(5):839\u0026ndash;848. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/17456916231180597\u003c/span\u003e\u003cspan address=\"10.1177/17456916231180597\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGlickman M, Sharot T (2025) How human\u0026ndash;AI feedback loops alter human perceptual, emotional and social judgements. Nat Hum Behav 9(2):345\u0026ndash;359. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41562-024-02077-2\u003c/span\u003e\u003cspan address=\"10.1038/s41562-024-02077-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoerisch F (2017), January 5 \u003cem\u003eTrack your New Year\u0026rsquo;s fitness goals with Google Calendar\u003c/em\u003e. Google. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://blog.google/products/calendar/track-your-new-years-fitness-goals-with-google-calendar/\u003c/span\u003e\u003cspan address=\"https://blog.google/products/calendar/track-your-new-years-fitness-goals-with-google-calendar/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeater B (2018), May 9 Maps walking navigation is Google\u0026rsquo;s most compelling use for AR yet. \u003cem\u003eTechCrunch\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://social.techcrunch.com/2018/05/08/maps-walking-navigation-is-googles-most-compelling-use-for-ar-yet/\u003c/span\u003e\u003cspan address=\"http://social.techcrunch.com/2018/05/08/maps-walking-navigation-is-googles-most-compelling-use-for-ar-yet/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeersmink R (2017) Distributed selves: Personal identity and extended memory systems. Synthese 194(8):3135\u0026ndash;3151. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11229-016-1102-4\u003c/span\u003e\u003cspan address=\"10.1007/s11229-016-1102-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeidegger M (2008) The Question Concerning Technology (W. Lovitt, Trans.). In \u003cem\u003eBasic Writings\u003c/em\u003e (Revised, Expanded ed. edition). Harper Perennial Modern Classics. (Original work published 1953)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHill K (2019), January 29 I Cut Google Out Of My Life. It Screwed Up Everything. \u003cem\u003eGizmodo\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://gizmodo.com/i-cut-google-out-of-my-life-it-screwed-up-everything-1830565500\u003c/span\u003e\u003cspan address=\"https://gizmodo.com/i-cut-google-out-of-my-life-it-screwed-up-everything-1830565500\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHookway N (2008) `Entering the blogosphere\u0026rsquo;: Some strategies for using blogs in social research. Qualitative Res 8(1):91\u0026ndash;113. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1468794107085298\u003c/span\u003e\u003cspan address=\"10.1177/1468794107085298\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoskins A (2011) 7/7 and connective memory: Interactional trajectories of remembering in post-scarcity culture. Memory Stud 4(3):269\u0026ndash;280. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1750698011402570\u003c/span\u003e\u003cspan address=\"10.1177/1750698011402570\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoskins A (2018) Memory of the Multitude: The End of Collective Memory. In: Hoskins A (ed) Digital Memory Studies: Media Pasts in Transition. Routledge, pp 85\u0026ndash;109\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoskins A (2024) AI and memory. \u003cem\u003eMemory Mind \u0026amp; Media\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/mem.2024.16\u003c/span\u003e\u003cspan address=\"10.1017/mem.2024.16\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIhde D (1990) Technology and the lifeworld: From garden to earth. Indiana University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIhde D (2009) Postphenomenology and technoscience: The Peking University lectures. SUNY\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaspers K (1957) Man in the Modern Age. Anchor Books\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarakayali N, Kostem B, Galip I (2018) Recommendation Systems as Technologies of the Self: Algorithmic Control and the Formation of Music Taste. Theory Cult Soc 35:3\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0263276417722391\u003c/span\u003e\u003cspan address=\"10.1177/0263276417722391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKitchin R, Dodge M (2011) Code/space: Software and everyday life. MIT Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eK\u0026oslash;ster A, Fernandez AV (2021) Investigating modes of being in the world: An introduction to Phenomenologically grounded qualitative research. Phenomenology Cogn Sci. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11097-020-09723-w\u003c/span\u003e\u003cspan address=\"10.1007/s11097-020-09723-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKozinets R (2015) Netnography: Redefined. SAGE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKozinets R (2019) Netnography: The Essential Guide to Qualitative Social Media Research. SAGE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKozinets R, Gambetti R (eds) (2021) Netnography unlimited: Understanding technoculture using qualitative social media research. Routledge/Taylor \u0026amp; Francis Group\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKozinets R (2024) Researching AI chatbots, platforms and the metaverse: Understanding today\u0026rsquo;s netnography. In R. W. Belk \u0026amp; C. Otnes (Eds.), \u003cem\u003eHandbook of Qualitative Research Methods in Marketing\u003c/em\u003e (pp. 185\u0026ndash;196). Routledge. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4337/9781035302727.00025\u003c/span\u003e\u003cspan address=\"10.4337/9781035302727.00025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrippendorff K (2018) Content analysis: An introduction to its methodology. Sage\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLardinois F (2018), May 8 Google Maps will soon give you better recommendations. \u003cem\u003eTechCrunch\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://techcrunch.com/2018/05/08/google-maps-will-soon-give-you-better-recommendations/\u003c/span\u003e\u003cspan address=\"https://techcrunch.com/2018/05/08/google-maps-will-soon-give-you-better-recommendations/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLatour B (1999) Pandora\u0026rsquo;s Hope: Essays on the Reality of Science Studies. Harvard University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLatour B (2005) Reassembling the social: An introduction to actor-network-theory. Oxford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLau J (2020), September 3 \u003cem\u003eGoogle Maps 101: How AI helps predict traffic and determine routes\u003c/em\u003e. Google. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://blog.google/products/maps/google-maps-101-how-ai-helps-predict-traffic-and-determine-routes/\u003c/span\u003e\u003cspan address=\"https://blog.google/products/maps/google-maps-101-how-ai-helps-predict-traffic-and-determine-routes/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLedwich M, Zaitsev A (2019) \u003cem\u003eAlgorithmic Extremism: Examining YouTube\u0026rsquo;s Rabbit Hole of Radicalization\u003c/em\u003e (No. arXiv:1912.11211). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1912.11211\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1912.11211\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLedwich M, Zaitsev A, Laukemper A (2022) Radical bubbles on YouTube? Revisiting algorithmic extremism with personalised recommendations. First Monday. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://firstmonday.org/ojs/index.php/fm/article/view/12552\u003c/span\u003e\u003cspan address=\"https://firstmonday.org/ojs/index.php/fm/article/view/12552\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLevitt HM, Motulsky SL, Wertz FJ, Morrow SL, Ponterotto JG (2017) Recommendations for designing and reviewing qualitative research in psychology: Promoting methodological integrity. Qualitative Psychol 4(1):2\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/qup0000082\u003c/span\u003e\u003cspan address=\"10.1037/qup0000082\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi F, Ruijs N, Lu Y (2023) Ethics \u0026amp; AI: A Systematic Review on Ethical Concerns and Related Strategies for Designing with AI in Healthcare. AI 4(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eArticle 1. https://doi.org/10.3390/ai4010003\u003c/span\u003e\u003cspan address=\"Article 1. 10.3390/ai4010003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLord C (2008) \u003cem\u003eEvolution of the electronic calendar: Introducing social calendaring\u003c/em\u003e. Citeseer\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaier Z (2016), September 30 Now you can see your Google Calendar events in Google Maps. \u003cem\u003eGoogle\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.blog.google/products/maps/now-you-can-see-your-google-calendar-events-google-maps/\u003c/span\u003e\u003cspan address=\"https://www.blog.google/products/maps/now-you-can-see-your-google-calendar-events-google-maps/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartini B, Bellisario D, Coletti P (2024) Human-Centered and Sustainable Artificial Intelligence in Industry 5.0: Challenges and Perspectives. Sustainability 16(13) Article 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su16135448\u003c/span\u003e\u003cspan address=\"10.3390/su16135448\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcAfee A, Brynjolfsson E (2017) Machine, Platform, Crowd: Harnessing Our Digital Future. W. W. Norton \u0026amp; Company\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMenary R (ed) (2010) The extended mind. MIT Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMetz C (2016), September 27 An Infusion of AI Makes Google Translate More Powerful Than Ever. \u003cem\u003eWired\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.wired.com/2016/09/google-claims-ai-breakthrough-machine-translation/\u003c/span\u003e\u003cspan address=\"https://www.wired.com/2016/09/google-claims-ai-breakthrough-machine-translation/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMhlanga D (2022) Human-Centered Artificial Intelligence: The Superlative Approach to Achieve Sustainable Development Goals in the Fourth Industrial Revolution. Sustainability 14(13) Article 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su14137804\u003c/span\u003e\u003cspan address=\"10.3390/su14137804\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMitcham C (1994) Thinking Through Technology: The Path Between Engineering and Philosophy. University of Chicago Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNewman-Griffis D (2024) \u003cem\u003eAI Thinking: A framework for rethinking artificial intelligence in practice\u003c/em\u003e (No. arXiv:2409.12922). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2409.12922\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2409.12922\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNimmo R (2011) Actor-Network Theory and Methodology: Social Research in a More-Than-Human World. Methodological Innovations Online 6(3):108\u0026ndash;119. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4256/mio.2011.010\u003c/span\u003e\u003cspan address=\"10.4256/mio.2011.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOzmen Garibay O, Winslow B, Andolina S, Antona M, Bodenschatz A, Grieman K, Jirotka M, Ten Holter CA (2023) Six human-centered artificial intelligence grand challenges. Int J Hum Comput Interact 39(3). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10447318.2022.2153320\u003c/span\u003e\u003cspan address=\"10.1080/10447318.2022.2153320\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasquale F (2015) The Algorithmic Self. Hedgehog Rev, 30\u0026ndash;45\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasquale F (2016) \u003cem\u003eThe Black Box Society: The Secret Algorithms That Control Money and Information\u003c/em\u003e (Reprint edition). Harvard University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePham A, Rubel A, Castro C (eds) (2021) Autonomy, Agency, and Responsibility. Algorithms and Autonomy: The Ethics of Automated Decision Systems. Cambridge University Press, pp 21\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/9781108895057.002\u003c/span\u003e\u003cspan address=\"10.1017/9781108895057.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePichai S (2013), May 15 \u003cem\u003eLive from Google I/O: Mo\u0026rsquo; screens, mo\u0026rsquo; goodness\u003c/em\u003e. The Keyword. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.blog.google/products/android/live-from-google-io-mo-screens-mo/\u003c/span\u003e\u003cspan address=\"https://www.blog.google/products/android/live-from-google-io-mo-screens-mo/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePickering A (2010) Material Culture and the Dance of Agency. \u003cem\u003eThe Oxford Handbook of Material Culture Studies\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/oxfordhb/9780199218714.013.0007\u003c/span\u003e\u003cspan address=\"10.1093/oxfordhb/9780199218714.013.0007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamnath J (2016), April 12 Find time for your goals with Google Calendar. \u003cem\u003eThe Keyword\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.blog.google/products/calendar/find-time-goals-google-calendar/\u003c/span\u003e\u003cspan address=\"https://www.blog.google/products/calendar/find-time-goals-google-calendar/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRicci F, Rokach L, Shapira B (2015) Recommender Systems: Introduction and Challenges. In: Ricci F, Rokach L, Shapira B (eds) Recommender Systems Handbook. Springer US, pp 1\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-1-4899-7637-6_1\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4899-7637-6_1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiedl MO (2019) Human-centered artificial intelligence and machine learning. Hum Behav Emerg Technol 1(1):33\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hbe2.117\u003c/span\u003e\u003cspan address=\"10.1002/hbe2.117\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRisko EF, Gilbert SJ (2016) Cognitive Offloading. Trends Cogn Sci 20(9):676\u0026ndash;688. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tics.2016.07.002\u003c/span\u003e\u003cspan address=\"10.1016/j.tics.2016.07.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRobert LP, Pierce C, Marquis L, Kim S, Alahmad R (2020) Designing fair AI for managing employees in organizations: A review, critique, and design agenda. Human\u0026ndash;Computer Interact 35(5\u0026ndash;6):545\u0026ndash;575. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/07370024.2020.1735391\u003c/span\u003e\u003cspan address=\"10.1080/07370024.2020.1735391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosa H (2013) Social acceleration: A new theory of modernity. Columbia University, Trans, J. Trejo-Mathys\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRyan RM, Deci EL (2011) A Self-Determination Theory Perspective on Social, Institutional, Cultural, and Economic Supports for Autonomy and Their Importance for Well-Being. In V. I. Chirkov, R. M. Ryan, \u0026amp; K. M. Sheldon (Eds.), \u003cem\u003eHuman Autonomy in Cross-Cultural Context: Perspectives on the Psychology of Agency, Freedom, and Well-Being\u003c/em\u003e (pp. 45\u0026ndash;64). Springer Netherlands. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-90-481-9667-8_3\u003c/span\u003e\u003cspan address=\"10.1007/978-90-481-9667-8_3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchieffer S (2016) April 27). Google Calendar for Android: Find a time for my meeting. Google Blog\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSearle JR (1983) Intentionality, an essay in the philosophy of mind. Cambridge University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSearle JR (2010) Making the social world: The structure of human civilization. Oxford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShneiderman B (2022) Human-centered AI. Oxford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimon HA (1971) Designing organizations for an information-rich world. In M. Greenberger (Ed.), \u003cem\u003eComputers, Communication, and the Public Interest\u003c/em\u003e (pp. 37\u0026ndash;72). The Johns Hopkins Press. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://digitalcollections.library.cmu.edu/awweb/awarchive?type=file\u0026amp;item=33748\u003c/span\u003e\u003cspan address=\"https://digitalcollections.library.cmu.edu/awweb/awarchive?type=file\u0026amp;item=33748\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith MN (2017) Intentions: Past, present, future. Philosophical Explorations 20(sup2):1\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/13869795.2017.1356360\u003c/span\u003e\u003cspan address=\"10.1080/13869795.2017.1356360\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSrivastava DK, Lilly JM, Feigh KM (2023) \u003cem\u003eImproving Operator Situation Awareness when Working with AI Recommender Systems\u003c/em\u003e (No. arXiv:2310.11370). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2310.11370\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2310.11370\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUmapathy V (2015), December 7 \u003cem\u003eAdd to-dos to your Google Calendar using Reminders\u003c/em\u003e. The Keyword. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://blog.google/products/calendar/add-to-dos-to-your-google-calendar/\u003c/span\u003e\u003cspan address=\"https://blog.google/products/calendar/add-to-dos-to-your-google-calendar/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Dijck J (2007) Mediated Memories in the Digital Age. Stanford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVarshney LR (2020) \u003cem\u003eRespect for Human Autonomy in Recommender Systems\u003c/em\u003e (No. arXiv:2009.02603). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2009.02603\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2009.02603\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang D, Ma X, Wang AY (2021) \u003cem\u003eHuman-Centered AI for Data Science: A Systematic Approach\u003c/em\u003e (No. arXiv:2110.01108). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2110.01108\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2110.01108\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J (2021), November 4 \u003cem\u003eGoogle Maps navigates its way to 10 billion installs\u003c/em\u003e. Android Police. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.androidpolice.com/google-maps-navigates-its-way-to-10-billion-installs/\u003c/span\u003e\u003cspan address=\"https://www.androidpolice.com/google-maps-navigates-its-way-to-10-billion-installs/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWard AF (2013) \u003cem\u003eOne with the Cloud: Why People Mistake the Internet\u0026rsquo;s Knowledge for Their Own\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dash.harvard.edu/handle/1/11004901\u003c/span\u003e\u003cspan address=\"https://dash.harvard.edu/handle/1/11004901\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWard AF, Duke K, Gneezy A, Bos MW (2017) Brain Drain: The Mere Presence of One\u0026rsquo;s Own Smartphone Reduces Available Cognitive Capacity. J Association Consumer Res 2(2):140\u0026ndash;154. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1086/691462\u003c/span\u003e\u003cspan address=\"10.1086/691462\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWellner G, Mykhailov D (2023) Caring in an Algorithmic World: Ethical Perspectives for Designers and Developers in Building AI Algorithms to Fight Fake News. \u003cem\u003eScience and Engineering Ethics\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11948-023-00450-4\u003c/span\u003e\u003cspan address=\"10.1007/s11948-023-00450-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWertenbroch K, Schrift RY, Alba JW, Barasch A, Bhattacharjee A, Giesler M, Knobe J, Lehmann DR, Matz S, Nave G, Parker JR, Puntoni S, Zheng Y, Zwebner Y (2020) Autonomy in consumer choice. Mark Lett 31(4):429\u0026ndash;439. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11002-020-09521-z\u003c/span\u003e\u003cspan address=\"10.1007/s11002-020-09521-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWoolf NH, Silver C (2017) Qualitative analysis using ATLAS.ti: The five-level QDA method. Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYadav PS (2025) \u003cem\u003eCognitive Sustainability in the Age of AI: A Philosophical Framework for Understanding Competency Erosion and Cognitive Stratification in Human-AI Systems\u003c/em\u003e (SSRN Scholarly Paper No. 5317987). Social Science Research Network. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.5317987\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.5317987\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZuboff S (2018) \u003cem\u003eThe Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power\u003c/em\u003e. PublicAffairs\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Human-centered AI, Recommendation Systems, Digital Autonomy, Algorithmic Decision-Making, Sustainable Innovation, User Empowerment, Netnography","lastPublishedDoi":"10.21203/rs.3.rs-9380629/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9380629/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study examines the impact of AI-powered recommendation systems. Using a netnographic approach, we analyzed user experiences with Google's recommendation systems ecosystem over five years (2016\u0026ndash;2020).\u003c/p\u003e \u003cp\u003eIntegrating post-phenomenological analysis, actor-network theory, and extended cognition theory, we used thematic analysis of 525 blog posts, 46 books, and 25 academic studies to reveal how recommendation systems mediate cognitive processes. Our findings revealed potential influences on human autonomy, intention, rationality, and memory. Three patterns were identified. First, users experience decisional delegation, where algorithmic suggestions gradually replace autonomous decision-making. Second, intentionality becomes hybridized as users struggle to distinguish self-generated intentions from algorithmically mediated ones. Third, memory as a basis for decisions and sense-making shifts from individual and collective to algorithmically mediated connective memory, where personal history becomes inseparable from platform-mediated data aggregation.\u003c/p\u003e \u003cp\u003eThese mediating effects challenge prevailing human-centered AI paradigms that prioritize human control and transparency without addressing potential cognitive impacts. We propose reconceptualizing HCAI around cognitive sustainability: preserving autonomous decision-making, maintaining authentic intention formation, and protecting memory sovereignty. Practical implications include implementing reflexivity mechanisms, establishing temporal boundaries for cognitive preservation, and developing assessment frameworks to evaluate long-term cognitive impacts beyond efficiency gains.\u003c/p\u003e","manuscriptTitle":"Beyond Efficiency: Rethinking Human-entered AI for Cognitive Sustainability","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-23 15:18:15","doi":"10.21203/rs.3.rs-9380629/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"143379301122362929126707064492521399615","date":"2026-05-19T10:54:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"124197217042323040167006917186381028625","date":"2026-05-18T16:42:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"23486857298054837682023999179529310376","date":"2026-05-08T16:40:53+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-15T17:52:33+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-15T09:20:44+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-15T09:20:00+00:00","index":"","fulltext":""},{"type":"submitted","content":"Cognition, Technology \u0026 Work","date":"2026-04-10T14:16:56+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"b07d4512-a3e1-4334-990b-87274cf37495","owner":[],"postedDate":"April 23rd, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"143379301122362929126707064492521399615","date":"2026-05-19T10:54:33+00:00","index":18,"fulltext":""},{"type":"reviewerAgreed","content":"124197217042323040167006917186381028625","date":"2026-05-18T16:42:27+00:00","index":17,"fulltext":""},{"type":"reviewerAgreed","content":"23486857298054837682023999179529310376","date":"2026-05-08T16:40:53+00:00","index":13,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-23T15:18:15+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-23 15:18:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9380629","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9380629","identity":"rs-9380629","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}