Sensemaking AI: Introducing a Research and Design Agenda for Human–AI Networks

Sensemaking AI: Introducing a Research and Design Agenda for Human–AI Networks | 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 Sensemaking AI: Introducing a Research and Design Agenda for Human–AI Networks Tina Comes This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7463619/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 19 Mar, 2026 Read the published version in EPJ Data Science → Version 1 posted 11 You are reading this latest preprint version Abstract Digital technologies and AI promise to optimise complex systems through data-driven decisions, predictive modelling, and anticipatory action. However, this optimisation imperative creates a fundamental paradox: as systems excel at achieving measurable objectives, they may erode the collective intelligence and adaptive capacity of our societies. Recognising this tension, the field of Human-Centred AI (HCAI) has emerged to ensure AI aligns with human values. However, research on HCAI often focuses on idealised interactions, neglecting the pressure, moral dilemmas, and social dynamics typical of today’s complex problems. This paper introduces and advocates for a paradigm shift towards Sensemaking AI : AI that supports collective meaning-making processes in evolving human-AI networks. This novel perspective recognises that algorithmic and AI systems actively participate in the social processes through which humans interpret information, coordinate responses, and adapt their values. Grounded in sensemaking and decision theory and informed by a scoping review of the HCAI literature, this paper identifies three connected research areas: (i) sensemaking-aware automation that preserves interpretive flexibility; (ii) collective agency for network-level control; and (iii) value-aware sensemaking that supports collective meaning-making. These principles form the basis for Sensemaking AI as a design and research agenda that prioritises collective meaning-making and democratic deliberation in networks. Scoping Review Sensemaking AI Human-AI interaction Decision Theory Human-Centred AI Complex Systems Collective Intelligence Optimisation: Human-AI networks Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction Our societies are complex dynamical systems increasingly shaped by digital technologies. As decision-makers grapple with increasing complexity, AI is portrayed as a ‘solution’ for its ability to process vast amounts of data and reduce the cognitive biases in human decisions by providing ‘objective’ or ‘data-driven’ optimal answers. Digital twins (Fan et al., 2021), automated decisions (Coppi et al., 2021), and predictive models for ‘anticipatory action’ (Kjærum & Madsen, 2025) promise efficient, scalable, rapid and cheap solutions to address complex and time-compressed problems. With the promise of ‘taming complexity’ (Gil et al., 2021), AI has become both a tool to understand complex systems and a driver that shapes their behaviour. As such, AI systems do not just optimise a system: they shape the informational and social environments via which humans understand their environment and respond to the challenges they are confronted with. Recognising this dual nature, this paper echoes the calls for Machine Behaviour (Rahwan et al., 2019), Human-Centered AI (HCAI) (Ozmen Garibay et al., 2023) and Hybrid Intelligence (Akata et al., 2020) that all conceptualise AI as a social-technical system. Especially HCAI has emerged as a prominent field advocating for explainable, fair, and transparent AI systems that keep humans “in the loop” rather than replacing them (Shneiderman, 2020). This human-centred approach seeks to address the optimisation paradox by designing AI that enhances rather than diminishes human agency. Yet HCAI is based on idealised situations that insufficiently address the time pressure, high stakes, volatility, moral dilemmas and shifting networks typical for decisions in today’s complex problems – even though these conditions have been shown to fundamentally change sensemaking and lead decision-makers to overlook, discard, or not act upon information (Paulus et al., 2022). In complex problems, where multiple interconnected challenges demand rapid coordination across scales and domains (Mahajan et al., 2022) this lack of understanding may lead to overlooking important implications of the use of AI, which in turn jeopardises the central tenet of meaningful human control (Cavalcante Siebert et al., 2023). Sensemaking offers a crucial lens and theoretical framework to address this tension because it focuses on how humans collectively interpret ambiguous situations, construct meaning from uncertain information, and adapt as understanding evolves (Weick, 1995). Unlike optimisation approaches with predetermined objectives, sensemaking embraces the continuous social processes through which values and objectives emerge and evolve. This is essential when dealing with complex problems that can only be addressed by decentralised interaction rather than top-down optimisation. This paper contributes in three ways to the ongoing discourse on optimisation and AI: building on foundations in sensemaking support systems (Muhren & Van de Walle, 2010; Seidel et al., 2018) and decision theory (French, 1986), it introduces Sensemaking AI as AI that actively supports sensemaking processes in evolving human-AI-networks , and delineates it from other paradigms such as HCAI (Shneiderman, 2020) or machine behaviour (Rahwan et al., 2019). Empirically, a scoping review of 101 academic articles from the Human-Centred AI (HCAI) literature examines how, and to what extent, current HCAI research addresses the dynamics of human-AI interaction in complex, time-sensitive, and morally fraught environments. Poly-crises serve here as illustrative cases to highlight the limitations of optimisation-centric AI design. P rogrammatically, the paper then develops design principles and a research agenda on sensemaking AI along three axes: sensemaking-aware automation that promotes interpretive flexibility; collective agency for network-level control; and value-aware sensemaking AI. As such, this paper is meant to inspire the reader and start a conversation across computer, cognitive and behavioural sciences that acknowledges that fundamentally we humans are social beings who continuously seek to make sense of our environment. 2. On the Impact of AI on Sensemaking and Decision-Making This section provides the theoretical foundations for understanding how AI transforms human sensemaking and decision-making. I first provide insights into Sensemaking as a theory and process for navigating complexity. Then, I explore the foundations of Human-Centred AI to showcase how the notion of control is shifting with the ubiquity of AI and the complexity of the problem addressed. From there, I move to sketch how AI systems reshape information-decision feedback dynamics in networked social-technical systems. 2.1 Sensemaking We are confronted with several accelerating poly-crises (Søgaard Jørgensen et al., 2024). Climate change, geopolitical instability, and increasing inequality and fragmentation represent “wicked problems” characterised by contested problem definitions, interdependencies, and evolving values (Rittel & Webber, 1973). Data-driven optimisation and AI promise clearly identifying the ‘best’ alternatives even– or especially – if problems are complex. However, by definition, wicked problems cannot be solved by traditional optimisation approaches (French, 2012). Weick's concept of sensemaking (Weick, 1993, 1995) provides the theoretical foundations to explain how humans navigate such ambiguity and uncertainty. Sensemaking is the process of meaning making, by which humans structure and process the stream of unfiltered, chaotic data, and turn it into meaningful and actionable information (Weick, 1993). Sensemaking is a creative process, wherein we humans construct ‘bridges’ to address uncertainties consisting of ideas, thoughts, emotions, feelings, and memories (Sharoda & Reddy, 2010). Importantly, sensemaking is a social process, through which decision-makers interact with their peers and the environment, they coordinate, and receive feedback through enactment, allowing the formation of collective action agendas (Maitlis & Christianson, 2014). Sensemaking also entails the process of identity construction, by which humans come to understand what is 'meaningful' in their own identities, teams and organisations (Helms Mills et al., 2010). Unlike rational decision-making models that assume clear problems and predetermined objectives (Gralla et al., 2016), sensemaking recognises that in complex environments, actors first construct what situations mean before determining their responses (Comes et al., 2020). Where optimisation assumes fixed objectives and measurable trade-offs, sensemaking analyses how those objectives emerge through interaction, feedback, and evolving understanding. Rather than converging on stable solutions, sensemaking sustains the ambiguity necessary for creative reinterpretation and collective learning (Weick & Sutcliffe, 2007). As AI systems increasingly filter information, prioritise, and suggest what to do, they inevitably become participants in these sensemaking dynamics and shape the cognitive and social foundations on which sensemaking depends. The question that emerges is not simply how to design AI that supports human decisions, but how to understand AI’s role in shaping the very processes through which our collective understanding and coordinated action emerge. 2.2 Human-Centred AI: Towards Control in Human-AI Networks The term human-centred AI is increasingly popular in response to the many concerns about AI risks and human agency in increasingly digitally mediated environments (Capel & Brereton, 2023). Instead of a world in which AI optimises our lives and takes control, human-centred AI presents a design paradigm to ensure that AI serves people and enhances human capabilities rather than replacing them (van Berkel et al., 2022). By definition, such an AI is trustworthy, reliable and safe for humans to use (Auernhammer, 2020; Shneiderman, 2020). Often, control is framed as a question of who carries out which task or holds responsibility for it. The optimal distribution of tasks between humans and machines has been a central question since the 1950s, when Fitts (1951) introduced a protocol to decide which tasks are better performed by machines or by humans in the context of air traffic control. Today, this question has evolved to address how AI can support and optimise human decisions in complex problems while avoiding the pitfalls of cognitive or motivational biases. Tasks are assigned to humans or machines based on their respective expertise to optimise performance (Tausch & Kluge, 2022), with humans being better at tasks requiring creative and social intelligence (Ponti & Seredko, 2022). If tasks are allocated to machines, rapidly questions become acute about human autonomy and control (Abbass, 2019). However, the question of control in human-AI systems extends beyond task allocation. The question is not merely about who (or what) decides. Rather, we need to ask how we can preserve meaningful human agency when AI systems shape the interpretive context in and through which we make choices. As AI filters, structures, and sequences information, it influences attention, salience, and relevance, thereby directing what we read, see, ignore, or find important. In doing so, AI systems (co-)construct the conditions in which human sensemaking unfolds. Many of the frameworks to understand control stem from understanding human operators of machines, ranging from submarines (Sheridan et al., 1978) to air traffic (Council, 1998). These frameworks put forward assume a single human operator interacting with one machine (Endsley, 2017; Parasuraman et al., 2000). Yet, addressing complex challenges requires decentralised approaches, where many people (and potentially algorithms) work together (Mahajan et al., 2022). Increasingly, there is a discussion around human-AI teams (HAT), “ a purposeful combination of human and cyber-physical elements that collaboratively pursue goals that are unachievable by either individually ” (Alix et al., 2021). Human-AI teams are composed of few people and machines (Endsley, 2023), whereas complex problems require the coordination of large-scale, ad-hoc and decentralised networks of humans and AI. Figure 1 summarises the evolution of research on human-machine interaction and control. With increasingly complex challenges and pervasiveness of AI, research needs to shift from AI that mimics human reasoning and single-operator systems to decentralised evolving human-AI networks that require coordination and human control. While many AI-supported decisions are intended as quick fixes and short-term optimisations, they often produce lasting consequences. Once enacted, decisions alter social networks, infrastructures, and expectations, creating path dependencies that narrow future choices (Webster, 2008). These effects are especially problematic when they shape collective sensemaking trajectories (Helms Mills et al., 2010), influencing how future situations are interpreted and what actions seem legitimate. Moreover, AI systems themselves evolve through feedback: large language models, for instance, adapt to user behaviour, reinforcing patterns of engagement and attachment (Kim et al., 2025) and producing recursive information bubbles (Jacob et al., 2025). Over time, both the humans and the AI intended to support the humans may drift away from the original intentions in ways that become difficult to detect, control or reverse. Preserving human control thus requires more than oversight: it demands mechanisms for temporal reflexivity : the ability to recognise deviations from purpose and intervene in unfolding trajectories when needed. These temporal shifts and the decentralised nature of human-AI networks challenge conventional notions of agency and control, particularly as AI becomes embedded in systems that evolve faster than human governance can follow. In the next section, I examine how these dynamics manifest via information-decision-feedback loops. 2.3 Optimisation Cascades and Information-Decision-Feedback AI and humans exist in a complex control relationship: AI influences human sensemaking and decision-making, and humans engineer, train, steer, and – possibly – control the AI (Rahwan et al., 2019). Despite this interdependence, information and decisions are typically studied independently, and decision-information-feedback is only marginally considered. By their very nature, poly-crises require urgent interventions despite their complexity. This urgency matters, as time compression alters information processing, information sharing, and human decision-making behaviour. With prospect theory in the 1970s (Kahneman & Tversky, 1979), the need to include cognitive aspects in risky decisions became prominently recognised and inspired a wealth of research pointing to the cognitive and motivational biases that they bring (Klein et al., 2010; Paulus et al., 2022; Weick, 1993). An obvious question then is: can AI improve human decision-making by overcoming the cognitive or motivational biases? To unpack this question, an analysis of human decision-making processes is helpful, see Figure 2, left side. The human process is driven by the interplay of sensemaking, coordination and decision-making, by which humans adapt to the changing informational and social environment. While often, it is assumed that decision-making is the guiding principle to organise coordination, research has shown that it is the interaction of sensemaking and decision-making (Comes et al., 2020; Gralla et al., 2016). This implies that as the understanding of the situation changes, also the objectives, preferences, values and thereby the required or desired solutions change, fundamentally contradicting the linear problem solving paradigm, by which first a problem is formalised, and then solved (Volkema, 1983). AI and optimisation algorithms fundamentally change how we perceive our environment, with whom and how we interact to find solutions (Kiesler & Sproull, 1992). For illustrative purposes, Figure 2 (right side) shows the most extreme case where one or several AI systems autonomously acquire and analyse data; optimise, decide and implement the decision in ‘ optimisation cascades’ . This is not meant as a prescription, but to contrast with the human sensemaking cycles. In this case, the creative, social and identity-forming process of sensemaking is replaced by data analysis. Optimisation cascades manifest across scales. At the individual level, recommendation algorithms and LLMs optimise for engagement based on personal information, shaping what people read or see and thereby gradually shifting user preferences and choices (Sharma et al., 2024) . At the policy level, performance metrics drive decisions under the umbrella of New Public Management, even if they only poorly capture intended outcomes. This phenomenon that Muller (2018) called the “tyranny of metrics” distorts our understanding of complex phenomena such as poverty, climate change, or welfare. A fundamental tension between human sensemaking cycles and optimisation cascades is that human processes are iterative social cycles where problem definitions and objectives co-evolve with shared understanding. In contrast, optimisation processes replace the underlying creative ambiguity with predetermined static metrics and automated execution of the optimal decision. The risk here is not only that machines make “wrong” decisions, but that the optimisation fundamentally impacts human sensemaking, shaping social orders, norms of our interactions, and what we perceive as ‘good’ or even thinkable solutions. Today, hybrid approaches, by which many humans and machines work together, are becoming the norm. An open question is how to coordinate the resulting dynamic networks of humans and AI, while controlling information-decision feedback loops. Here, coordination is defined as the set of procedures by which teams plan, organise, orchestrate and integrate their activities to achieve shared goals (Malone et al., 1994). As such, coordination entails activities such as information sharing, planning, task allocation or scheduling (Neale et al., 2004). Several studies have confirmed that decision performance on distributed tasks improves if individuals know who has access to what information (Stasser & Titus, 1985; Stewart & Stasser, 1995), which is problematic in complex networks and with black box AI models. Further, if AI – and especially generative AI – optimises information based on past interactions, it can introduce or reinforce existing biases (Sharma et al., 2024), which are then amplified via path-dependencies in sensemaking and decision trajectories. Given the broad impact of AI and optimisation on human sensemaking and decision-making in complex networks, the next sections examine how and in how far current Human-Centred AI literature addresses these fundamental challenges. 3. Methods To assess how the literature navigates the tension between data-driven optimisation and human meaning-making in uncertain and evolving contexts, I focus on the HCAI literature. HCAI addresses the relationship between human agency and AI systems. As such, it provides an important lens for evaluating how AI design accounts for the cognitive, social, and ethical dynamics of complex decisions. To explore this evolving body of work, I conducted a scoping review. Scoping reviews are designed to determine the scope or coverage of an emergent body of literature (Munn et al., 2018) and identify gaps in emerging interdisciplinary fields such (Sadek et al., 2024). The scoping review was conducted on rayyan (https://www.rayyan.ai/) and followed the extended PRISMA protocol for scoping reviews (Tricco et al., 2018). Figure 3 summarises the screening and analysis process. 3.1 Article Identification The search was based on Scopus, because of its comprehensive coverage of interdisciplinary sources. To determine how human values and decisions are represented in the HCAI literature, the search focused on human-centred AI, searching combinations of “human-centred AI”, “human-centered AI” OR “HCAI” with “Decision support”, “Values” OR “Principles”. This framing focuses on decision problems to reveal the limitations of optimisation-driven AI . The search was limited to articles in English, focusing on original manuscripts, case studies and perspective papers. To account for the rapid growth of the field, a preliminary pilot was conducted in June 2024, and the final search was executed on February 14, 2025. Excluding duplicates, the search led to 419 articles. 3.2 Screening The titles, abstracts and keywords of all articles were screened. Articles had to focus on supporting specific decision-making or sensemaking tasks and discuss the distribution of the tasks among humans and machines under the framework of human-centred AI to ensure that the different HCAI concepts or principles could be mapped out. This initial screening process resulted in 185 papers that were retained for the full-text review. 3.3 Full-text Eligibility The articles were then analysed based on the extent of discussion on human-AI interaction, decision-making/ sensemaking and HCAI concepts. Studies were excluded if they: (i) focused on physical interaction (e.g., robotics); (ii) addressed only interface design without engaging HCAI principles; (iii) referred only to conventional statistics, not AI; or (iv) were categorised as reviews or tutorials, or (v) were inaccessible, not in English, see Figure 3. 3.4 Analysis and Extraction Framework The analysis began by mapping publication trends and methods across the reviewed studies. For the methods used, I categorise (i) perspective and conceptual papers that theorise or aim to guide the use of AI; (ii) empirical studies that observes how humans interact with machines in real life; (iii) behavioural experiments; (iv) design studies that build and implement the AI, also including technical studies, and (v) surveys. Further, I distinguish the field , for which the AI is designed, from the type of AI or algorithm , ranging from conversational agents and chatbots, large language models and generative AI to machine learning. From there, I investigate the role of the human vis a vis the machines to understand how and in how far humans are – indeed – central to HCAI, and what the implications are for optimisation, sensemaking and decision-making. Questions are who has decision authority (the human or the machine); and how many humans are interacting with how many machines. I conceptualise this in terms of the following categories: (i) AI supports a single (human) user (who is responsible for the decision); (ii) the AI makes the decision autonomously; (iii) AI and humans act as a team, where the AI interacts with and advises multiple humans, or (iv) one AI interacts with many humans, going beyond the concept of a team. Finally, the review assessed which HCAI design principles and concerns are most prominent in the literature, and how they co-occur. Because many terms in the literature—such as fairness, explainability, and automation—blur the boundary between normative values and technical features, I treat them collectively under the umbrella of HCAI design principles . This includes both ethical principles (e.g., solidarity, equity) and system-level properties (e.g., transparency, control, privacy). The initial coding categories were established on the basis of major guidelines (EC, 2019; OECD, 2019; UNESCO, 2022), as well as the seminal HCAI literature (Shneiderman, 2020). Through bottom-up coding concepts were added. An analysis of frequency and co-occurrence patterns was then used to explore dominant principles and gaps in the HCAI literature. Table 1 provides an overview of the dimensions of analysis and categories used. Table 1: Overview of the Analysis & Extraction Framework for the Scoping Review Dimension Categories Research Method Perspective/conceptual; qualitative empirical; behavioural experiments; design and tool developments; surveys Field of Application Medicine & Health; Public policy & governance, Business; Education; Crisis & Safety; Manufacturing; Software Engineering; Transport; Other Type of AI AI agnostic/unspecified; Machine Learning, Natural Language Programming (NLP); Generative AI; Agentic AI Relation of Human And AI One AI supports one single human; AI decides autonomously; one AI interacts with many humans; Human-AI teams Decision Authority Human only, Machine only, Shared (interactive), Distributed HCAI DESIGN Principles Considered Explainability, Fairness, Trust, Transparency, Accountability, Solidarity, Contextualisation, Empowerment, Safety, Humanity, Control, Agency, Privacy, Useability, Responsibility, Automation, Equity, Situational Awareness, Bias 3.5 Design Principles for Sensemaking AI In addition to the scoping review, this paper develops a set of theoretical arguments and proposes three interconnected design principles for Sensemaking AI with the aim of starting to address conceptual blind spots of the current literature. The design principles proposed are developed through an abductive process (Sætre & Van de Ven, 2021) that iteratively links theoretical framing from Sensemaking and Decision-Making theories and empirical patterns from the review with illustrative examples from poly-crises to develop a research and design agenda for AI that supports meaning-making in complex systems. 4. Results This section presents the results of the scoping review and examines how HCAI research engages with the tensions between optimisation, human agency, and sensemaking in complex environments. By analysing patterns in methods, applications, AI system types, roles of humans and AI, and HCAI design principles, this results section aims to provide insights into how the field has evolved and where important gaps may remain. 4.1 The How, What and Where of HCAI Along with the rise of AI, there is also growing interest at the intersection of human-centred AI and decisions sparked by the first HCAI publication by Shneiderman (2020). The distribution of research methods (Figure 4) shows a highly diverse set of research methods, in which experimental, design and perspective approaches dominate, each with almost 30 papers. Under design methods, papers are classified that design, build, and test Human-Centred AI applications, e.g., (Elahi et al., 2021; Erlei, 2024; Sun, 2022), indicating a focus on development and evaluation. In contrast, the many perspective and opinion papers emphasise theoretical and conceptual discussions, primarily regarding the value perspective in HCAI– especially in sensitive contexts such as education, health or crisis management (Comes, 2024; Kattnig et al., 2024; van Leersum & Maathuis, 2025). Controlled experiments often study cognition and behaviour by focusing on how users interact with an AI. These studies link HCAI to Human-Computer-Interaction and often test for the impact of specific principles such as fairness or explainability (Flathmann et al., 2023; Gajos & Mamykina, 2022). Qualitative methods that study the use of AI in situ and surveys that focus on perception and use of technology especially in work processes are less common (Bingley et al., 2023; Herrmann & Pfeiffer, 2023). Figure 5 shows the distribution of the types of AI that are considered. Strikingly, almost 40% of the publications theorise or analyse ‘ AI’ as a generic term, being agnostic of the specific algorithm (Akula & Garibay, 2021; Bingley et al., 2023; Hoch et al., 2022). Machine learning (ML), including deep learning, is the most common approach specified with more than 35 % of publications, either alone or in combination with other methods such as natural language processing (NLP) and generative AI . Increasingly, generative AI papers are a part of the HCAI literature (Buçinca, 2024; Erlei, 2024; Kattnig et al., 2024). Agentic AI , which refers to AI systems with autonomous decision-making capabilities (Acharya et al., 2025), is also increasingly well-represented with almost 10% of publications, indicating a growing interest in AI autonomy and adaptive behaviour – most often via humans interacting with an autonomous agent (Criscuolo & Dolci, 2024; Gou et al., 2024). The review clearly shows that HCAI has made its entrance in a diversity of fields, cf. Figure 6. Applications in health and medicine are most prominent (19/101), most often in the context of clinical decision support systems for diagnosis or treatment e.g., (Lee et al., 2022; Van Berkel et al., 2023; Verma et al., 2023). This is followed by public policy and governance (Lee et al., 2019; Lettieri et al., 2023; Stapleton et al., 2022) including several papers that discuss sustainability aspects (Sigfrids et al., 2023). Business and managerial applications (Freire et al., 2023; Hoch et al., 2022) follow. In education (Chaudhry et al., 2022; Duan et al., 2024) and crisis management, papers also discuss the risks of AI especially regarding the introduction of new biases (Chaudhry et al., 2022). ‘Other’ is a highly diverse category, including law, agriculture, and language. 4.2 The Role of AI in HCAI When it comes to the roles of the AI, Figure 7 shows that with 65/101 papers, the large majority of papers focus on an AI supporting an individual decision-maker, ranging from supporting elite sports coaches (Comes, 2024; Maiden et al., 2023) to supporting elderly app users in smart cities (Elahi et al., 2021), from nurses steering patients to a hospital (Li et al., 2024) to managers making strategic decisions (Passlack et al., 2024). Fewer papers (13/101) discuss Human-Centred AI for autonomous decision-making by which the AI fully automates decisions (Bingley et al., 2023; He et al., 2023; Jin et al., 2022; Lee et al., 2019; Nabizadeh Rafsanjani & Nabizadeh, 2023; Shulner-Tal et al., 2023; Suchan et al., 2021; Yazdanpanah et al., 2021). Despite the increasing prevalence of AI in society, only nine papers discuss one AI that supports many humans, primarily in the context of crowdsourcing (Sprenkamp et al., 2025). Further thirteen papers study human-AI-teams, and the implications for group dynamics (Bansal et al., 2021; Cooke et al., 2020; Flathmann et al., 2023; Riedl, 2019), for instance in medical teams (Hagemann et al., 2023; Verma et al., 2023) or manufacturing (Hoch et al., 2022). What is, however, missing are studies that analyse the broader societal implications, collective intelligence, and democratic processes whereby many humans work with many AI systems that impact information flows and decision-making. When many actors optimise their choices based on optimised input by AI algorithms, their collective behaviour can generate system-wide effects that no single actor anticipated or can control. If AI is implemented at scale and optimised decisions propagate through and even shape the networks of our interaction, what are the emergent effects on human sensemaking and decision-making over time? And how can these emergent effects in complex human–AI networks be understood and controlled? The principles that guide the interaction of one human with one AI or one AI with a small group of humans may not suffice for analysing the dynamics of complex networks in which many humans are supported by many machines. These dynamics raise questions about automation, coordination, and the preservation of meaningful human control. 4.3 What HCAI Optimises for Decisions are inherently linked to what we value. AI that supports human decision-making must therefore also grapple with fundamental questions about which values and principles should guide their design and operation. As outlined in the methods section, the HCAI literature blurs the boundary between normative values and technical requirements. Figure 8 shows the concepts currently used as design principles and guiding values in the HCAI literature. Clearly, current HCAI research emphasises explainability (n=45) (Sun, 2022), trust (n=29) (Liao & Sundar, 2022), transparency (n=24) (Kunar et al., 2024), and fairness (n=23) (Kattnig et al., 2024), while other concepts ranging from equity (n=3) (Akula & Garibay, 2021) to solidarity (n=3) (Sigfrids et al., 2023), humanity (n=1, ‘Other’) (Comes, 2024), or safety (n=2, ‘Other’) (Zhang et al., 2024) receive far less attention, even though they are key to human decision-making, especially in poly-crises. Given the analytical nature of AI research, it may not be surprising that also HCAI research prioritises concepts that can be easily measured or optimised for, such as explainability, fairness or trust. This creates a fundamental problem: explainability metrics become substitutes for accountability; fairness for justice; and trust replaces meaningful human oversight. The result is systems optimised for measurable proxies rather than the deeper values or principles they are supposed to represent. By focusing on what can be optimised for, the field of HCAI may therefore inadvertently reproduce the very reductionism it seeks to address Moving from individual concepts to patterns of co-occurrence, Figure 9 shows that explainability in combination with trust (n=17), fairness (n=8) and situational awareness (n=8) dominates the discussion. At the same time, Figure 9 highlights limited attention to the link between other concepts. Despite the calls for contextualising AI (Sloane et al., 2023) to the environments and social dynamics of its use, the link between contextualisation and principles such as fairness or accountability is under-explored. This is an important gap since many decisions are highly situational as discussed in Section 2. Surprisingly, agency, autonomy and control are less frequently associated with concerns for fairness, accountability and privacy, suggesting a research gap in how we balance the need for efficient, rapid and automated decisions with human oversight and value deliberation. In sum, this scoping review shows that current HCAI focuses on dyadic interactions, measurable principles, and task distribution between humans and machines. What is missing is a deeper discussion and understanding of how optimisation for specific types of interactions and role, tasks, or principles reshapes the very contexts in which we humans make meaning. 5. Discussion: From Human-Centred to Sensemaking AI The results of the scoping review show a fundamental tension: while HCAI research has made important advances in improving individual-level explainability, fairness, or trust, it has yet to grapple with what happens when these optimised interactions scale to complex, networked environments where many humans work with many AI systems. In such environments, optimisation reshapes the informational and social conditions under which actors interpret, coordinate and decide – the terrain of sensemaking itself. This creates an optimisation paradox : the more we optimise individual human-AI interactions, the less equipped we may become to handle the emergent, collective challenges that define complex socio-technical systems. This paradox motivates a reorientation from model‑centric improvements toward AI that sustains collective meaning‑making in evolving human–AI networks. This paradox becomes particularly visible in poly-crises, which are complex, decentralised, fraught with dilemmas and marked by ‘ time running out’ (Comes, 2024; Levin et al., 2012). This combination fundamentally changes human sensemaking and decision-making behaviour, thereby also altering the evolving human-AI-interactions. HCAI research predominantly frames interactions as generic, not contextualised, as dyadic, not networked and as static, not dynamically evolving. The networks involved in crises, however, are inherently dynamic and uncertain, demanding flexible and context-sensitive AI systems (Jennings et al., 2014). Building on these observations, I propose three shifts: Sensemaking‑aware automation, collective agency for network‑level control, and value‑aware sensemaking. Together, these shifts provide a research and design agenda for Sensemaking AI. 5.1 Sensemaking-Aware Automation It is tempting to assume that machines can optimise our meaning-making—make it faster, more efficient, less biased. But automation shapes the trajectories of interpretation by narrowing how problems are framed. Although the review shows the increasing recognition of human-AI-teaming in situational awareness (Endsley, 2023; Gajos & Mamykina, 2022) or developing shared mental models (Hoch et al., 2022; Thompson, 2021), most research focuses on computer-supported information sharing or decision-making. Sensemaking is treated as an individual cognitive process that can be supported through better information presentation, rather than recognising it as a fundamentally collective process of meaning construction that emerges through social interaction and network dynamics. In addition, the AI is seen as a (neutral) tool that mediates human-human interaction (Heyndels, 2023), rather than viewing AI as an actor that optimises information flows and thereby shapes sensemaking and decision-making. As humans rely routinely on AI, however, the relation between a human ‘operator’ and an AI tool supporting the human has become blurry. Automation does not just take over individual tasks or decisions. Rather, automation reshapes how we understand what matters by impacting the process of meaning-making, through which we individually and collectively define and understand our choices. Via sensemaking trajectories, initial perceptions – as moderated or provided by an AI - can become deeply engrained (Comes et al., 2020), leading to path-dependencies that remain hidden within the current task-based frameworks. As AI systems continuously filter, prioritise, and present information, they do not merely ‘support’ decision-making but shape human cognition, reinforcing certain narratives while marginalising others. In essence, sensemaking depends on ambiguity , the interpretive flexibility that enables creative reframing when understanding evolves (Weick, 2015). The question becomes not whether to automate information processing or decision-making, but how to ensure such automation supports rather than constrains the interpretive flexibility that enables creative reframing. A way ahead in achieving interpretive flexibility may be graceful degradation (Ploeg et al., 2014) – designed fallback from higher levels of automation that force systems to slow, expose provenance, invite dissent, and hand decision authority to people if needed. Interpretive flexibility, however, is not an individual cognitive capacity that can be preserved through dyadic human-AI interactions. Sensemaking is fundamentally social—it emerges through collective interaction, shared interpretation, and distributed meaning-making across networks of actors (Weick, 1995). Yet current HCAI research largely overlooks this social dimension. Especially the role of AI in identity construction, by which identity is continuously shaped through social interaction and feedback (Weick, 1995), has not received attention yet. AI, as an actor, influences how humans construct their professional and social identities by mediating access to information, influencing real or perceived agency, and reinforcing or challenging organisational and societal norms. Especially with the increasing personalisation of generative AI, there is a risk that identities become shaped by an algorithm that amplifies biases and leads to echo chambers. When AI systems optimise for engagement, productivity, or efficiency, they reshape the very questions humans ask about meaning and purpose. In addition, there is a growing concern about the risk of de-skilling, where the reliance on AI leads to an erosion of critical thinking (Sellen & Horvitz, 2024). What is more, complex problems are often characterised by moral dilemmas. Yet off-loading morally challenging decisions to a machine may lead to moral de-skilling (Vallor, 2015). As such, the impact of AI on perceived or real responsibility for others, and on the evolution of human competences and skills are a concern that must be addressed in sensemaking aware automation. To address this gap, Sensemaking AI can draw on different bodies of literature as summarised in Figure 10. First, there is wealth of research on group decision-making (Hollingshead et al., 1993), information sharing and sensemaking (Stasser & Titus, 1985) (left box in Figure 10; see also Figure 2) dedicated to how teams and groups differ from individuals, which is largely neglected in the HCAI literature. This gap becomes evident when we consider our finding that only 22 (9+13) out of 101 papers address situations where AI systems interact with multiple humans simultaneously. At the other end of the spectrum, research on collective machine behaviour and multi-AI coordination (right side in Figure 10) focuses on coordinating different artificial agents (Stone et al., 2010). While 13 papers in the review discuss autonomous systems, none of them focuses how the coordination of these systems can or should be designed to benefit humans. Here, the role of trust, loyalty, or cognitive and behavioural factors that are important in human interactions and group decision-making are discarded; networks and groups are formed based on optimal skillsets or available resources. I argue that Sensemaking AI should address the reality of complex networks where multiple types of interactions occur simultaneously: humans engaging in collective sensemaking with each other, humans interacting with multiple AI systems that shape their information environments and thereby their decisions, and AI systems that coordinate or influence each other. This implies that theories of Sensemaking AI recognise that all these interactions constitute a single, complex system where human sensemaking, AI mediation, and algorithmic coordination are fundamentally intertwined and mutually constitutive. This requires integrating theories on group decisions and sensemaking into HCAI research and incorporating task distribution and information prioritisation protocols from collective machine behaviour research. Building on the integrated theoretical framework and recognising that AI fundamentally influences social networks, shared meaning-making and values, we need to then focus on Sensemaking AI as a design principle and ask: which sensemaking, coordination processes and decisions do we want to or need to optimise or automate, and why? Answering this question, especially given the urgency and moral dilemmas pertaining to poly-crises, requires addressing research questions such as: How does AI impact shared identity construction and how can collective moral de-skilling be avoided? How do optimisation algorithms impact collective sensemaking trajectories and thereby shape the interpretive flexibility that is crucial for sensemaking? Answering these questions requires expanding the current theoretical foundations and conduct interdisciplinary research that focuses on large-scale and longitudinal studies on network dynamics as the central focus. Here, longitudinal empirical studies with foundations in group decision making (e.g., information pooling, shared mental models) can be combined with insights from machine behaviour (e.g., prioritisation protocols) and computational models from complexity science to capture the interplay between automation and evolving sensemaking. Experimental designs that measure interpretive flexibility, identity construction and (moral) deskilling before and after the use of an AI in different constellations of groups and teams can create new insights into sensemaking trajectories; based on these empirical insight, agent-based models (Nespeca et al., 2021) can simulate information sharing and sensemaking dynamics to explore path-dependencies and conditions or tipping points that lead to the erosion of interpretive flexibility in networks. 5.2 Collective Agency for Network‑Level Control The discussion about automation is inherently connected to questions of control, autonomy and agency. However, this review shows limited attention for control mechanisms beyond individual human-AI interactions, with concepts like 'control' (n=12) and 'agency' (n=10) receiving far less attention than principles like explainability (n=45), see Figure 8. Even ‘autonomy’ (n=21) is most often discussed in the context of trust (6/21) and fairness (6/21) rather than through the lens of control (4/21), see Figure 9. Current principled approaches to control and accountability overlook the complexity arising from the many diverse interactions of humans and machines. For instance, the OECD guidelines specify that “ AI actors should be accountable for the proper functioning of AI systems” (OECD, 2019). But what if it is precisely the ‘proper functioning’ that leads to undesired consequences or harmful cascading effects? When optimisation decisions propagate through networks of human and AI actors, the “problem of many hands” and subsequent responsibility gaps occur (Matthias, 2004). This problem cannot simply be solved by distributing responsibility (Coeckelbergh, 2020) since it is not clear who bears responsibility when properly functioning optimisation algorithms produce undesirable outcomes at systems level. Preserving human control in such dynamic systems thus requires temporal reflexivity (see Section 2) , i.e., the ability to recognise when optimisation drifts away from the original purpose and to intervene in time to change the unfolding trajectories when needed . Moreover, research has shown that decisions shape physical, informational, and social networks, which in turn influence the information accessible (Comes et al., 2020) to human actors and AI agents. When algorithms optimise traffic and energy flows, financial markets, or entire smart cities, they create optimised environments, in which human choices are increasingly constrained by algorithmic assumptions about what should be optimised, and how. In these contexts, traditional concepts of control—rooted in task allocation and oversight—become inadequate since the challenge is not only controlling what machines do , but preserving spaces for human interaction and meaning-making within emergent, decentralised networks. As such, the question of human control becomes: how can human agency be preserved when optimisation algorithms and AI increasingly dominate information flows and decision architectures? How do decision-information feedback loops influence the long-term evolution of control structures in human-AI networks across spatial and temporal scales? Addressing control in emergent human-AI networks requires moving beyond traditional oversight models toward networked agency. Control theory and cybernetics, originally developed by Wiener (1948), provides the theoretical framework for understanding adaptive regulation: control theory provides a framework for modelling decision loops where human and AI agents dynamically adjust their actions based on new information, constraints, and goals. Cybernetics stresses the need for self-correcting incentives and governance. As such, cybernetics has also been suggested as a way to coordinate decentralised AI networks in autonomic computing (De Wolf & Holvoet, 2003) and more recently as a governance principle for humans and technology (Zwitter, 2024). Rather than centralised monitoring, cybernetic approaches enable network-level self-regulation where control emerges through distributed feedback loops and adaptive responses to changing conditions, e.g., via meta‑signals on uncertainty and impact, circuit breakers for cascading automations, auditability of decision–information feedback loops. This shift reframes control from actions or outcomes to designing self‑correcting conditions , under which collective agency can emerge and evolve with human-AI networks. 5.3 Value-Aware Sensemaking AI: Processes and Boundaries Value-aware Sensemaking AI refers to AI that makes value assumptions explicit and revisable. As such, value-aware Sensemaking AI needs to distinguish between process principles (how values are formed, contested and revised) and content principles. Such systems make value assumptions explicit and revisable, facilitating how values are surfaced, balanced and formalised while preserving space for disagreement if trade-offs violate moral boundaries. The HCAI literature recognises that AI systems need to be designed to “ understand humans ” including the norms and values that govern our actions (Riedl, 2019). Even though Shneiderman (2020) proposed HCAI as a design process , this review shows ‘ human-centred’ has largely become a synonym to explainable, fair, accountable, transparent and trusted AI systems, see Figure 8. These principles are often treated as generic optimisable requirements that systems can be built from irrespective of the context. Maybe not surprisingly, a similar view is presented by the various guidelines, standards and regulatory frameworks for the design and use of AI. The UNESCO recommendations on the Ethics of AI (UNESCO, 2022), the OECD Recommendation of the Council on Artificial Intelligence (OECD, 2019), the European Commission’s recommendations by the High-Level Expert Group on AI (EC, 2019) the IEEE standards for Ethically Aligned Design of Autonomous and Intelligent Systems (IEEE, 2019), and the EU AI Act establish important foundations around transparency, accountability, trust and fairness. However, they all operate under the assumption that values can be translated into generic, stable, measurable principles. This suggests a tendency to optimise for what can be measured while neglecting values that may be essential because they resist quantification. In addition, current HCAI principles assume that all principles can be achieved simultaneously. Yet, there are inherent conflicts across principles or what society values, and these conflicts cannot always be reconciled, e.g., when climate justice collides with economic stability, transparency with privacy, or control with personal freedom. There is an expanding literature that highlights that humans refuse making such trade-offs because they are seen as morally problematic, or taboo (Chorus et al., 2018; Tetlock, 2003). Formalising such trade-offs in an optimisation then risks treating them as commensurable, thereby eroding their role as ethical boundaries. The underlying challenge is: AI principles are viewed in separation from the contextualised objectives, preferences, attitudes, or emotions that drive human sensemaking and decision-making (van Berkel et al., 2022), as well as from the consequences that occur if AI is used at scale. Based on sensemaking theory, I argue that the meaning of key principles depends on continuous collective (re-)interpretation. I do not advocate for abandoning principled approaches, but rather for distinguishing between process principles that support collective meaning-making and democratic deliberation from content principles that may predetermine its outcomes. This shift requires integrating approaches that recognise value formation as an output of human-AI interaction. Social choice ethics for AI design (Baum, 2020) provides a framework for this, emphasising questions of standing ( who participates in value construction?), measurement ( how are diverse perspectives translated into system design?), and aggregation ( how do we coordinate across potentially conflicting value systems?). This process-centered approach creates an important challenge: how to translate the outcomes of collective deliberation into formal specifications without undermining the integrity of the process? To ensure that the results of collective value formation can guide AI development, outcomes need to be linked to formal decision theoretical frameworks that translate values into objective functions and operational trade-offs. Research is needed to formalise abstract goals such as equity (Coleman et al., 2024; Holguin-Veras et al., 2013) and to explore the dynamic nature and structure of trade-offs for instance for intangible or sacred goods via taboo trade-offs (Daw et al., 2015; Lu et al., 2021). At the same time, democratic deliberation may identify domains, values or decisions that cannot be translated, and where preserving ambiguity, maintaining human judgment, and sustaining ongoing deliberation is more important than efficiency. This leads to the question: how can AI systems recognise and respect the boundaries of their own applicability as determined via deliberation? By addressing these challenges, research on Sensemaking AI can combine the cognitive, behavioural, social and ethical elements needed to move towards AI that supports rather than constrains collective meaning-making. Taken together, the three shifts towards Sensemaking AI repositions AI from a tool that optimises towards generic objectives in dyadic relations to an actor that shapes collective cognition in networks. AI and humans, together, are tasked with sustaining interpretive flexibility, networked control and value-awareness at scale. 6. Conclusion Human-Centred AI (HCAI) has been put forward as a paradigm to design AI the supports humans by advocating for design principles ensuring that AI is explainable, fair, transparent and trustworthy. Yet, this scoping review shows that these principles are largely operationalised for dyadic interactions where one human works with one AI, and that the focus is on a relatively narrow set of values that can be readily operationalised. Even though I acknowledge that this review provides only a snapshot, and that the field of (HC)AI is rapidly evolving, this framing is too narrow for the complex and often time-compressed decisions that we are facing today. The ubiquity of AI, and the optimisation of information sharing and processing, reconfigures the informational and social conditions under which humans interpret situations and decide. AI reshapes the way we make sense of our environment. Against this backdrop, this paper advocates for a paradigm shift towards Sensemaking AI: AI that supports collective meaning-making in evolving human-AI networks. Conceptually, the paper synthesises sensemaking and decision theory with literatures on coordination and machine behaviour to characterise AI as an actor within socio‑technical systems. Via a scoping review of the HCAI literature, this paper highlights that current research focuses on individual support and generic, measurable principles. Gaps persist in our understanding of how (human-centred) AI impacts and reshapes sensemaking and decision-making over time. Together, these strands motivate three interconnected directions for Sensemaking AI research and design: Sensemaking-aware automation: AI shapes sensemaking trajectories, reinforcing certain narratives while marginalising others. Research needs to expand beyond dyadic human-AI interactions and integrate group decision and collective machine behaviour theories to understand how automation impacts collective sensemaking in dynamic networks where many humans work with many algorithms. Future research questions include: How does AI-driven automation influence identity construction, (moral) de-skilling and collective meaning-making over time? What mechanisms can mitigate path dependencies, and how to preserve the interpretive flexibility essential for creative reinterpretation in networked systems. Collective agency for networked control: in complex systems, oversight of individual components can never ensure control of the whole system. Therefore, AI understood as an actor in complex networks poses a challenge for conceptualising and maintaining human control and oversight, especially since over time optimisation cascades create decision-information feedback loops. Sensemaking AI reframes control as networked agency. Research needs to analyse how to preserve collective agency when optimisation logic increasingly dominates decision architectures. Drawing on control theory and cybernetics, this requires designing conditions that enable network-level self-regulation through distributed feedback loops and adaptive responses to changing conditions. Value-aware Sensemaking: Current principled AI frameworks lack the adaptability required for complex, dynamic decisions. Recognising that some values—such as dignity, justice, and humanity—resist translation into optimisable metrics and cannot be traded off, Sensemaking AI distinguishes process principles , which support ongoing democratic deliberation and contextualisation , from content principles that risk pre‑empting it. Methodologically, this calls for pipelines that (i) enable participatory formation and revision of objectives, (ii) translate deliberative outcomes into formal decision models where appropriate , and (iii) specify boundaries of applicability where automated optimisation should defer to human judgment and sustain ambiguity. The aim is not to abandon principles, but to embed them in processes that keep values contestable and revisable at scale. The shift towards Sensemaking AI has actionable implications for research and practice: the focus of AI studies has to shift from individual decision-makers to studies that acknowledge AI as embedded in evolving social-technical networks. This requires a shift of methods towards integrating longitudinal, large-scale empirical studies with methods from complexity science to trace how information sharing, analysis, explanation and automation affect interpretive flexibility, sensemaking and coordination over time. AI design should integrate and test mechanisms to ensure a diversity of inputs, allow for surfacing dissent rather than focusing on convergence, and allow to shift from automation to human deliberation when needed. AI governance needs to move towards architectures that integrate feedback mechanisms and incentives, rather than promoting static principles. The contribution of this paper lies in challenging the imperative within current Human-Centred AI literature to optimise AI systems to become more transparent or fair. Instead, Sensemaking AI is proposed as a concrete alternative that recognises AI as an actor within complex socio-technical systems that shapes collective meaning making and decision architectures. As such, Sensemaking AI needs to be designed to sustain the interpretive, social, and ethical capacities on which sensemaking depends. Abbreviations AI: Artificial Intelligence EC: European Commission HCAI: Human-Centred Artificial Intelligence LLM Large Language Model NLP: Natural Language Processing OECD: Organisation for Economic Co-operation and Development PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses Declarations Statements Clinical trial number: not applicable as this is not a clinical study Ethics approval and consent to participate: Not applicable since there was no participation of human subjectis in the study. Consent for publication: Not applicable since this was the work of the author. Availability of data and material: Not applicable. Competing interests: The author has no competing interests to declare that are relevant to the content of this article. Funding: No funding was received to assist with the preparation of this manuscript. Authors' contributions: Tina Comes is the sole author of this manuscript. Acknowledgements: Not applicable. References Abbass, H. A. (2019). Social Integration of Artificial Intelligence: Functions, Automation Allocation Logic and Human-Autonomy Trust. Cognitive Computation , 11 (2), 159-171. https://doi.org/10.1007/s12559-018-9619-0 Acharya, D. B., Kuppan, K., & Divya, B. (2025). Agentic AI: Autonomous Intelligence for Complex Goals—A Comprehensive Survey. IEEE Access , 13 , 18912-18936. https://doi.org/10.1109/ACCESS.2025.3532853 Akata, Z., Balliet, D., Rijke, M. d., Dignum, F., Dignum, V., Eiben, G., Fokkens, A., Grossi, D., Hindriks, K., Hoos, H., Hung, H., Jonker, C., Monz, C., Neerincx, M., Oliehoek, F., Prakken, H., Schlobach, S., Gaag, L. v. d., Harmelen, F. v.,…Welling, M. (2020). A Research Agenda for Hybrid Intelligence: Augmenting Human Intellect With Collaborative, Adaptive, Responsible, and Explainable Artificial Intelligence. Computer , 53 (8), 18-28. https://doi.org/10.1109/MC.2020.2996587 Akula, R., & Garibay, I. (2021, 2021//). Ethical AI for Social Good. HCI International 2021 - Late Breaking Papers: Multimodality, eXtended Reality, and Artificial Intelligence, Cham. Alix, C., Lafond, D., Mattioli, J., Heer, J. D., Chattington, M., & Robic, P. O. (2021, 14-18 June 2021). Empowering Adaptive Human Autonomy Collaboration with Artificial Intelligence. 2021 16th International Conference of System of Systems Engineering (SoSE), Auernhammer, J. (2020). Human-centered AI: The role of Human-centered Design Research in the development of AI. Bansal, G., Wu, T., Zhou, J., Fok, R., Nushi, B., Kamar, E., Ribeiro, M. T., & Weld, D. (2021, 2021). Does the whole exceed its parts? the effect of ai explanations on complementary team performance. CHI, New York. Baum, S. D. (2020). Social choice ethics in artificial intelligence. AI & society , 35 (1), 165-176. https://doi.org/10.1007/s00146-017-0760-1 Bingley, W. J., Haslam, S. A., Steffens, N. K., Gillespie, N., Worthy, P., Curtis, C., Lockey, S., Bialkowski, A., Ko, R. K. L., & Wiles, J. (2023). Enlarging the model of the human at the heart of human-centered AI: A social self-determination model of AI system impact. New Ideas in Psychology , 70 . https://doi.org/doi:10.1016/j.newideapsych.2023.101025 Buçinca, Z. (2024). Optimizing Decision-Maker's Intrinsic Motivation for Effective Human-AI Decision-Making. Extended Abstracts of the CHI Conference on Human Factors in Computing Systems . https://doi.org/info:doi/10.1145/3613905.3638179 Capel, T., & Brereton, M. (2023). What is Human-Centered about Human-Centered AI? A Map of the Research Landscape Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, Hamburg, Germany. https://doi-org.tudelft.idm.oclc.org/10.1145/3544548.3580959 Cavalcante Siebert, L., Lupetti, M. L., Aizenberg, E., Beckers, N., Zgonnikov, A., Veluwenkamp, H., Abbink, D., Giaccardi, E., Houben, G.-J., & Jonker, C. M. (2023). Meaningful human control: actionable properties for AI system development. AI and Ethics , 3 (1), 241-255. Chaudhry, M. A., Cukurova, M., & Luckin, R. (2022, 2022//). A Transparency Index Framework for AI in Education. Artificial Intelligence in Education. Posters and Late Breaking Results, Workshops and Tutorials, Industry and Innovation Tracks, Practitioners’ and Doctoral Consortium, Cham. Chorus, C. G., Pudāne, B., Mouter, N., & Campbell, D. (2018). Taboo trade-off aversion: A discrete choice model and empirical analysis. Journal of Choice Modelling , 27 , 37-49. https://doi.org/https://doi.org/10.1016/j.jocm.2017.09.002 Coeckelbergh, M. (2020). Artificial intelligence, responsibility attribution, and a relational justification of explainability. Science and engineering ethics , 26 (4), 2051-2068. Coleman, N., Li, X., Comes, T., & Mostafavi, A. (2024). Weaving equity into infrastructure resilience research: a decadal review and future directions. npj Natural Hazards , 1 (1), 25. https://doi.org/10.1038/s44304-024-00022-x Comes, T. (2024). AI for crisis decisions. Ethics and Information Technology , 26 (1), 12. https://doi.org/10.1007/s10676-024-09750-0 Comes, T., Van de Walle, B., & Van Wassenhove, L. (2020). The coordination‐information bubble in humanitarian response: theoretical foundations and empirical investigations. Production and Operations Management , 29 (11), 2484-2507. Cooke, N., Demir, M., & Huang, L. (2020, 2020). A framework for human-autonomy team research. HCII 2020, Copenhagen, Denmark. Coppi, G., Moreno Jimenez, R., & Kyriazi, S. (2021). Explicability of humanitarian AI: a matter of principles. Journal of International Humanitarian Action , 6 (1), 19. Council, N. R. (1998). The future of air traffic control: Human operators and automation (C. Wickens, A. Mavor, R. Parasuraman, & J. McGee, Eds.). National Academies Press. Criscuolo, C., & Dolci, T. (2024, 13-16 May 2024). Exploring Fairness Interpretability with FairnessFriend: A Chatbot Solution. 2024 IEEE 40th International Conference on Data Engineering Workshops (ICDEW), Daw, T. M., Coulthard, S., Cheung, W. W. L., Brown, K., Abunge, C., Galafassi, D., Peterson, G. D., McClanahan, T. R., Omukoto, J. O., & Munyi, L. (2015). Evaluating taboo trade-offs in ecosystems services and human well-being. Proceedings of the National Academy of Sciences , 112 (22), 6949-6954. https://doi.org/10.1073/pnas.1414900112 De Wolf, T., & Holvoet, T. (2003, 2003). Towards autonomic computing: agent-based modelling, dynamical systems analysis, and decentralised control. Duan, X., Pei, B., Ambrose, G. A., Hershkovitz, A., Cheng, Y., & Wang, C. (2024). Towards transparent and trustworthy prediction of student learning achievement by including instructors as co-designers: a case study. Education and Information Technologies , 29 (3), 3075-3096. https://doi.org/10.1007/s10639-023-11954-8 EC. (2019). Ethics guidelines for trustworthy AI . https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai Elahi, H., Castiglione, A., Wang, G., & Geman, O. (2021). A human-centered artificial intelligence approach for privacy protection of elderly App users in smart cities. Neurocomputing , 444 , 189-202. https://doi.org/https://doi.org/10.1016/j.neucom.2020.06.149 Endsley, M. R. (2017). From here to autonomy: lessons learned from human–automation research. Human factors , 59 (1), 5-27. Endsley, M. R. (2023). Supporting Human-AI Teams:Transparency, explainability, and situation awareness. Computers in Human Behavior , 140 , 107574. https://doi.org/https://doi.org/10.1016/j.chb.2022.107574 Erlei, A. (2024). Understanding Choice Independence and Error Types in Human-AI Collaboration. Proceedings of the CHI Conference on Human Factors in Computing Systems . https://doi.org/info:doi/10.1145/3613904.3641946 Fan, C., Zhang, C., Yahja, A., & Mostafavi, A. (2021). Disaster City Digital Twin: A vision for integrating artificial and human intelligence for disaster management. International Journal of Information Management , 56 , 102049. Fitts, P. M. (1951). Human engineering for an effective air-navigation and traffic-control system. Flathmann, C., Schelble, B. G., Rosopa, P. J., McNeese, N. J., Mallick, R., & Madathil, K. C. (2023). Examining the impact of varying levels of AI teammate influence on human-AI teams. International Journal of Human Computer Studies , 177 . https://doi.org/doi:10.1016/j.ijhcs.2023.103061 Freire, S. K., Niforatos, E., Wang, C., Ruiz-Arenas, S., Foosherian, M., Wellsandt, S., & Bozzon, A. (2023). Lessons Learned from Designing and Evaluating CLAICA: A Continuously Learning AI Cognitive Assistant Proceedings of the 28th International Conference on Intelligent User Interfaces, Sydney, NSW, Australia. https://doi-org.tudelft.idm.oclc.org/10.1145/3581641.3584042 French, S. (1986). Decision Theory: An introduction of the Mathematics of Rationality . Ellis Harwood Limited. French, S. (2012). Cynefin, statistics and decision analysis. Journal of the Operational Research Society . https://doi.org/10.1057/jors.2012.23 Gajos, K. Z., & Mamykina, L. (2022). Do People Engage Cognitively with AI? Impact of AI Assistance on Incidental Learning Proceedings of the 27th International Conference on Intelligent User Interfaces, Helsinki, Finland. https://doi-org.tudelft.idm.oclc.org/10.1145/3490099.3511138 Gil, Y., Garijo, D., Khider, D., Knoblock, C. A., Ratnakar, V., Osorio, M., Vargas, H., Pham, M., Pujara, J., & Shbita, B. (2021). Artificial intelligence for modeling complex systems: taming the complexity of expert models to improve decision making. ACM Transactions on Interactive Intelligent Systems , 11 (2), 1-49. Gou, J., Liang, Q., Wang, Z., & Dabić, M. (2024). Affordances and Constraints of Automation and Augmentation: Lessons Learned From Development of a Human-AI Collaboration Business Simulation Platform. Journal of Global Information Management , 32 (1), 1-27. https://doi.org/https://doi.org/10.4018/JGIM.357260 Gralla, E., Goentzel, J., & Fine, C. (2016). Problem Formulation and Solution Mechanisms : A Behavioral Study of Humanitarian Transportation Planning. POMS , 25 (1), 22-35. https://doi.org/10.1111/poms.12496 Hagemann, V., Rieth, M., Suresh, A., & Kirchner, F. (2023). Human-AI teams—Challenges for a team-centered AI at work. Frontiers in Artificial Intelligence , 6 . https://doi.org/doi:10.3389/frai.2023.1252897 He, J., Piorkowski, D., Muller, M. J., Brimijoin, K., Houde, S., & Weisz, J. D. (2023, 2023). Understanding How Task Dimensions Impact Automation Preferences with a Conversational Task Assistant. CHI AutomationXP, Hamburg. Helms Mills, J., Thurlow, A., & Mills, A. J. (2010). Making sense of sensemaking: the critical sensemaking approach. Qualitative Research in Organizations and Management: An International Journal , 5 (2), 182-195. https://doi.org/10.1108/17465641011068857 Herrmann, T., & Pfeiffer, S. (2023). Keeping the organization in the loop: a socio-technical extension of human-centered artificial intelligence. AI & SOCIETY , 38 (4), 1523-1542. https://doi.org/10.1007/s00146-022-01391-5 Heyndels, S. (2023). Technology and neutrality. Philosophy & Technology , 36 (4), 75. Hoch, T., Heinzl, B., Czech, G., Khan, M., Waibel, P., Bachhofner, S., Kiesling, E., & Moser, B. (2022). Teaming. AI: enabling human-AI teaming intelligence in manufacturing I-ESA, Valencia, Spain. Holguin-Veras, J., Perez, N., Jaller, M., Van Wassenhove, L. N., & Aros-Vera, F. (2013). On the appropriate objective function for post-disaster humanitarian logistics models. Journal of Operations Management , 31 (5), 262-280. https://doi.org/10.1016/j.jom.2013.06.002 Hollingshead, A. B., McGrath, J. E., & O'Connor, K. M. (1993). Group Task Performance and Communication Technology: A Longitudinal Study of Computer-Mediated Versus Face-to-Face Work Groups. Small Group Research , 24 (3), 307-333. https://doi.org/10.1177/1046496493243003 IEEE. (2019). Ethically Aligned Design - A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems. In Ethically Aligned Design - A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems (pp. 1-294): IEEE. Jacob, C., Kerrigan, P., & Bastos, M. (2025). The chat-chamber effect: Trusting the AI hallucination. Big Data & Society , 12 (1), 20539517241306345. Jennings, N. R., Moreau, L., Nicholson, D., Ramchurn, S., Roberts, S., Rodden, T., & Rogers, A. (2014). Human-agent collectives. Communications of the ACM , 57 (12), 80-88. Jin, L., Boden, A., & Shajalal, M. (2022). Automated Decision Making Systems in Smart Homes: A Study on User Engagement and Design (Vol. 3154). https://doi.org/doi: Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica , 47 (2), 263-291. Kattnig, M., Angerschmid, A., Reichel, T., & Kern, R. (2024). Assessing trustworthy AI: Technical and legal perspectives of fairness in AI. Computer Law & Security Review , 55 , 106053. https://doi.org/https://doi.org/10.1016/j.clsr.2024.106053 Kiesler, S., & Sproull, L. (1992). Group decision making and communication technology. Organizational Behavior and Human Decision Processes , 52 (1), 96-123. https://doi.org/https://doi.org/10.1016/0749-5978(92)90047-B Kim, J. S., Kim, M., & Baek, T. H. (2025). Enhancing User Experience With a Generative AI Chatbot. International Journal of Human–Computer Interaction , 41 (1), 651-663. https://doi.org/10.1080/10447318.2024.2311971 Kjærum, A., & Madsen, B. S. (2025). Pushing the boundaries of anticipatory action using machine learning. Data & Policy , 7 , e8. Klein, G., Calderwood, R., & Clinton-Cirocco, A. (2010). Rapid Decision Making on the Fire Ground: The Original Study Plus a Postscript. Journal of Cognitive Engineering and Decision Making , 4 (3), 186-209. https://doi.org/10.1518/155534310X12844000801203 Kunar, M. A., Montana, G., & Watson, D. G. (2024). Increasing transparency of computer-aided detection impairs decision-making in visual search. Psychonomic Bulletin & Review . https://doi.org/10.3758/s13423-024-02601-5 Lee, M. H., Siewiorek, D. P., Smailagic, A., Bernardino, A., & Bermúdez i Badia, S. (2022, 2022). Towards efficient annotations for a human-ai collaborative, clinical decision support system: A case study on physical stroke rehabilitation assessment. IUI, Helsinki, Finland. Lee, M. K., Kusbit, D., Kahng, A., Kim, J. T., Yuan, X., Chan, A., See, D., Noothigattu, R., Lee, S., Psomas, A., & Procaccia, A. D. (2019). Webuildai: Participatory framework for algorithmic governance. Proceedings of the ACM on Human-Computer Interaction , 3 . https://doi.org/doi:10.1145/3359283 Lettieri, N., Guarino, A., Zaccagnino, R., & Malandrino, D. (2023). Keeping judges in the loop: a human–machine collaboration strategy against the blind spots of AI in criminal justice. Soft Computing , 27 (16), 11275-11293. https://doi.org/10.1007/s00500-023-08604-z Levin, K., Cashore, B., Bernstein, S., & Auld, G. (2012). Overcoming the tragedy of super wicked problems: constraining our future selves to ameliorate global climate change. Policy Sciences , 45 (2), 123-152. Li, X., Zong, Q., & Cheng, M. (2024). The Impact of Medical Explainable Artificial Intelligence on Nurses’ Innovation Behaviour: A Structural Equation Modelling Approach. Journal of Nursing Management , 2024 (1), 8885760. https://doi.org/https://doi.org/10.1155/2024/8885760 Liao, Q. V., & Sundar, S. S. (2022). Designing for Responsible Trust in AI Systems: A Communication Perspective Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, Seoul, Republic of Korea. https://doi-org.tudelft.idm.oclc.org/10.1145/3531146.3533182 Lu, N., Liu, L., Yu, D., & Fu, B. (2021). Navigating trade-offs in the social-ecological systems. Current Opinion in Environmental Sustainability , 48 , 77-84. https://doi.org/https://doi.org/10.1016/j.cosust.2020.10.014 Mahajan, S., Hausladen, C. I., Sánchez-Vaquerizo, J. A., Korecki, M., & Helbing, D. (2022). Participatory resilience: Surviving, recovering and improving together. Sustainable Cities and Society , 83 , 103942. Maiden, N., Lockerbie, J., Zachos, K., Wolf, A., & Brown, A. (2023). Designing new digital tools to augment human creative thinking at work: An application in elite sports coaching. Expert Systems , 40 (3), e13194. https://doi.org/https://doi.org/10.1111/exsy.13194 Maitlis, S., & Christianson, M. (2014). Sensemaking in Organizations: Taking Stock and Moving Forward. The Academy of Management Annals , 8 (1), 57-125. https://doi.org/10.1080/19416520.2014.873177 Malone, T. W., Malone, T. W., Crowston, K., & Crowston, K. (1994). The interdisciplinary study of coordination. ACM Computing Surveys , 26 (1), 87-119. https://doi.org/http://doi.acm.org/10.1145/174666.174668 Matthias, A. (2004). The responsibility gap: Ascribing responsibility for the actions of learning automata. Ethics and Information Technology , 6 (3), 175-183. https://doi.org/10.1007/s10676-004-3422-1 Muhren, W. J., & Van de Walle, B. (2010). A Call for Sensemaking Support Systems in Crisis Management. In R. Babuška & F. C. A. Groen (Eds.), Interactive Collaborative Information Systems (pp. 425-452). Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-11688-9_16 Muller, J. (2018). The tyranny of metrics . Princeton University Press. Munn, Z., Peters, M. D. J., Stern, C., Tufanaru, C., McArthur, A., & Aromataris, E. (2018). Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Medical Research Methodology , 18 (1), 143. https://doi.org/10.1186/s12874-018-0611-x Nabizadeh Rafsanjani, H., & Nabizadeh, A. H. (2023). Towards human-centered artificial intelligence (AI) in architecture, engineering, and construction (AEC) industry. Computers in Human Behavior Reports , 11 . https://doi.org/doi:10.1016/j.chbr.2023.100319 Neale, D. C., Carroll, J. M., & Rosson, M. B. (2004). Evaluating computer-supported cooperative work: models and frameworks Proceedings of the 2004 ACM conference on Computer supported cooperative work, Chicago, Illinois, USA. https://doi-org.tudelft.idm.oclc.org/10.1145/1031607.1031626 Nespeca, V., Comes, T., & Brazier, F. (2021). A methodology to develop agent-based models for policy design in socio-technical systems based on qualitative inquiry. In M. Czupryna & B. Kamiński (Eds.), Social Simulation Conference 2021 . Springer. OECD. (2019). Recommendation of the Council on Artificial Intelligence . https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449 Ozmen Garibay, O., Winslow, B., Andolina, S., Antona, M., Bodenschatz, A., Coursaris, C., Falco, G., Fiore, S. M., Garibay, I., Grieman, K., Havens, J. C., Jirotka, M., Kacorri, H., Karwowski, W., Kider, J., Konstan, J., Koon, S., Lopez-Gonzalez, M., Maifeld-Carucci, I.,…Xu, W. (2023). Six Human-Centered Artificial Intelligence Grand Challenges. International Journal of Human–Computer Interaction , 39 (3), 391-437. https://doi.org/10.1080/10447318.2022.2153320 Parasuraman, R., Sheridan, T. B., & Wickens, C. D. (2000). A model for types and levels of human interaction with automation. IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans , 30 (3), 286-297. https://doi.org/10.1109/3468.844354 Passlack, N., Hammerschmidt, T., & Posegga, O. (2024). It was(n’t) me: vignette experiments on managers’ responsibility attribution in AI-advised decision-making. Behaviour & Information Technology , 1-30. https://doi.org/10.1080/0144929X.2024.2431050 Paulus, D., Fathi, R., Fiedrich, F., de Walle, B. V., & Comes, T. (2022). On the Interplay of Data and Cognitive Bias in Crisis Information Management. Information Systems Frontiers . https://doi.org/10.1007/s10796-022-10241-0 Ploeg, J., Semsar-Kazerooni, E., Lijster, G., Van de Wouw, N., & Nijmeijer, H. (2014). Graceful degradation of cooperative adaptive cruise control. IEEE Transactions on Intelligent Transportation Systems , 16 (1), 488-497. Ponti, M., & Seredko, A. (2022). Human-machine-learning integration and task allocation in citizen science. Humanities and Social Sciences Communications , 9 (1), 48. https://doi.org/10.1057/s41599-022-01049-z Rahwan, I., Cebrian, M., Obradovich, N., Bongard, J., Bonnefon, J.-F., Breazeal, C., Crandall, J. W., Christakis, N. A., Couzin, I. D., & Jackson, M. O. (2019). Machine behaviour. Nature , 568 (7753), 477-486. Riedl, M. O. (2019). Human‐centered artificial intelligence and machine learning. Human behavior and emerging technologies , 1 (1), 33-36. Rittel, H. W. J., & Webber, M. M. (1973). Dilemmas in a general theory of planning. Policy Sciences , 4 (2), 155-169. Sadek, M., Kallina, E., Bohné, T., Mougenot, C., Calvo, R. A., & Cave, S. (2024). Challenges of responsible AI in practice: scoping review and recommended actions. AI & society . https://doi.org/10.1007/s00146-024-01880-9 Sætre, A. S., & Van de Ven, A. (2021). Generating theory by abduction. Academy of Management Review , 46 (4), 684-701. Seidel, S., Chandra Kruse, L., Székely, N., Gau, M., & Stieger, D. (2018). Design principles for sensemaking support systems in environmental sustainability transformations. European Journal of Information Systems , 27 (2), 221-247. https://doi.org/10.1057/s41303-017-0039-0 Sellen, A., & Horvitz, E. (2024). The Rise of the AI Co-Pilot: Lessons for Design from Aviation and Beyond. Commun. ACM , 67 (7), 18–23. https://doi.org/10.1145/3637865 Sharma, N., Liao, Q. V., & Xiao, Z. (2024). Generative Echo Chamber? Effect of LLM-Powered Search Systems on Diverse Information Seeking Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, Honolulu, HI, USA. https://doi-org.tudelft.idm.oclc.org/10.1145/3613904.3642459 Sharoda, P. A., & Reddy, M. C. (2010). Understanding Together: Sensemaking in Collaborative Information Seeking CSCW, Sheridan, T. B., Verplank, W. L., & Brooks, T. L. (1978, 1978). Human/computer control of undersea teleoperators. Shneiderman, B. (2020). Human-centered artificial intelligence: Reliable, safe & trustworthy. International Journal of Human–Computer Interaction , 36 (6), 495-504. Shulner-Tal, A., Kuflik, T., & Kliger, D. (2023). Enhancing Fairness Perception–Towards Human-Centred AI and Personalized Explanations Understanding the Factors Influencing Laypeople’s Fairness Perceptions of Algorithmic Decisions. International Journal of Human-Computer Interaction , 39 (7), 1455-1482. https://doi.org/doi:10.1080/10447318.2022.2095705 Sigfrids, A., Leikas, J., Salo-Pöntinen, H., & Koskimies, E. (2023). Human-centricity in AI governance: A systemic approach [Perspective]. Frontiers in Artificial Intelligence , 6 . https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2023.976887 Sloane, M., Solano-Kamaiko, I. R., Yuan, J., Dasgupta, A., & Stoyanovich, J. (2023). Introducing contextual transparency for automated decision systems. Nature Machine Intelligence , 5 (3), 187-195. https://doi.org/10.1038/s42256-023-00623-7 Søgaard Jørgensen, P., Jansen, R. E. V., Avila Ortega, D. I., Wang-Erlandsson, L., Donges, J. F., Österblom, H., Olsson, P., Nyström, M., Lade, S. J., & Hahn, T. (2024). Evolution of the polycrisis: Anthropocene traps that challenge global sustainability. Philosophical Transactions of the Royal Society B , 379 (1893), 20220261. Sprenkamp, K., Dolata, M., Schwabe, G., & Zavolokina, L. (2025). Data-driven intelligence in crisis: The case of Ukrainian refugee management. Government Information Quarterly , 42 (1), 101978. https://doi.org/https://doi.org/10.1016/j.giq.2024.101978 Stapleton, L., Lee, M. H., Qing, D., Wright, M., Chouldechova, A., Holstein, K., Wu, Z. S., & Zhu, H. (2022). Imagining new futures beyond predictive systems in child welfare: A qualitative study with impacted stakeholders Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, Seoul, Republic of Korea. https://doi-org.tudelft.idm.oclc.org/10.1145/3531146.3533177 Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision making: Biased information sampling during discussion. Journal of Personality and Social Psychology , 48 (6), 1467-1478. https://doi.org/10.1037/0022-3514.48.6.1467 Stewart, D. D., & Stasser, G. (1995). Expert role assignment and information sampling during collective recall and decision making. Journal of Personality and Social Psychology , 69 (4), 619-628. Stone, P., Kaminka, G., Kraus, S., & Rosenschein, J. (2010, 2010). Ad hoc autonomous agent teams: Collaboration without pre-coordination. Suchan, J., Bhatt, M., & Varadarajan, S. (2021). Commonsense visual sensemaking for autonomous driving – On generalised neurosymbolic online abduction integrating vision and semantics. Artificial Intelligence , 299 . https://doi.org/doi:10.1016/j.artint.2021.103522 Sun, J. (2022). Investigating Explainability of Generative AI for Code through Scenario-based Design. 27th International Conference on Intelligent User Interfaces . https://doi.org/info:doi/10.1145/3490099.3511119 Tausch, A., & Kluge, A. (2022). The best task allocation process is to decide on one’s own: effects of the allocation agent in human–robot interaction on perceived work characteristics and satisfaction. Cognition, Technology & Work , 24 (1), 39-55. https://doi.org/10.1007/s10111-020-00656-7 Tetlock, P. E. (2003). Thinking the unthinkable: sacred values and taboo cognitions. Trends in Cognitive Sciences , 7 (7), 320-324. https://doi.org/https://doi.org/10.1016/S1364-6613(03)00135-9 Thompson, J. (2021). Mental Models and Interpretability in AI Fairness Tools and Code Environments. In C. Stephanidis, M. Kurosu, J. Y. C. Chen, G. Fragomeni, N. Streitz, S. i. Konomi, H. Degen, & S. Ntoa, HCI International 2021 - Late Breaking Papers: Multimodality, eXtended Reality, and Artificial Intelligence Cham. Tricco, A. C., Lillie, E., Zarin, W., O'Brien, K. K., Colquhoun, H., Levac, D., Moher, D., Peters, M. D. J., Horsley, T., & Weeks, L. (2018). PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Annals of internal medicine , 169 (7), 467-473. UNESCO. (2022). Recommendation on the Ethics of Artificial Intelligence . https://unesdoc.unesco.org/ark:/48223/pf0000381137 Vallor, S. (2015). Moral Deskilling and Upskilling in a New Machine Age: Reflections on the Ambiguous Future of Character. Philosophy & Technology , 28 (1), 107-124. https://doi.org/10.1007/s13347-014-0156-9 Van Berkel, N., Bellio, M., Skov, M. B., & Blandford, A. (2023). Measurements, algorithms, and presentations of reality: Framing interactions with AI-enabled decision support. ACM Transactions on Computer-Human Interaction , 30 (2), 1-33. van Berkel, N., Tag, B., Goncalves, J., & Hosio, S. (2022). Human-centred artificial intelligence: a contextual morality perspective. Behaviour & Information Technology , 41 (3), 502-518. https://doi.org/10.1080/0144929X.2020.1818828 van Leersum, C. M., & Maathuis, C. (2025). Human centred explainable AI decision-making in healthcare. Journal of Responsible Technology , 21 , 100108. https://doi.org/https://doi.org/10.1016/j.jrt.2025.100108 Verma, H., Mlynar, J., Schaer, R., Reichenbach, J., Jreige, M., Prior, J., Evéquoz, F., & Depeursinge, A. (2023, 2023). Rethinking the role of AI with physicians in oncology: revealing perspectives from clinical and research workflows. CHI Human Factors, Volkema, R. J. (1983). Problem Formulation in Planning and Design. Management Science , 29 (6), 639-652. https://doi.org/10.1287/mnsc.29.6.639 Webster, M. (2008). Incorporating path dependency into decision-analytic methods: An application to global climate-change policy. Decision Analysis , 5 (2), 60-75. Weick, K., & Sutcliffe, K. (2007). Managing the unexpected: resilient performance in an age of uncertainty, second edition. http://dl.acm.org/citation.cfm?id=1408051 Weick, K. E. (1993). The Collapse of Sensemaking in Organizations: The Mann Gulch Disaster. Administrative Science Quarterly , 38 (4), 628-652. Weick, K. E. (1995). Sensemaking in organizations (Vol. 3). Sage publications Thousand Oaks, CA. Weick, K. E. (2015). Ambiguity as grasp: The reworking of sense. Journal of Contingencies and Crisis Management , 23 (2), 117-123. Wiener, N. (1948). Cybernetics. Scientific American , 179 (5), 14-19. Yazdanpanah, V., Gerding, E. H., Stein, S., Dastani, M., Jonker, C. M., & Norman, T. J. (2021). Responsibility Research for Trustworthy Autonomous Systems AAMAS, Online. Zhang, R., Flathmann, C., Musick, G., Schelble, B., McNeese, N. J., Knijnenburg, B., & Duan, W. (2024). I Know This Looks Bad, But I Can Explain: Understanding When AI Should Explain Actions In Human-AI Teams. ACM Trans. Interact. Intell. Syst. , 14 (1), Article 6. https://doi.org/10.1145/3635474 Zwitter, A. (2024). Cybernetic governance: implications of technology convergence on governance convergence. Ethics and Information Technology , 26 (2), 24. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 19 Mar, 2026 Read the published version in EPJ Data Science → Version 1 posted Editorial decision: Revision requested 23 Oct, 2025 Reviews received at journal 21 Oct, 2025 Reviews received at journal 14 Oct, 2025 Reviews received at journal 02 Oct, 2025 Reviewers agreed at journal 22 Sep, 2025 Reviewers agreed at journal 17 Sep, 2025 Reviewers agreed at journal 11 Sep, 2025 Reviewers invited by journal 04 Sep, 2025 Editor assigned by journal 02 Sep, 2025 Submission checks completed at journal 01 Sep, 2025 First submitted to journal 26 Aug, 2025 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-7463619","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":512950700,"identity":"ea573c52-f0f4-456a-99bf-244f1a36e0a1","order_by":0,"name":"Tina Comes","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAvUlEQVRIiWNgGAWjYBADOSA2YHhgA6TYCSjlgdLGPCAtCWkGDAzMRGpJ7CFaiz0D88NHNyrq0vdLJG98kJDwh8GcsC1sxsY5Zw7n9kikFRskJBgwWDYT1MLDJp3bdgCoJcdMIvGHAYPBYaK0/KtL55HIMf8BsoVILQ3MCUAtZgzEaTkM8suxw4Y9Z54VSyQkGPMQ1MLe3vzwcU5NnTx7e/LGDx8S5OQMjjcQ0IMepDxYVY2CUTAKRsEoIA0AAF+lNuZCg+8bAAAAAElFTkSuQmCC","orcid":"","institution":"Delft University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Tina","middleName":"","lastName":"Comes","suffix":""}],"badges":[],"createdAt":"2025-08-26 13:53:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7463619/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7463619/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1140/epjds/s13688-026-00634-5","type":"published","date":"2026-03-19T15:58:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":91088396,"identity":"ecd97196-731d-4dde-a760-a7b3754dc07b","added_by":"auto","created_at":"2025-09-11 12:39:39","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":58996,"visible":true,"origin":"","legend":"\u003cp\u003eThe need to understand evolving human-AI networks\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/6460e60fc16b34a9c8b4cf8b.png"},{"id":91088699,"identity":"4a33be28-4817-4b40-a872-2512cc88fe51","added_by":"auto","created_at":"2025-09-11 12:47:39","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":94828,"visible":true,"origin":"","legend":"\u003cp\u003eThe cycle of sensemaking and decision-making. Comparing human-centred (left) and fully automated processes (right).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/69319144ffd1990b7471fe09.png"},{"id":91088397,"identity":"59096932-a219-4805-aa36-ecff62e8442c","added_by":"auto","created_at":"2025-09-11 12:39:39","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":51673,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA-Scoping Review Protocol followed for this study\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/b891470e79daeb8149d18ab9.png"},{"id":91088429,"identity":"4407be36-c492-4771-b610-36c8c7d347fd","added_by":"auto","created_at":"2025-09-11 12:39:40","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":32287,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of research methods across the analysed papers (n=101)\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/1679f3fc07a75712d072f79a.png"},{"id":91089650,"identity":"c9379643-a98b-4d9a-85c0-a0caae090928","added_by":"auto","created_at":"2025-09-11 12:55:39","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":25944,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of the different types of AI in the HCAI literature.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/954191fc1e456dee8060c706.png"},{"id":91088707,"identity":"b8bbd1fd-4feb-4e0e-8896-5bc047f111f8","added_by":"auto","created_at":"2025-09-11 12:47:39","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":45279,"visible":true,"origin":"","legend":"\u003cp\u003eHCAI is increasingly prominent in a diversity of fields of application\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/7dd8a3f137d6fbc5e61753a0.png"},{"id":91088399,"identity":"92133aba-adc5-4d27-9c5e-6a46ffddc3f3","added_by":"auto","created_at":"2025-09-11 12:39:39","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":31638,"visible":true,"origin":"","legend":"\u003cp\u003eHuman-Centred AI and the role of AI. Papers that discuss multiple roles of AI are counted multiple times.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/ab9a6d8389b420d28953ef85.png"},{"id":91088401,"identity":"eca3d173-8eec-41ea-bcb7-6007bdd170cb","added_by":"auto","created_at":"2025-09-11 12:39:39","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":51623,"visible":true,"origin":"","legend":"\u003cp\u003eFrequency of concepts in HCAI papers. Showing concepts with at least three mentions across the 101 papers\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/57731e46ce3d7d1d9f3af3a1.png"},{"id":91088705,"identity":"63d02325-d3c9-4d04-b59a-75795632ebb6","added_by":"auto","created_at":"2025-09-11 12:47:39","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":105677,"visible":true,"origin":"","legend":"\u003cp\u003eHeatmap showing co-occurrence of concepts related to HCAI\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/8496ba88174a835541fcf792.png"},{"id":105223287,"identity":"0112e52c-3e7b-4d56-bf12-af95c7183337","added_by":"auto","created_at":"2026-03-23 16:02:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1821214,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7463619/v1/5930aba0-cfa8-43eb-b2ec-99f377bb4f1a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eSensemaking AI: Introducing a Research and Design Agenda for Human–AI Networks\u003c/p\u003e","fulltext":[{"header":"1.\tIntroduction","content":"\u003cp\u003eOur societies are complex dynamical systems increasingly shaped by digital technologies. As decision-makers grapple with increasing complexity, AI is portrayed as a ‘solution’ for its ability to process vast amounts of data and reduce the cognitive biases in human decisions by providing ‘objective’ or ‘data-driven’ optimal answers. Digital twins (Fan et al., 2021), automated decisions (Coppi et al., 2021), and predictive models for ‘anticipatory action’ (Kjærum \u0026amp; Madsen, 2025) promise efficient, scalable, rapid and cheap solutions to address complex and time-compressed problems.\u003c/p\u003e\n\u003cp\u003eWith the promise of ‘taming complexity’ (Gil et al., 2021), AI has become both a tool to understand complex systems and a driver that shapes their behaviour. As such, AI systems do not just optimise a system: they shape the\u0026nbsp;informational and social environments via which humans understand their environment and respond to the challenges they are confronted with. Recognising this dual nature, this paper echoes the calls for \u003cem\u003eMachine Behaviour\u003c/em\u003e (Rahwan et al., 2019), \u003cem\u003eHuman-Centered AI\u003c/em\u003e (HCAI)\u0026nbsp;(Ozmen Garibay et al., 2023)\u0026nbsp;and \u003cem\u003eHybrid Intelligence\u003c/em\u003e (Akata et al., 2020)\u0026nbsp;that all conceptualise AI as a social-technical system. Especially HCAI\u0026nbsp;has emerged as a prominent field advocating for explainable, fair, and transparent AI systems that keep humans “in the loop” rather than replacing them\u0026nbsp;(Shneiderman, 2020). This human-centred approach seeks to address the optimisation paradox by designing AI that enhances rather than diminishes human agency.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eYet HCAI is based on idealised situations that insufficiently\u0026nbsp;address the time pressure, high stakes, volatility, moral dilemmas and shifting networks typical for decisions in today’s complex problems – even though these conditions have been shown to fundamentally change sensemaking and lead decision-makers to overlook, discard, or not act upon information (Paulus et al., 2022). In complex problems, where multiple interconnected challenges demand rapid coordination across scales and domains (Mahajan et al., 2022) this lack of understanding may lead to overlooking important implications of the use of AI, which in turn jeopardises the central tenet of meaningful human control (Cavalcante Siebert et al., 2023).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSensemaking\u003c/em\u003e offers a crucial lens and theoretical framework to address this tension because it focuses on how humans collectively interpret ambiguous situations, construct meaning from uncertain information, and adapt as understanding evolves (Weick, 1995). Unlike optimisation approaches with predetermined objectives, sensemaking embraces the continuous social processes through which values and objectives emerge and evolve. This is essential when dealing with complex problems that can only be addressed by decentralised interaction rather than top-down optimisation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis paper contributes in three ways to the ongoing discourse on optimisation and AI: building on foundations in sensemaking support systems (Muhren \u0026amp; Van de Walle, 2010; Seidel et al., 2018) and decision theory (French, 1986), it introduces \u003cem\u003eSensemaking AI\u003c/em\u003e as \u003cem\u003eAI that actively supports sensemaking processes in evolving human-AI-networks\u003c/em\u003e, and delineates it from other paradigms such as HCAI (Shneiderman, 2020) or machine behaviour (Rahwan et al., 2019). Empirically, a scoping review of 101 academic articles from the Human-Centred AI (HCAI) literature examines how, and to what extent, current HCAI research addresses the dynamics of human-AI interaction in complex, time-sensitive, and morally fraught environments. Poly-crises serve here as illustrative cases to highlight the limitations of optimisation-centric AI design. P\u003cstrong\u003erogrammatically,\u0026nbsp;\u003c/strong\u003ethe paper then develops design principles and a research agenda on sensemaking AI along three axes: sensemaking-aware automation that promotes interpretive flexibility; collective agency for network-level control; and value-aware sensemaking AI. As such, this paper is meant to inspire the reader and start a conversation across computer, cognitive and behavioural sciences that acknowledges that fundamentally we humans are social beings who continuously seek to make sense of our environment.\u0026nbsp;\u003c/p\u003e"},{"header":"2.\tOn the Impact of AI on Sensemaking and Decision-Making ","content":"\u003cp\u003eThis section provides the theoretical foundations for understanding how AI transforms human sensemaking and decision-making. I first provide insights into Sensemaking as a theory and process for navigating complexity. Then, I explore the foundations of Human-Centred AI to showcase how the notion of control is shifting with the ubiquity of AI and the complexity of the problem addressed. From there, I move to sketch how AI systems reshape information-decision feedback dynamics in networked social-technical systems.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e2.1 Sensemaking\u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eWe are confronted with several accelerating poly-crises (S\u0026oslash;gaard J\u0026oslash;rgensen et al., 2024). Climate change, geopolitical instability, and increasing inequality and fragmentation represent \u0026ldquo;wicked problems\u0026rdquo; characterised by contested problem definitions, interdependencies, and evolving values (Rittel \u0026amp; Webber, 1973). Data-driven optimisation and AI promise clearly identifying the \u0026lsquo;best\u0026rsquo; alternatives even\u0026ndash; or especially \u0026ndash; if problems are complex. However, by definition, wicked problems cannot be solved by traditional optimisation approaches (French, 2012).\u003c/p\u003e\n\u003cp\u003eWeick\u0026apos;s concept of sensemaking (Weick, 1993, 1995) provides the theoretical foundations to explain how humans navigate such ambiguity and uncertainty. Sensemaking is the process of meaning making, by which humans structure and process the stream of unfiltered, chaotic data, and turn it into meaningful and actionable information (Weick, 1993). Sensemaking is a creative process, wherein we humans construct \u0026lsquo;bridges\u0026rsquo; to address uncertainties consisting of ideas, thoughts, emotions, feelings, and memories (Sharoda \u0026amp; Reddy, 2010). Importantly, sensemaking is a social process, through which decision-makers interact with their peers and the environment, they coordinate, and receive feedback through enactment, allowing the formation of collective action agendas (Maitlis \u0026amp; Christianson, 2014). Sensemaking also entails the process of identity construction, by which humans come to understand what is \u0026apos;meaningful\u0026apos; in their own identities, teams and organisations (Helms Mills et al., 2010).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eUnlike rational decision-making models that assume clear problems and predetermined objectives (Gralla et al., 2016), sensemaking recognises that in complex environments, actors first construct what situations mean before determining their responses (Comes et al., 2020). Where optimisation assumes fixed objectives and measurable trade-offs, sensemaking analyses how those objectives emerge through interaction, feedback, and evolving understanding. Rather than converging on stable solutions, sensemaking sustains the ambiguity necessary for creative reinterpretation and collective learning (Weick \u0026amp; Sutcliffe, 2007).\u003c/p\u003e\n\u003cp\u003eAs AI systems increasingly filter information, prioritise, and suggest what to do, they inevitably become participants in these sensemaking dynamics and shape the cognitive and social foundations on which sensemaking depends. The question that emerges is not simply how to design AI that supports human decisions, but how to understand AI\u0026rsquo;s role in shaping the very processes through which our collective understanding and coordinated action emerge.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e2.2 Human-Centred AI: Towards Control in Human-AI Networks\u003c/h2\u003e\n\u003cp\u003eThe term human-centred AI is increasingly popular in response to the many concerns about AI risks and human agency in increasingly digitally mediated environments (Capel \u0026amp; Brereton, 2023). Instead of a world in which AI optimises our lives and takes control, human-centred AI presents a design paradigm to ensure that AI serves people and enhances human capabilities rather than replacing them (van Berkel et al., 2022). By definition, such an AI is trustworthy, reliable and safe for humans to use (Auernhammer, 2020; Shneiderman, 2020).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOften, control is framed as a question of who carries out which task or holds responsibility for it. The optimal distribution of tasks between humans and machines has been a central question since the 1950s, when Fitts (1951) introduced a protocol to decide which tasks are better performed by machines or by humans in the context of air traffic control. Today, this question has evolved to address how AI can support and optimise human decisions in complex problems while avoiding the pitfalls of cognitive or motivational biases. Tasks are assigned to humans or machines based on their respective expertise to optimise performance (Tausch \u0026amp; Kluge, 2022), with humans being better at tasks requiring creative and social intelligence (Ponti \u0026amp; Seredko, 2022). If tasks are allocated to machines, rapidly questions become acute about human autonomy and control (Abbass, 2019).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHowever, the question of control in human-AI systems extends beyond task allocation. The question is not merely about \u003cem\u003ewho\u003c/em\u003e (or what) decides. Rather, we need to ask how we can preserve meaningful human agency when AI systems shape the interpretive context in and through which we make choices. As AI filters, structures, and sequences information, it influences attention, salience, and relevance, thereby directing what we read, see, ignore, or find important. In doing so, AI systems (co-)construct the conditions in which human sensemaking unfolds.\u003c/p\u003e\n\u003cp\u003eMany of the frameworks to understand control stem from understanding human operators of machines, ranging from submarines (Sheridan et al., 1978) to air traffic (Council, 1998). These frameworks put forward assume a single human operator interacting with one machine (Endsley, 2017; Parasuraman et al., 2000). Yet, addressing complex challenges requires decentralised approaches, where many people (and potentially algorithms) work together (Mahajan et al., 2022). Increasingly, there is a discussion around human-AI teams (HAT), \u0026ldquo;\u003cem\u003ea purposeful combination of human and cyber-physical elements that collaboratively pursue goals that are unachievable by either individually\u003c/em\u003e\u0026rdquo; (Alix et al., 2021). Human-AI teams are composed of few people and machines (Endsley, 2023), whereas complex problems require the coordination of large-scale, ad-hoc and decentralised networks of humans and AI. Figure 1 summarises the evolution of research on human-machine interaction and control. With increasingly complex challenges and pervasiveness of AI, research needs to shift from AI that mimics human reasoning and single-operator systems to decentralised evolving human-AI networks that require coordination and human control.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhile many AI-supported decisions are intended as quick fixes and short-term optimisations, they often produce lasting consequences. Once enacted, decisions alter social networks, infrastructures, and expectations, creating path dependencies that narrow future choices (Webster, 2008). These effects are especially problematic when they shape collective sensemaking trajectories (Helms Mills et al., 2010), influencing how future situations are interpreted and what actions seem legitimate. Moreover, AI systems themselves evolve through feedback: large language models, for instance, adapt to user behaviour, reinforcing patterns of engagement and attachment (Kim et al., 2025) and producing recursive information bubbles (Jacob et al., 2025). Over time, both the humans and the AI intended to support the humans may drift away from the original intentions\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ein ways that become difficult to detect, control or reverse. Preserving human control thus requires more than oversight: it demands mechanisms for \u003cem\u003etemporal reflexivity\u003c/em\u003e: the ability to recognise deviations from purpose and intervene in unfolding trajectories when needed.\u003c/p\u003e\n\u003cp\u003eThese temporal shifts and the decentralised nature of human-AI networks challenge conventional notions of agency and control, particularly as AI becomes embedded in systems that evolve faster than human governance can follow. In the next section, I examine how these dynamics manifest via information-decision-feedback loops.\u003c/p\u003e\n\u003ch2\u003e2.3 Optimisation Cascades and Information-Decision-Feedback\u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eAI and humans exist in a complex control relationship: AI influences human sensemaking and decision-making, and humans engineer, train, steer, and \u0026ndash; possibly \u0026ndash; control the AI (Rahwan et al., 2019). Despite this interdependence, information and decisions are typically studied independently, and decision-information-feedback is only marginally considered.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBy their very nature, poly-crises require urgent interventions despite their complexity. This urgency matters, as time compression alters information processing, information sharing, and human decision-making behaviour. With prospect theory in the 1970s (Kahneman \u0026amp; Tversky, 1979), the need to include cognitive aspects in risky decisions became prominently recognised and inspired a wealth of research pointing to the cognitive and motivational biases that they bring (Klein et al., 2010; Paulus et al., 2022; Weick, 1993). An obvious question then is: can AI improve human decision-making by overcoming the cognitive or motivational biases?\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo unpack this question, an analysis of human decision-making processes is helpful, see Figure 2, left side. The human process is driven by the interplay of sensemaking, coordination and decision-making, by which humans adapt to the changing informational and social environment. While often, it is assumed that decision-making is the guiding principle to organise coordination, research has shown that it is the interaction of sensemaking and decision-making (Comes et al., 2020; Gralla et al., 2016). This implies that as the understanding of the situation changes, also the objectives, preferences, values and thereby the required or desired solutions change, fundamentally contradicting the linear problem solving paradigm, by which first a problem is formalised, and then solved (Volkema, 1983).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAI and optimisation algorithms fundamentally change how we perceive our environment, with whom and how we interact to find solutions (Kiesler \u0026amp; Sproull, 1992). For illustrative purposes, Figure 2 (right side) shows the most extreme case where one or several AI systems autonomously acquire and analyse data; optimise, decide and implement the decision in \u0026lsquo;\u003cem\u003eoptimisation cascades\u0026rsquo;\u003c/em\u003e. This is not meant as a prescription, but to contrast with the human sensemaking cycles. In this case, the creative, social and identity-forming process of sensemaking is replaced by data analysis. \u003cem\u003eOptimisation cascades manifest across scales. At the individual level, recommendation algorithms and LLMs optimise for engagement based on personal information, shaping what people read or see and thereby gradually shifting user preferences and choices\u0026nbsp;\u003c/em\u003e\u003cem\u003e(Sharma et al., 2024)\u003c/em\u003e\u003cem\u003e. At the policy level, performance metrics drive decisions under the umbrella of New Public Management, even if they only poorly capture intended outcomes. This phenomenon that\u0026nbsp;\u003c/em\u003e\u003cem\u003eMuller (2018)\u003c/em\u003e\u003cem\u003e\u0026nbsp;called the \u0026ldquo;tyranny of metrics\u0026rdquo;\u003c/em\u003e\u003cem\u003e\u0026nbsp;distorts\u003c/em\u003e our understanding of complex phenomena such as poverty, climate change, or welfare.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eA fundamental tension between human sensemaking cycles and optimisation cascades is that human processes are iterative social cycles where problem definitions and objectives co-evolve with shared understanding. In contrast, optimisation processes replace the underlying creative ambiguity with predetermined static metrics and automated execution of the optimal decision. The risk here is not only that machines make \u0026ldquo;wrong\u0026rdquo; decisions, but that the optimisation fundamentally impacts human sensemaking, shaping\u0026nbsp;\u003c/em\u003esocial orders, norms of our interactions, and what we perceive as \u0026lsquo;good\u0026rsquo; or even thinkable solutions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eToday, hybrid approaches, by which many humans and machines work together, are becoming the norm. An open question is how to coordinate the resulting dynamic networks of humans and AI, while controlling information-decision feedback loops. Here, coordination is defined as the set of procedures by which teams plan, organise, orchestrate and integrate their activities to achieve shared goals (Malone et al., 1994). As such, coordination entails activities such as information sharing, planning, task allocation or scheduling (Neale et al., 2004). Several studies have confirmed that decision performance on distributed tasks improves if individuals know who has access to what information (Stasser \u0026amp; Titus, 1985; Stewart \u0026amp; Stasser, 1995), which is problematic in complex networks and with black box AI models. Further, if AI \u0026ndash; and especially generative AI \u0026ndash; optimises information based on past interactions, it can introduce or reinforce existing biases (Sharma et al., 2024), which are then amplified via path-dependencies in sensemaking and decision trajectories. Given the broad impact of AI and optimisation on human sensemaking and decision-making in complex networks, the next sections examine how and in how far current Human-Centred AI literature addresses these fundamental challenges.\u003c/p\u003e"},{"header":"3.\tMethods","content":"\u003cp\u003eTo assess how the literature navigates the tension between data-driven optimisation and human meaning-making in uncertain and evolving contexts, I focus on the HCAI literature. HCAI addresses the relationship between human agency and AI systems. As such, it provides an important lens for evaluating how AI design accounts for the cognitive, social, and ethical dynamics of complex decisions. To explore this evolving body of work, I conducted a scoping review. Scoping reviews are designed to determine the scope or coverage of an emergent body of literature (Munn et al., 2018) and identify gaps in emerging interdisciplinary fields such (Sadek et al., 2024). \u0026nbsp; The scoping review was conducted on rayyan (https://www.rayyan.ai/) and followed the extended PRISMA protocol for scoping reviews (Tricco et al., 2018). Figure 3 summarises the screening and analysis process.\u003c/p\u003e\n\u003ch2\u003e3.1 Article Identification\u003c/h2\u003e\n\u003cp\u003eThe search was based on Scopus, because of its comprehensive coverage of interdisciplinary sources. To determine how human values and decisions are represented in the HCAI literature, the search focused on human-centred AI, searching combinations of \u0026ldquo;human-centred AI\u0026rdquo;, \u0026ldquo;human-centered AI\u0026rdquo; OR \u0026ldquo;HCAI\u0026rdquo; with \u0026ldquo;Decision support\u0026rdquo;, \u0026ldquo;Values\u0026rdquo; OR \u0026ldquo;Principles\u0026rdquo;. \u003cstrong\u003eThis framing focuses on decision problems to reveal the limitations of optimisation-driven AI\u003c/strong\u003e. The search was limited to articles in English, focusing on original manuscripts, case studies and perspective papers. \u003cstrong\u003eTo account for the rapid growth of the field, a preliminary pilot was conducted in June 2024, and the final search was executed on February 14, 2025.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eExcluding duplicates, the search led to 419 articles.\u003c/p\u003e\n\u003ch2\u003e3.2 Screening\u003c/h2\u003e\n\u003cp\u003eThe titles, abstracts and keywords of all articles were screened. Articles had to focus on supporting specific decision-making or sensemaking tasks and discuss the distribution of the tasks among humans and machines under the framework of human-centred AI to ensure that the different HCAI concepts or principles could be mapped out. This initial screening process resulted in 185 papers that were retained for the full-text review.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e3.3 Full-text Eligibility\u003c/h2\u003e\n\u003cp\u003eThe articles were then analysed based on the extent of discussion on human-AI interaction, decision-making/ sensemaking and HCAI concepts. Studies were excluded if they: (i) focused on physical interaction (e.g., robotics); (ii) addressed only interface design without engaging HCAI principles; (iii) referred only to conventional statistics, not AI; or (iv) were categorised as reviews or tutorials, or (v) were inaccessible, not in English, see Figure 3.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e3.4 Analysis and Extraction Framework\u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eThe analysis began by mapping publication trends and methods across the reviewed studies. For the \u003cem\u003emethods\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/em\u003eused, I categorise (i) perspective and conceptual papers that theorise or aim to guide the use of AI; (ii) empirical studies that observes how humans interact with machines in real life; (iii) behavioural experiments; (iv) design studies that build and implement the AI, also including technical studies, and (v) surveys. Further, I distinguish the \u003cem\u003efield\u003c/em\u003e, for which the AI is designed, from the\u0026nbsp;\u003cem\u003etype of AI or algorithm\u003c/em\u003e, ranging from conversational agents and chatbots, large language models and generative AI to machine learning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFrom there, I investigate the\u0026nbsp;\u003cem\u003erole\u003c/em\u003e of the human vis a vis the machines to understand how and in how far humans are \u0026ndash; indeed \u0026ndash; central to HCAI, and what the implications are for optimisation, sensemaking and decision-making. Questions are who has decision authority (the human or the machine); and how many humans are interacting with how many machines. I conceptualise this in terms of the following categories: (i) AI supports a single (human) user (who is responsible for the decision); (ii) the AI makes the decision autonomously; (iii) AI and humans act as a team, where the AI interacts with and advises multiple humans, or (iv) one AI interacts with many humans, going beyond the concept of a team.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFinally, the review assessed which \u003cstrong\u003e\u003cem\u003eHCAI design principles and concerns\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eare most prominent in the literature, and how they co-occur. Because many terms in the literature\u0026mdash;such as fairness, explainability, and automation\u0026mdash;blur the boundary between normative values and technical features, I treat them collectively under the umbrella of \u003cem\u003eHCAI design principles\u003c/em\u003e. This includes both ethical principles (e.g., solidarity, equity) and system-level properties (e.g., transparency, control, privacy). The initial coding categories were established on the basis of major guidelines (EC, 2019; OECD, 2019; UNESCO, 2022), as well as the seminal HCAI literature (Shneiderman, 2020). Through bottom-up coding concepts were added. An analysis of frequency and co-occurrence patterns was then used to explore dominant principles and gaps in the HCAI literature. Table 1 provides an overview of the dimensions of analysis and categories used.\u003c/p\u003e\n\u003cp\u003eTable 1: Overview of the Analysis \u0026amp; Extraction Framework for the Scoping Review\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDimension\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCategories\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eResearch Method\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePerspective/conceptual; qualitative empirical; behavioural experiments; design and tool developments; surveys\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eField of Application\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMedicine \u0026amp; Health; Public policy \u0026amp; governance, Business; Education; Crisis \u0026amp; Safety; Manufacturing; Software Engineering; Transport; Other\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eType of AI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAI agnostic/unspecified; Machine Learning, Natural Language Programming (NLP); Generative AI; Agentic AI\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRelation of Human And AI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eOne AI supports one single human; AI decides autonomously; one AI interacts with many humans; Human-AI teams\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDecision Authority\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHuman only, Machine only, Shared (interactive), Distributed\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHCAI DESIGN Principles Considered\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eExplainability, Fairness, Trust, Transparency, Accountability, Solidarity, Contextualisation, Empowerment, Safety, Humanity, Control, Agency, Privacy, Useability, Responsibility, Automation, Equity, Situational Awareness, Bias\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e3.5 Design Principles for Sensemaking AI\u003c/h2\u003e\n\u003cp\u003eIn addition to the scoping review, this paper develops a set of theoretical arguments and proposes three interconnected design principles for Sensemaking AI with the aim of starting to address conceptual blind spots of the current literature. The design principles proposed are developed through an abductive process (S\u0026aelig;tre \u0026amp; Van de Ven, 2021) that iteratively links theoretical framing from Sensemaking and Decision-Making theories and empirical patterns from the review with illustrative examples from poly-crises to develop a research and design agenda for AI that supports meaning-making in complex systems.\u003c/p\u003e"},{"header":"4.\tResults ","content":"\u003cp\u003eThis section presents the results of the scoping review and examines how HCAI research engages with the tensions between optimisation, human agency, and sensemaking in complex environments. By analysing patterns in methods, applications, AI system types, roles of humans and AI, and HCAI design principles, this results section aims to provide insights into how the field has evolved and where important gaps may remain.\u003cbr\u003e\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e4.1 The How, What and Where of HCAI\u003c/h2\u003e\n\u003cp\u003eAlong with the rise of AI, there is also growing interest at the intersection of human-centred AI and decisions sparked by the first HCAI publication by Shneiderman (2020). The distribution of research methods (Figure 4) shows a highly diverse set of research methods, in which experimental, \u003cstrong\u003edesign\u003c/strong\u003e and \u003cstrong\u003eperspective\u003c/strong\u003e approaches dominate, each with almost 30 papers. Under \u003cstrong\u003edesign\u0026nbsp;\u003c/strong\u003emethods, papers are classified that design, build, and test Human-Centred AI applications, e.g., (Elahi et al., 2021; Erlei, 2024; Sun, 2022), indicating a focus on development and evaluation. \u003cstrong\u003eIn contrast,\u0026nbsp;\u003c/strong\u003ethe many perspective and opinion papers emphasise theoretical and conceptual discussions, primarily regarding the value perspective in HCAI\u0026ndash; especially in sensitive contexts such as education, health or crisis management (Comes, 2024; Kattnig et al., 2024; van Leersum \u0026amp; Maathuis, 2025). Controlled \u003cstrong\u003eexperiments\u003c/strong\u003eoften study cognition and behaviour by focusing on how users interact with an AI. These studies link HCAI to Human-Computer-Interaction and often test for the impact of specific principles such as fairness or explainability (Flathmann et al., 2023; Gajos \u0026amp; Mamykina, 2022). Qualitative methods that study the use of AI in situ and surveys that focus on perception and use of technology especially in work processes are less common (Bingley et al., 2023; Herrmann \u0026amp; Pfeiffer, 2023).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFigure 5 shows the distribution of the types of AI that are considered. Strikingly, almost 40% of the publications theorise or analyse \u0026lsquo;\u003cem\u003eAI\u0026rsquo;\u003c/em\u003e as a generic term, being agnostic of the specific algorithm (Akula \u0026amp; Garibay, 2021; Bingley et al., 2023; Hoch et al., 2022). \u003cstrong\u003eMachine learning (ML), including deep learning,\u003c/strong\u003e is the most common approach specified with more than 35 % of publications, either alone or in combination with other methods such as \u003cstrong\u003enatural language processing (NLP)\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;\u003cstrong\u003egenerative AI\u003c/strong\u003e. \u003cstrong\u003eIncreasingly, generative AI\u003c/strong\u003e\u0026nbsp;\u003c/strong\u003epapers are a part of the HCAI literature (Bu\u0026ccedil;inca, 2024; Erlei, 2024; Kattnig et al., 2024). \u003cstrong\u003eAgentic AI\u003c/strong\u003e\u003cstrong\u003e,\u0026nbsp;\u003c/strong\u003ewhich refers to AI systems with autonomous decision-making capabilities (Acharya et al., 2025), is also increasingly well-represented with almost 10% of publications, indicating a growing interest in AI autonomy and adaptive behaviour \u0026ndash; most often via humans interacting with an autonomous agent (Criscuolo \u0026amp; Dolci, 2024; Gou et al., 2024).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe review clearly shows that HCAI has made its entrance in a diversity of fields, cf. Figure 6. Applications in health and medicine are most prominent (19/101), most often in the context of clinical decision support systems for diagnosis or treatment e.g., (Lee et al., 2022; Van Berkel et al., 2023; Verma et al., 2023). This is followed by public policy and governance (Lee et al., 2019; Lettieri et al., 2023; Stapleton et al., 2022) including several papers that discuss sustainability aspects (Sigfrids et al., 2023). Business and managerial applications (Freire et al., 2023; Hoch et al., 2022) follow. In education (Chaudhry et al., 2022; Duan et al., 2024) and crisis management, papers also discuss the risks of AI especially regarding the introduction of new biases (Chaudhry et al., 2022). \u0026lsquo;Other\u0026rsquo; is a highly diverse category, including law, agriculture, and language.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e4.2 The Role of AI in HCAI\u003c/h2\u003e\n\u003cp\u003eWhen it comes to the roles of the AI, Figure 7 shows that with 65/101 papers, the large majority of papers focus on an AI supporting an individual decision-maker, ranging from supporting elite sports coaches (Comes, 2024; Maiden et al., 2023) to supporting elderly app users in smart cities (Elahi et al., 2021), from nurses steering patients to a hospital (Li et al., 2024) to managers making strategic decisions (Passlack et al., 2024). Fewer papers (13/101) discuss Human-Centred AI for autonomous decision-making by which the AI fully automates decisions (Bingley et al., 2023; He et al., 2023; Jin et al., 2022; Lee et al., 2019; Nabizadeh Rafsanjani \u0026amp; Nabizadeh, 2023; Shulner-Tal et al., 2023; Suchan et al., 2021; Yazdanpanah et al., 2021). Despite the increasing prevalence of AI in society, only nine papers discuss one AI that supports many humans, primarily in the context of crowdsourcing (Sprenkamp et al., 2025). Further thirteen papers study human-AI-teams, and the implications for group dynamics (Bansal et al., 2021; Cooke et al., 2020; Flathmann et al., 2023; Riedl, 2019), for instance in medical teams (Hagemann et al., 2023; Verma et al., 2023) or manufacturing (Hoch et al., 2022).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhat is, however, missing are studies that analyse the broader societal implications, collective intelligence, and democratic processes whereby many humans work with many AI systems that impact information flows and decision-making. When many actors optimise their choices based on optimised input by AI algorithms, their collective behaviour can generate system-wide effects that no single actor anticipated or can control. If AI is implemented at scale and optimised decisions propagate through and even shape the networks of our interaction, what are the emergent effects on human sensemaking and decision-making over time? And how can these emergent effects in complex human\u0026ndash;AI networks be understood and controlled? The principles that guide the interaction of one human with one AI or one AI with a small group of humans may not suffice for analysing the dynamics of complex networks in which many humans are supported by many machines. These dynamics raise questions about automation, coordination, and the preservation of meaningful human control.\u003c/p\u003e\n\u003ch2\u003e4.3 What HCAI Optimises for\u003c/h2\u003e\n\u003cp\u003eDecisions are inherently linked to what we value. AI that supports human decision-making must therefore also grapple with fundamental questions about which values and principles should guide their design and operation. As outlined in the methods section, the HCAI literature blurs the boundary between normative values and technical requirements. Figure 8 shows the concepts currently used as design principles and guiding values in the HCAI literature. Clearly, current HCAI research emphasises explainability (n=45) (Sun, 2022), trust (n=29) (Liao \u0026amp; Sundar, 2022), transparency (n=24) (Kunar et al., 2024), and fairness (n=23) (Kattnig et al., 2024), while other concepts ranging from equity (n=3) (Akula \u0026amp; Garibay, 2021) to solidarity (n=3) (Sigfrids et al., 2023), humanity (n=1, \u0026lsquo;Other\u0026rsquo;) (Comes, 2024), or safety (n=2, \u0026lsquo;Other\u0026rsquo;) (Zhang et al., 2024) receive far less attention, even though they are key to human decision-making, especially in poly-crises.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGiven the analytical nature of AI research, it may not be surprising that also HCAI research prioritises concepts that can be easily measured or optimised for, such as explainability, fairness or trust. This creates a fundamental problem: explainability metrics become substitutes for accountability; fairness for justice; and trust replaces meaningful human oversight. The result is systems optimised for measurable proxies rather than the deeper values or principles they are supposed to represent. By focusing on what can be optimised for, the field of HCAI may therefore inadvertently reproduce the very reductionism it seeks to address\u003c/p\u003e\n\u003cp\u003eMoving from individual concepts to patterns of co-occurrence, Figure 9 shows that explainability in combination with trust (n=17), fairness (n=8) and situational awareness (n=8) dominates the discussion. At the same time, Figure 9 highlights limited attention to the link between other concepts. Despite the calls for contextualising AI (Sloane et al., 2023) to the environments and social dynamics of its use, the link between contextualisation and principles such as fairness or accountability is under-explored. This is an important gap since many decisions are highly situational as discussed in Section 2. Surprisingly, agency, autonomy and control are less frequently associated with concerns for fairness, accountability and privacy, suggesting a research gap in how we balance the need for efficient, rapid and automated decisions with human oversight and value deliberation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn sum, this scoping review shows that current HCAI focuses on dyadic interactions, measurable principles, and task distribution between humans and machines. What is missing is a deeper discussion and understanding of how optimisation for specific types of interactions and role, tasks, or principles reshapes the very contexts in which we humans make meaning.\u0026nbsp;\u003c/p\u003e"},{"header":"5.\tDiscussion: From Human-Centred to Sensemaking AI ","content":"\u003cp\u003eThe results of the scoping review show a fundamental tension: while HCAI research has made important advances in improving individual-level explainability, fairness, or trust, it has yet to grapple with what happens when these optimised interactions scale to complex, networked environments where many humans work with many AI systems. In such environments, optimisation reshapes the informational and social conditions under which actors interpret, coordinate and decide \u0026ndash; the terrain of sensemaking itself. This creates an \u003cem\u003eoptimisation paradox\u003c/em\u003e: the more we optimise individual human-AI interactions, the less equipped we may become to handle the emergent, collective challenges that define complex socio-technical systems. This paradox motivates a reorientation from model‑centric improvements toward AI that sustains collective meaning‑making in evolving human\u0026ndash;AI networks.\u003c/p\u003e\n\u003cp\u003eThis paradox becomes particularly visible in poly-crises, which are complex, decentralised, fraught with dilemmas and marked by \u0026lsquo;\u003cem\u003etime running out\u0026rsquo;\u0026nbsp;\u003c/em\u003e(Comes, 2024; Levin et al., 2012). This combination fundamentally changes human sensemaking and decision-making behaviour, thereby also altering the evolving human-AI-interactions.\u0026nbsp;HCAI research predominantly frames interactions as generic, not contextualised, as dyadic, not networked and as static, not dynamically evolving. The networks involved in crises, however, are inherently dynamic and uncertain, demanding flexible and context-sensitive AI systems\u0026nbsp;(Jennings et al., 2014). Building on these observations, I propose three shifts: Sensemaking‑aware automation, collective agency for network‑level control, and value‑aware sensemaking. Together, these shifts provide a research and design agenda for Sensemaking AI.\u003c/p\u003e\n\u003ch2\u003e5.1 \u0026nbsp;Sensemaking-Aware Automation\u003c/h2\u003e\n\u003cp\u003eIt is tempting to assume that machines can optimise our meaning-making\u0026mdash;make it faster, more efficient, less biased. But automation shapes the trajectories of interpretation by narrowing how problems are framed. Although the review shows the increasing recognition of human-AI-teaming in situational awareness (Endsley, 2023; Gajos \u0026amp; Mamykina, 2022) or developing shared mental models (Hoch et al., 2022; Thompson, 2021), most research focuses on computer-supported information sharing or decision-making. Sensemaking is treated as an individual cognitive process that can be supported through better information presentation, rather than recognising it as a fundamentally \u003cstrong\u003e\u003cem\u003ecollective\u003c/em\u003e\u003c/strong\u003e process of meaning construction that emerges through social interaction and network dynamics. In addition, the AI is seen as a (neutral) tool that mediates human-human interaction (Heyndels, 2023), rather than viewing AI as an \u003cem\u003eactor\u003c/em\u003e that optimises information flows and thereby shapes sensemaking and decision-making. As humans rely routinely on AI, however, the relation between a human \u0026lsquo;operator\u0026rsquo; and an AI tool supporting the human has become blurry.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAutomation does not just take over individual tasks or decisions. Rather, automation reshapes how we understand what matters by impacting the process of meaning-making, through which we individually and collectively define and understand our choices. Via sensemaking trajectories, initial perceptions \u0026ndash; as moderated or provided by an AI - can become deeply engrained (Comes et al., 2020), leading to path-dependencies that remain hidden within the current task-based frameworks. As AI systems continuously filter, prioritise, and present information, they do not merely \u0026lsquo;support\u0026rsquo; decision-making but shape human cognition, reinforcing certain narratives while marginalising others. In essence, sensemaking depends on \u003cem\u003eambiguity\u003c/em\u003e, the interpretive flexibility that enables creative reframing when understanding evolves (Weick, 2015). The question becomes not whether to automate information processing or decision-making, but how to ensure such automation supports rather than constrains the interpretive flexibility that enables creative reframing. \u003cem\u003eA way ahead in achieving interpretive flexibility may be\u003c/em\u003e\u003cstrong\u003egraceful degradation\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e(Ploeg et al., 2014)\u003c/strong\u003e\u003cem\u003e\u003cstrong\u003e\u0026nbsp;\u0026ndash;\u0026nbsp;\u003c/strong\u003e\u003c/em\u003e\u003cem\u003edesigned fallback from higher levels of automation that force systems to slow, expose provenance, invite dissent, and hand decision authority to people if needed.\u0026nbsp;\u003c/em\u003eInterpretive flexibility, however, is not an individual cognitive capacity that can be preserved through dyadic human-AI interactions. Sensemaking is fundamentally social\u0026mdash;it emerges through collective interaction, shared interpretation, and distributed meaning-making across networks of actors\u0026nbsp;(Weick, 1995). Yet current HCAI research largely overlooks this social dimension.\u003c/p\u003e\n\u003cp\u003eEspecially the role of AI in identity construction, by which identity is continuously shaped through social interaction and feedback (Weick, 1995), has not received attention yet. AI, as an actor, influences how humans construct their professional and social identities by mediating access to information, influencing real or perceived agency, and reinforcing or challenging organisational and societal norms. Especially with the increasing personalisation of generative AI, there is a risk that identities become shaped by an algorithm that amplifies biases and leads to echo chambers. When AI systems optimise for engagement, productivity, or efficiency, they reshape the very questions humans ask about meaning and purpose. In addition, there is a growing concern about the risk of de-skilling, where the reliance on AI leads to an erosion of critical thinking (Sellen \u0026amp; Horvitz, 2024). What is more, complex problems are often characterised by moral dilemmas. Yet off-loading morally challenging decisions to a machine may lead to \u003cem\u003emoral de-skilling\u003c/em\u003e (Vallor, 2015). As such, the impact of AI on perceived or real responsibility for others, and on the evolution of human competences and skills are a concern that must be addressed in sensemaking aware automation.\u003c/p\u003e\n\u003cp\u003eTo address this gap, Sensemaking AI can draw on different bodies of literature as summarised in Figure 10. First, there is wealth of research on group decision-making (Hollingshead et al., 1993), information sharing and sensemaking (Stasser \u0026amp; Titus, 1985) (left box in Figure 10; see also Figure 2) dedicated to how teams and groups differ from individuals, which is largely neglected in the HCAI literature. This gap becomes evident when we consider our finding that only 22 (9+13) out of 101 papers address situations where AI systems interact with multiple humans simultaneously. At the other end of the spectrum, research on collective machine behaviour and multi-AI coordination (right side in Figure 10) focuses on coordinating different artificial agents (Stone et al., 2010). While 13 papers in the review discuss autonomous systems, none of them focuses how the coordination of these systems can or should be designed to benefit humans. \u0026nbsp;Here, the role of trust, loyalty, or cognitive and behavioural factors that are important in human interactions and group decision-making are discarded; networks and groups are formed based on optimal skillsets or available resources.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eI argue that Sensemaking AI should address the reality of complex networks where multiple types of interactions occur simultaneously: humans engaging in collective sensemaking with each other, humans interacting with multiple AI systems that shape their information environments and thereby their decisions, and AI systems that coordinate or influence each other. This implies that theories of Sensemaking AI recognise that all these interactions constitute a single, complex system where human sensemaking, AI mediation, and algorithmic coordination are fundamentally intertwined and mutually constitutive. This requires integrating theories on group decisions and sensemaking into HCAI research and incorporating task distribution and information prioritisation protocols from collective machine behaviour research.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBuilding on the integrated theoretical framework and recognising that AI fundamentally influences social networks, shared meaning-making and values, we need to then focus on Sensemaking AI as a design principle and ask: which sensemaking, coordination processes and decisions do we want to or need to optimise or automate, and why? Answering this question, especially given the urgency and moral dilemmas pertaining to poly-crises, requires addressing research questions such as: How does AI impact shared identity construction and how can collective moral de-skilling be avoided? How do optimisation algorithms impact collective sensemaking trajectories and thereby shape the interpretive flexibility that is crucial for sensemaking?\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnswering these questions requires expanding the current theoretical foundations and conduct interdisciplinary research that focuses on large-scale and longitudinal studies on network dynamics as the central focus. Here, longitudinal empirical studies with foundations in group decision making (e.g., information pooling, shared mental models) can be combined with insights from machine behaviour (e.g., prioritisation protocols) and computational models from complexity science to capture the interplay between automation and evolving sensemaking. Experimental designs that measure interpretive flexibility, identity construction and (moral) deskilling before and after the use of an AI in different constellations of groups and teams can create new insights into sensemaking trajectories; based on these empirical insight, agent-based models (Nespeca et al., 2021) can simulate information sharing and sensemaking dynamics to explore path-dependencies and conditions or tipping points that lead to the erosion of interpretive flexibility in networks.\u003c/p\u003e\n\u003ch2\u003e5.2 Collective Agency for Network‑Level Control\u003c/h2\u003e\n\u003cp\u003eThe discussion about automation is inherently connected to questions of control, autonomy and agency. However, this review shows limited attention for control mechanisms beyond individual human-AI interactions, with concepts like \u0026apos;control\u0026apos; (n=12) and \u0026apos;agency\u0026apos; (n=10) receiving far less attention than principles like explainability (n=45), see Figure 8. Even \u0026lsquo;autonomy\u0026rsquo; (n=21) is most often discussed in the context of trust (6/21) and fairness (6/21) rather than through the lens of control (4/21), see Figure 9.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCurrent principled approaches to control and accountability overlook the complexity arising from the many diverse interactions of humans and machines. For instance, the OECD guidelines specify that \u0026ldquo;\u003cem\u003eAI actors should be accountable for the proper functioning of AI systems\u0026rdquo;\u003c/em\u003e (OECD, 2019). But what if it is precisely the \u0026lsquo;proper functioning\u0026rsquo; that leads to undesired consequences or harmful cascading effects? When optimisation decisions propagate through networks of human and AI actors, the \u0026ldquo;problem of many hands\u0026rdquo; and subsequent responsibility gaps occur (Matthias, 2004). This problem cannot simply be solved by distributing responsibility (Coeckelbergh, 2020) since it is not clear who bears responsibility when properly functioning optimisation algorithms produce undesirable outcomes at systems level. \u003cstrong\u003ePreserving human control in such dynamic systems thus requires\u0026nbsp;\u003c/strong\u003e\u003cem\u003etemporal reflexivity (see Section 2)\u003c/em\u003e\u003cstrong\u003e\u003cem\u003e,\u003c/em\u003e i.e., the ability to recognise when optimisation\u0026nbsp;\u003c/strong\u003edrifts away from the original purpose and to intervene in time to change the unfolding trajectories when needed\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMoreover, research has shown that decisions shape physical, informational, and social networks, which in turn influence the information accessible (Comes et al., 2020) to human actors and AI agents. When algorithms optimise traffic and energy flows, financial markets, or entire smart cities, they create optimised environments, in which human choices are increasingly constrained by algorithmic assumptions about what should be optimised, and how. In these contexts, traditional concepts of control\u0026mdash;rooted in task allocation and oversight\u0026mdash;become inadequate since the challenge is not only controlling what machines \u003cem\u003edo\u003c/em\u003e, but preserving spaces for human interaction and meaning-making within emergent, decentralised networks. As such, the question of human control becomes: how can human agency be preserved when optimisation algorithms and AI increasingly dominate information flows and decision architectures? How do decision-information feedback loops influence the long-term evolution of control structures in human-AI networks across spatial and temporal scales?\u003c/p\u003e\n\u003cp\u003eAddressing control in emergent human-AI networks requires moving beyond traditional oversight models toward networked agency. Control theory and cybernetics, originally developed by Wiener (1948), \u0026nbsp; provides the theoretical framework for understanding adaptive regulation: control theory provides a framework for modelling decision loops where human and AI agents dynamically adjust their actions based on new information, constraints, and goals. Cybernetics stresses the need for self-correcting incentives and governance. As such, cybernetics has also been suggested as a way to coordinate decentralised AI networks in autonomic computing (De Wolf \u0026amp; Holvoet, 2003) and more recently as a governance principle for humans and technology (Zwitter, 2024). Rather than centralised monitoring, cybernetic approaches enable \u003cstrong\u003enetwork-level self-regulation\u003c/strong\u003e where control emerges through distributed feedback loops and adaptive responses to changing conditions, e.g., via meta‑signals on uncertainty and impact, circuit breakers for cascading automations, auditability of decision\u0026ndash;information feedback loops. This shift\u003cem\u003e\u0026nbsp;\u003cem\u003ereframes control from actions or outcomes to designing\u0026nbsp;\u003c/em\u003e\u003c/em\u003e\u003cstrong\u003eself‑correcting conditions\u003c/strong\u003e\u003cem\u003e, under which collective agency can emerge and evolve with human-AI networks.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003e5.3 Value-Aware Sensemaking AI: Processes and Boundaries\u003c/h2\u003e\n\u003cp\u003eValue-aware Sensemaking AI refers to AI that makes value assumptions explicit and revisable. As such, value-aware Sensemaking AI needs to distinguish between process principles (how values are formed, contested and revised) and content principles. Such systems make value assumptions explicit and revisable, facilitating how values are surfaced, balanced and formalised while preserving space for disagreement if trade-offs violate moral boundaries.\u003c/p\u003e\n\u003cp\u003eThe HCAI literature recognises that AI systems need to be designed to \u0026ldquo;\u003cem\u003eunderstand humans\u003c/em\u003e\u0026rdquo; including the norms and values that govern our actions (Riedl, 2019). Even though Shneiderman (2020) proposed HCAI as a design \u003cem\u003eprocess\u003c/em\u003e, this review shows \u0026lsquo;\u003cem\u003ehuman-centred\u0026rsquo;\u003c/em\u003e has largely become a synonym to explainable, fair, accountable, transparent and trusted AI systems, see Figure 8. These principles are often treated as generic optimisable requirements that systems can be built from irrespective of the context. Maybe not surprisingly, a similar view is presented by the various guidelines, standards and regulatory frameworks for the design and use of AI. The UNESCO recommendations on the Ethics of AI (UNESCO, 2022), the OECD Recommendation of the Council on Artificial Intelligence (OECD, 2019), the European Commission\u0026rsquo;s recommendations by the High-Level Expert Group on AI (EC, 2019) the IEEE standards for Ethically Aligned Design of Autonomous and Intelligent Systems (IEEE, 2019), and the EU AI Act establish important foundations around transparency, accountability, trust and fairness. However, they all operate under the assumption that values can be translated into generic, stable, measurable principles. This suggests a tendency to optimise for what can be measured while neglecting values that may be essential \u003cem\u003ebecause\u003c/em\u003e they resist quantification.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn addition, current HCAI principles assume that all principles can be achieved simultaneously. Yet, there are inherent conflicts across principles or what society values, and these conflicts cannot always be reconciled, e.g., when climate justice collides with economic stability, transparency with privacy, or control with personal freedom. There is an expanding literature that highlights that humans refuse making such trade-offs because they are seen as morally problematic, or taboo (Chorus et al., 2018; Tetlock, 2003). Formalising such trade-offs in an optimisation then risks treating them as commensurable, thereby eroding their role as ethical boundaries.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eunderlying challenge is:\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eAI principles are viewed in separation from the contextualised objectives, preferences, attitudes, or emotions that drive human sensemaking and decision-making (van Berkel et al., 2022), as well as from the consequences that occur if AI is used at scale. Based on sensemaking theory, I argue that the meaning of key principles depends on continuous collective (re-)interpretation. I do not advocate for abandoning principled approaches, but rather for distinguishing between process principles that support collective meaning-making and democratic deliberation from content principles that may predetermine its outcomes. This shift requires integrating approaches that recognise value formation as an \u003cem\u003eoutput\u003c/em\u003e of human-AI interaction. Social choice ethics for AI design (Baum, 2020) provides a framework for this, emphasising questions of standing (\u003cem\u003ewho\u003c/em\u003e participates in value construction?), measurement (\u003cem\u003ehow\u003c/em\u003e are diverse perspectives translated into system design?), and aggregation (\u003cem\u003ehow\u003c/em\u003e do we coordinate across potentially conflicting value systems?).\u003c/p\u003e\n\u003cp\u003eThis process-centered approach creates an important challenge: how to translate the outcomes of collective deliberation into formal specifications without undermining the integrity of the process? To ensure that the results of collective value formation can guide AI development, outcomes need to be linked to formal decision theoretical frameworks that translate values into objective functions and operational trade-offs. Research is needed to formalise abstract goals such as equity (Coleman et al., 2024; Holguin-Veras et al., 2013) and to explore the dynamic nature and structure of trade-offs for instance for intangible or sacred goods via taboo trade-offs (Daw et al., 2015; Lu et al., 2021). At the same time, democratic deliberation may identify domains, values or decisions that cannot be translated, and where preserving ambiguity, maintaining human judgment, and sustaining ongoing deliberation is more important than efficiency. This leads to the question: how can AI systems recognise and respect the boundaries of their own applicability as determined via deliberation? By addressing these challenges, research on Sensemaking AI can combine the cognitive, behavioural, social and ethical elements needed to move towards AI that supports rather than constrains collective meaning-making.\u003c/p\u003e\n\u003cp\u003eTaken together, the three shifts towards Sensemaking AI repositions AI from a tool that optimises towards generic objectives in dyadic relations to an actor that shapes \u003cstrong\u003ecollective cognition in networks. AI and humans, together, are\u0026nbsp;\u003c/strong\u003etasked with sustaining interpretive flexibility, networked control and value-awareness at scale.\u0026nbsp;\u003c/p\u003e"},{"header":"6.\tConclusion ","content":"\u003cp\u003eHuman-Centred AI (HCAI) has been put forward as a paradigm to design AI the supports humans by advocating for design principles ensuring that AI is explainable, fair, transparent and trustworthy. Yet, this scoping review shows that these principles are largely operationalised for dyadic interactions where one human works with one AI, and that the focus is on a relatively narrow set of values that can be readily operationalised. Even though I acknowledge that this review provides only a snapshot, and that the field of (HC)AI is rapidly evolving, this framing is too narrow for the complex and often time-compressed decisions that we are facing today. The ubiquity of AI, and the optimisation of information sharing and processing, reconfigures \u003cstrong\u003ethe informational and social conditions\u003c/strong\u003e under which humans interpret situations and decide. AI reshapes the way we make sense of our environment.\u003c/p\u003e\n\u003cp\u003eAgainst this backdrop, this paper advocates for a paradigm shift towards Sensemaking AI: AI that supports collective meaning-making in evolving human-AI networks. Conceptually, the paper synthesises sensemaking and decision theory with literatures on coordination and machine behaviour to characterise AI as an \u003cstrong\u003eactor\u003c/strong\u003e within socio‑technical systems. Via a scoping review of the HCAI literature, this paper highlights that current research focuses on individual support and generic, measurable principles. Gaps persist in our understanding of how (human-centred) AI impacts and reshapes sensemaking and decision-making over time. Together, these strands motivate three interconnected directions for Sensemaking AI research and design:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eSensemaking-aware automation:\u003c/strong\u003e AI shapes sensemaking trajectories, reinforcing certain narratives while marginalising others. Research needs to expand beyond dyadic human-AI interactions and integrate group decision and collective machine behaviour theories to understand how automation impacts collective sensemaking in dynamic networks where many humans work with many algorithms. Future research questions include: How does AI-driven automation influence identity construction, (moral) de-skilling and collective meaning-making over time? What mechanisms can mitigate path dependencies, and how to preserve the interpretive flexibility essential for creative reinterpretation in networked systems.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCollective agency for networked control:\u003c/strong\u003e in complex systems, oversight of individual components can never ensure control of the whole system. Therefore, AI understood as an actor in complex networks poses a challenge for conceptualising and maintaining human control and oversight, especially since over time optimisation cascades create decision-information feedback loops. Sensemaking AI reframes control as networked agency. Research needs to analyse how to preserve collective agency when optimisation logic increasingly dominates decision architectures. Drawing on control theory and cybernetics, this requires designing conditions that enable network-level self-regulation through distributed feedback loops and adaptive responses to changing conditions.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eValue-aware Sensemaking:\u003c/strong\u003e Current principled AI frameworks lack the adaptability required for complex, dynamic decisions. Recognising that some values\u0026mdash;such as dignity, justice, and humanity\u0026mdash;resist translation into optimisable metrics and cannot be traded off, Sensemaking AI distinguishes \u003cstrong\u003eprocess principles\u003c/strong\u003e\u003cstrong\u003e,\u0026nbsp;\u003c/strong\u003ewhich support\u003cstrong\u003e\u0026nbsp;\u003cstrong\u003eongoing democratic deliberation and contextualisation\u003c/strong\u003e\u003c/strong\u003e, from \u003cstrong\u003econtent principles\u003c/strong\u003e that risk pre‑empting it. Methodologically, this calls for pipelines that (i) enable \u003cstrong\u003eparticipatory formation and revision\u003c/strong\u003e of objectives, (ii) translate deliberative outcomes into \u003cstrong\u003eformal decision models\u003c/strong\u003e \u003cem\u003ewhere appropriate\u003c/em\u003e, and (iii) specify \u003cstrong\u003eboundaries of applicability\u003c/strong\u003e where automated optimisation should \u003cstrong\u003edefer to human judgment\u003c/strong\u003e and sustain ambiguity. The aim is not to abandon principles, but to embed them in processes that keep values \u003cstrong\u003econtestable and revisable\u003c/strong\u003e at scale.\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe shift towards Sensemaking AI has actionable implications for research and practice: the focus of AI studies has to shift from individual decision-makers to studies that acknowledge AI as embedded in evolving social-technical networks. This requires a shift of methods towards integrating longitudinal, large-scale empirical studies with methods from complexity science to trace how information sharing, analysis, explanation and automation affect interpretive flexibility, sensemaking and coordination over time. AI design should integrate and test mechanisms to ensure a diversity of inputs, allow for surfacing dissent rather than focusing on convergence, and allow to shift from automation to human deliberation when needed. AI governance needs to move towards architectures that integrate feedback mechanisms and incentives, rather than promoting static principles.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe contribution of this paper lies in challenging the imperative within current Human-Centred AI literature to optimise AI systems to become more transparent or fair. Instead, Sensemaking AI is proposed as a concrete alternative that recognises AI as an actor within complex socio-technical systems that shapes collective meaning making and decision architectures. As such, Sensemaking AI needs to be designed to sustain the \u003cstrong\u003einterpretive, social, and ethical capacities\u003c/strong\u003e on which sensemaking depends. \u0026nbsp;\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAI: Artificial Intelligence\u003c/p\u003e\n\u003cp\u003eEC: European Commission\u003c/p\u003e\n\u003cp\u003eHCAI: Human-Centred Artificial Intelligence\u003c/p\u003e\n\u003cp\u003eLLM Large Language Model\u003cbr\u003e\u0026nbsp;NLP: Natural Language Processing\u003c/p\u003e\n\u003cp\u003eOECD: Organisation for Economic Co-operation and Development\u003c/p\u003e\n\u003cp\u003ePRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eStatements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eClinical trial number: not applicable as this is not a clinical study\u003c/p\u003e\n\u003cp\u003eEthics approval and consent to participate: Not applicable since there was no participation of human subjectis in the study.\u003cbr\u003e\u0026nbsp;Consent for publication: Not applicable since this was the work of the author.\u003cbr\u003e\u0026nbsp;Availability of data and material: Not applicable.\u003cbr\u003e\u0026nbsp;Competing interests: The author has no competing interests to declare that are relevant to the content of this article.\u003cbr\u003e\u0026nbsp;Funding: No funding was received to assist with the preparation of this manuscript.\u003cbr\u003e\u0026nbsp;Authors' contributions: Tina Comes is the sole author of this manuscript.\u003cbr\u003e\u0026nbsp;Acknowledgements: Not applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbbass, H. A. (2019). Social Integration of Artificial Intelligence: Functions, Automation Allocation Logic and Human-Autonomy Trust. \u003cem\u003eCognitive Computation\u003c/em\u003e,\u003cem\u003e 11\u003c/em\u003e(2), 159-171. https://doi.org/10.1007/s12559-018-9619-0 \u003c/li\u003e\n\u003cli\u003eAcharya, D. B., Kuppan, K., \u0026amp; Divya, B. (2025). Agentic AI: Autonomous Intelligence for Complex Goals\u0026mdash;A Comprehensive Survey. \u003cem\u003eIEEE Access\u003c/em\u003e,\u003cem\u003e 13\u003c/em\u003e, 18912-18936. https://doi.org/10.1109/ACCESS.2025.3532853 \u003c/li\u003e\n\u003cli\u003eAkata, Z., Balliet, D., Rijke, M. d., Dignum, F., Dignum, V., Eiben, G., Fokkens, A., Grossi, D., Hindriks, K., Hoos, H., Hung, H., Jonker, C., Monz, C., Neerincx, M., Oliehoek, F., Prakken, H., Schlobach, S., Gaag, L. v. d., Harmelen, F. v.,\u0026hellip;Welling, M. (2020). A Research Agenda for Hybrid Intelligence: Augmenting Human Intellect With Collaborative, Adaptive, Responsible, and Explainable Artificial Intelligence. \u003cem\u003eComputer\u003c/em\u003e,\u003cem\u003e 53\u003c/em\u003e(8), 18-28. https://doi.org/10.1109/MC.2020.2996587 \u003c/li\u003e\n\u003cli\u003eAkula, R., \u0026amp; Garibay, I. (2021, 2021//). Ethical AI for Social Good. HCI International 2021 - Late Breaking Papers: Multimodality, eXtended Reality, and Artificial Intelligence, Cham.\u003c/li\u003e\n\u003cli\u003eAlix, C., Lafond, D., Mattioli, J., Heer, J. D., Chattington, M., \u0026amp; Robic, P. O. (2021, 14-18 June 2021). Empowering Adaptive Human Autonomy Collaboration with Artificial Intelligence. 2021 16th International Conference of System of Systems Engineering (SoSE), \u003c/li\u003e\n\u003cli\u003eAuernhammer, J. (2020). Human-centered AI: The role of Human-centered Design Research in the development of AI. \u003c/li\u003e\n\u003cli\u003eBansal, G., Wu, T., Zhou, J., Fok, R., Nushi, B., Kamar, E., Ribeiro, M. T., \u0026amp; Weld, D. (2021, 2021). Does the whole exceed its parts? the effect of ai explanations on complementary team performance. CHI, New York.\u003c/li\u003e\n\u003cli\u003eBaum, S. D. (2020). Social choice ethics in artificial intelligence. \u003cem\u003eAI \u0026amp; society\u003c/em\u003e,\u003cem\u003e 35\u003c/em\u003e(1), 165-176. https://doi.org/10.1007/s00146-017-0760-1 \u003c/li\u003e\n\u003cli\u003eBingley, W. J., Haslam, S. A., Steffens, N. K., Gillespie, N., Worthy, P., Curtis, C., Lockey, S., Bialkowski, A., Ko, R. K. L., \u0026amp; Wiles, J. (2023). Enlarging the model of the human at the heart of human-centered AI: A social self-determination model of AI system impact. \u003cem\u003eNew Ideas in Psychology\u003c/em\u003e,\u003cem\u003e 70\u003c/em\u003e. https://doi.org/doi:10.1016/j.newideapsych.2023.101025 \u003c/li\u003e\n\u003cli\u003eBu\u0026ccedil;inca, Z. (2024). Optimizing Decision-Maker\u0026apos;s Intrinsic Motivation for Effective Human-AI Decision-Making. \u003cem\u003eExtended Abstracts of the CHI Conference on Human Factors in Computing Systems\u003c/em\u003e. https://doi.org/info:doi/10.1145/3613905.3638179 \u003c/li\u003e\n\u003cli\u003eCapel, T., \u0026amp; Brereton, M. (2023). \u003cem\u003eWhat is Human-Centered about Human-Centered AI? A Map of the Research Landscape\u003c/em\u003e Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, Hamburg, Germany. https://doi-org.tudelft.idm.oclc.org/10.1145/3544548.3580959\u003c/li\u003e\n\u003cli\u003eCavalcante Siebert, L., Lupetti, M. L., Aizenberg, E., Beckers, N., Zgonnikov, A., Veluwenkamp, H., Abbink, D., Giaccardi, E., Houben, G.-J., \u0026amp; Jonker, C. M. (2023). Meaningful human control: actionable properties for AI system development. \u003cem\u003eAI and Ethics\u003c/em\u003e,\u003cem\u003e 3\u003c/em\u003e(1), 241-255. \u003c/li\u003e\n\u003cli\u003eChaudhry, M. A., Cukurova, M., \u0026amp; Luckin, R. (2022, 2022//). A Transparency Index Framework for AI in Education. Artificial Intelligence in Education. Posters and Late Breaking Results, Workshops and Tutorials, Industry and Innovation Tracks, Practitioners\u0026rsquo; and Doctoral Consortium, Cham.\u003c/li\u003e\n\u003cli\u003eChorus, C. G., Pudāne, B., Mouter, N., \u0026amp; Campbell, D. (2018). Taboo trade-off aversion: A discrete choice model and empirical analysis. \u003cem\u003eJournal of Choice Modelling\u003c/em\u003e,\u003cem\u003e 27\u003c/em\u003e, 37-49. https://doi.org/https://doi.org/10.1016/j.jocm.2017.09.002 \u003c/li\u003e\n\u003cli\u003eCoeckelbergh, M. (2020). Artificial intelligence, responsibility attribution, and a relational justification of explainability. \u003cem\u003eScience and engineering ethics\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(4), 2051-2068. \u003c/li\u003e\n\u003cli\u003eColeman, N., Li, X., Comes, T., \u0026amp; Mostafavi, A. (2024). Weaving equity into infrastructure resilience research: a decadal review and future directions. \u003cem\u003enpj Natural Hazards\u003c/em\u003e,\u003cem\u003e 1\u003c/em\u003e(1), 25. https://doi.org/10.1038/s44304-024-00022-x \u003c/li\u003e\n\u003cli\u003eComes, T. (2024). AI for crisis decisions. \u003cem\u003eEthics and Information Technology\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(1), 12. https://doi.org/10.1007/s10676-024-09750-0 \u003c/li\u003e\n\u003cli\u003eComes, T., Van de Walle, B., \u0026amp; Van Wassenhove, L. (2020). The coordination‐information bubble in humanitarian response: theoretical foundations and empirical investigations. \u003cem\u003eProduction and Operations Management\u003c/em\u003e,\u003cem\u003e 29\u003c/em\u003e(11), 2484-2507. \u003c/li\u003e\n\u003cli\u003eCooke, N., Demir, M., \u0026amp; Huang, L. (2020, 2020). A framework for human-autonomy team research. HCII 2020, Copenhagen, Denmark.\u003c/li\u003e\n\u003cli\u003eCoppi, G., Moreno Jimenez, R., \u0026amp; Kyriazi, S. (2021). Explicability of humanitarian AI: a matter of principles. \u003cem\u003eJournal of International Humanitarian Action\u003c/em\u003e,\u003cem\u003e 6\u003c/em\u003e(1), 19. \u003c/li\u003e\n\u003cli\u003eCouncil, N. R. (1998). \u003cem\u003eThe future of air traffic control: Human operators and automation\u003c/em\u003e (C. Wickens, A. Mavor, R. Parasuraman, \u0026amp; J. McGee, Eds.). National Academies Press. \u003c/li\u003e\n\u003cli\u003eCriscuolo, C., \u0026amp; Dolci, T. (2024, 13-16 May 2024). Exploring Fairness Interpretability with FairnessFriend: A Chatbot Solution. 2024 IEEE 40th International Conference on Data Engineering Workshops (ICDEW), \u003c/li\u003e\n\u003cli\u003eDaw, T. M., Coulthard, S., Cheung, W. W. L., Brown, K., Abunge, C., Galafassi, D., Peterson, G. D., McClanahan, T. R., Omukoto, J. O., \u0026amp; Munyi, L. (2015). Evaluating taboo trade-offs in ecosystems services and human well-being. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e,\u003cem\u003e 112\u003c/em\u003e(22), 6949-6954. https://doi.org/10.1073/pnas.1414900112 \u003c/li\u003e\n\u003cli\u003eDe Wolf, T., \u0026amp; Holvoet, T. (2003, 2003). Towards autonomic computing: agent-based modelling, dynamical systems analysis, and decentralised control. \u003c/li\u003e\n\u003cli\u003eDuan, X., Pei, B., Ambrose, G. A., Hershkovitz, A., Cheng, Y., \u0026amp; Wang, C. (2024). Towards transparent and trustworthy prediction of student learning achievement by including instructors as co-designers: a case study. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e,\u003cem\u003e 29\u003c/em\u003e(3), 3075-3096. https://doi.org/10.1007/s10639-023-11954-8 \u003c/li\u003e\n\u003cli\u003eEC. (2019). \u003cem\u003eEthics guidelines for trustworthy AI\u003c/em\u003e. https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai\u003c/li\u003e\n\u003cli\u003eElahi, H., Castiglione, A., Wang, G., \u0026amp; Geman, O. (2021). A human-centered artificial intelligence approach for privacy protection of elderly App users in smart cities. \u003cem\u003eNeurocomputing\u003c/em\u003e,\u003cem\u003e 444\u003c/em\u003e, 189-202. https://doi.org/https://doi.org/10.1016/j.neucom.2020.06.149 \u003c/li\u003e\n\u003cli\u003eEndsley, M. R. (2017). From here to autonomy: lessons learned from human\u0026ndash;automation research. \u003cem\u003eHuman factors\u003c/em\u003e,\u003cem\u003e 59\u003c/em\u003e(1), 5-27. \u003c/li\u003e\n\u003cli\u003eEndsley, M. R. (2023). Supporting Human-AI Teams:Transparency, explainability, and situation awareness. \u003cem\u003eComputers in Human Behavior\u003c/em\u003e,\u003cem\u003e 140\u003c/em\u003e, 107574. https://doi.org/https://doi.org/10.1016/j.chb.2022.107574 \u003c/li\u003e\n\u003cli\u003eErlei, A. (2024). Understanding Choice Independence and Error Types in Human-AI Collaboration. \u003cem\u003eProceedings of the CHI Conference on Human Factors in Computing Systems\u003c/em\u003e. https://doi.org/info:doi/10.1145/3613904.3641946 \u003c/li\u003e\n\u003cli\u003eFan, C., Zhang, C., Yahja, A., \u0026amp; Mostafavi, A. (2021). Disaster City Digital Twin: A vision for integrating artificial and human intelligence for disaster management. \u003cem\u003eInternational Journal of Information Management\u003c/em\u003e,\u003cem\u003e 56\u003c/em\u003e, 102049. \u003c/li\u003e\n\u003cli\u003eFitts, P. M. (1951). Human engineering for an effective air-navigation and traffic-control system. \u003c/li\u003e\n\u003cli\u003eFlathmann, C., Schelble, B. G., Rosopa, P. J., McNeese, N. J., Mallick, R., \u0026amp; Madathil, K. C. (2023). Examining the impact of varying levels of AI teammate influence on human-AI teams. \u003cem\u003eInternational Journal of Human Computer Studies\u003c/em\u003e,\u003cem\u003e 177\u003c/em\u003e. https://doi.org/doi:10.1016/j.ijhcs.2023.103061 \u003c/li\u003e\n\u003cli\u003eFreire, S. K., Niforatos, E., Wang, C., Ruiz-Arenas, S., Foosherian, M., Wellsandt, S., \u0026amp; Bozzon, A. (2023). \u003cem\u003eLessons Learned from Designing and Evaluating CLAICA: A Continuously Learning AI Cognitive Assistant\u003c/em\u003e Proceedings of the 28th International Conference on Intelligent User Interfaces, Sydney, NSW, Australia. https://doi-org.tudelft.idm.oclc.org/10.1145/3581641.3584042\u003c/li\u003e\n\u003cli\u003eFrench, S. (1986). \u003cem\u003eDecision Theory: An introduction of the Mathematics of Rationality\u003c/em\u003e. Ellis Harwood Limited. \u003c/li\u003e\n\u003cli\u003eFrench, S. (2012). Cynefin, statistics and decision analysis. \u003cem\u003eJournal of the Operational Research Society\u003c/em\u003e. https://doi.org/10.1057/jors.2012.23 \u003c/li\u003e\n\u003cli\u003eGajos, K. Z., \u0026amp; Mamykina, L. (2022). \u003cem\u003eDo People Engage Cognitively with AI? Impact of AI Assistance on Incidental Learning\u003c/em\u003e Proceedings of the 27th International Conference on Intelligent User Interfaces, Helsinki, Finland. https://doi-org.tudelft.idm.oclc.org/10.1145/3490099.3511138\u003c/li\u003e\n\u003cli\u003eGil, Y., Garijo, D., Khider, D., Knoblock, C. A., Ratnakar, V., Osorio, M., Vargas, H., Pham, M., Pujara, J., \u0026amp; Shbita, B. (2021). Artificial intelligence for modeling complex systems: taming the complexity of expert models to improve decision making. \u003cem\u003eACM Transactions on Interactive Intelligent Systems\u003c/em\u003e,\u003cem\u003e 11\u003c/em\u003e(2), 1-49. \u003c/li\u003e\n\u003cli\u003eGou, J., Liang, Q., Wang, Z., \u0026amp; Dabić, M. (2024). Affordances and Constraints of Automation and Augmentation: Lessons Learned From Development of a Human-AI Collaboration Business Simulation Platform. \u003cem\u003eJournal of Global Information Management\u003c/em\u003e,\u003cem\u003e 32\u003c/em\u003e(1), 1-27. https://doi.org/https://doi.org/10.4018/JGIM.357260 \u003c/li\u003e\n\u003cli\u003eGralla, E., Goentzel, J., \u0026amp; Fine, C. (2016). Problem Formulation and Solution Mechanisms : A Behavioral Study of Humanitarian Transportation Planning. \u003cem\u003ePOMS\u003c/em\u003e,\u003cem\u003e 25\u003c/em\u003e(1), 22-35. https://doi.org/10.1111/poms.12496 \u003c/li\u003e\n\u003cli\u003eHagemann, V., Rieth, M., Suresh, A., \u0026amp; Kirchner, F. (2023). Human-AI teams\u0026mdash;Challenges for a team-centered AI at work. \u003cem\u003eFrontiers in Artificial Intelligence\u003c/em\u003e,\u003cem\u003e 6\u003c/em\u003e. https://doi.org/doi:10.3389/frai.2023.1252897 \u003c/li\u003e\n\u003cli\u003eHe, J., Piorkowski, D., Muller, M. J., Brimijoin, K., Houde, S., \u0026amp; Weisz, J. D. (2023, 2023). Understanding How Task Dimensions Impact Automation Preferences with a Conversational Task Assistant. CHI AutomationXP, Hamburg.\u003c/li\u003e\n\u003cli\u003eHelms Mills, J., Thurlow, A., \u0026amp; Mills, A. J. (2010). Making sense of sensemaking: the critical sensemaking approach. \u003cem\u003eQualitative Research in Organizations and Management: An International Journal\u003c/em\u003e,\u003cem\u003e 5\u003c/em\u003e(2), 182-195. https://doi.org/10.1108/17465641011068857 \u003c/li\u003e\n\u003cli\u003eHerrmann, T., \u0026amp; Pfeiffer, S. (2023). Keeping the organization in the loop: a socio-technical extension of human-centered artificial intelligence. \u003cem\u003eAI \u0026amp; SOCIETY\u003c/em\u003e,\u003cem\u003e 38\u003c/em\u003e(4), 1523-1542. https://doi.org/10.1007/s00146-022-01391-5 \u003c/li\u003e\n\u003cli\u003eHeyndels, S. (2023). Technology and neutrality. \u003cem\u003ePhilosophy \u0026amp; Technology\u003c/em\u003e,\u003cem\u003e 36\u003c/em\u003e(4), 75. \u003c/li\u003e\n\u003cli\u003eHoch, T., Heinzl, B., Czech, G., Khan, M., Waibel, P., Bachhofner, S., Kiesling, E., \u0026amp; Moser, B. (2022). \u003cem\u003eTeaming. AI: enabling human-AI teaming intelligence in manufacturing\u003c/em\u003e I-ESA, Valencia, Spain. \u003c/li\u003e\n\u003cli\u003eHolguin-Veras, J., Perez, N., Jaller, M., Van Wassenhove, L. N., \u0026amp; Aros-Vera, F. (2013). On the appropriate objective function for post-disaster humanitarian logistics models. \u003cem\u003eJournal of Operations Management\u003c/em\u003e,\u003cem\u003e 31\u003c/em\u003e(5), 262-280. https://doi.org/10.1016/j.jom.2013.06.002 \u003c/li\u003e\n\u003cli\u003eHollingshead, A. B., McGrath, J. E., \u0026amp; O\u0026apos;Connor, K. M. (1993). Group Task Performance and Communication Technology: A Longitudinal Study of Computer-Mediated Versus Face-to-Face Work Groups. \u003cem\u003eSmall Group Research\u003c/em\u003e,\u003cem\u003e 24\u003c/em\u003e(3), 307-333. https://doi.org/10.1177/1046496493243003 \u003c/li\u003e\n\u003cli\u003eIEEE. (2019). Ethically Aligned Design - A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems. In \u003cem\u003eEthically Aligned Design - A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems\u003c/em\u003e (pp. 1-294): IEEE.\u003c/li\u003e\n\u003cli\u003eJacob, C., Kerrigan, P., \u0026amp; Bastos, M. (2025). The chat-chamber effect: Trusting the AI hallucination. \u003cem\u003eBig Data \u0026amp; Society\u003c/em\u003e,\u003cem\u003e 12\u003c/em\u003e(1), 20539517241306345. \u003c/li\u003e\n\u003cli\u003eJennings, N. R., Moreau, L., Nicholson, D., Ramchurn, S., Roberts, S., Rodden, T., \u0026amp; Rogers, A. (2014). Human-agent collectives. \u003cem\u003eCommunications of the ACM\u003c/em\u003e,\u003cem\u003e 57\u003c/em\u003e(12), 80-88. \u003c/li\u003e\n\u003cli\u003eJin, L., Boden, A., \u0026amp; Shajalal, M. (2022). \u003cem\u003eAutomated Decision Making Systems in Smart Homes: A Study on User Engagement and Design\u003c/em\u003e (Vol. 3154). https://doi.org/doi: \u003c/li\u003e\n\u003cli\u003eKahneman, D., \u0026amp; Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. \u003cem\u003eEconometrica\u003c/em\u003e,\u003cem\u003e 47\u003c/em\u003e(2), 263-291. \u003c/li\u003e\n\u003cli\u003eKattnig, M., Angerschmid, A., Reichel, T., \u0026amp; Kern, R. (2024). Assessing trustworthy AI: Technical and legal perspectives of fairness in AI. \u003cem\u003eComputer Law \u0026amp; Security Review\u003c/em\u003e,\u003cem\u003e 55\u003c/em\u003e, 106053. https://doi.org/https://doi.org/10.1016/j.clsr.2024.106053 \u003c/li\u003e\n\u003cli\u003eKiesler, S., \u0026amp; Sproull, L. (1992). Group decision making and communication technology. \u003cem\u003eOrganizational Behavior and Human Decision Processes\u003c/em\u003e,\u003cem\u003e 52\u003c/em\u003e(1), 96-123. https://doi.org/https://doi.org/10.1016/0749-5978(92)90047-B \u003c/li\u003e\n\u003cli\u003eKim, J. S., Kim, M., \u0026amp; Baek, T. H. (2025). Enhancing User Experience With a Generative AI Chatbot. \u003cem\u003eInternational Journal of Human\u0026ndash;Computer Interaction\u003c/em\u003e,\u003cem\u003e 41\u003c/em\u003e(1), 651-663. https://doi.org/10.1080/10447318.2024.2311971 \u003c/li\u003e\n\u003cli\u003eKj\u0026aelig;rum, A., \u0026amp; Madsen, B. S. (2025). Pushing the boundaries of anticipatory action using machine learning. \u003cem\u003eData \u0026amp; Policy\u003c/em\u003e,\u003cem\u003e 7\u003c/em\u003e, e8. \u003c/li\u003e\n\u003cli\u003eKlein, G., Calderwood, R., \u0026amp; Clinton-Cirocco, A. (2010). Rapid Decision Making on the Fire Ground: The Original Study Plus a Postscript. \u003cem\u003eJournal of Cognitive Engineering and Decision Making\u003c/em\u003e,\u003cem\u003e 4\u003c/em\u003e(3), 186-209. https://doi.org/10.1518/155534310X12844000801203 \u003c/li\u003e\n\u003cli\u003eKunar, M. A., Montana, G., \u0026amp; Watson, D. G. (2024). Increasing transparency of computer-aided detection impairs decision-making in visual search. \u003cem\u003ePsychonomic Bulletin \u0026amp; Review\u003c/em\u003e. https://doi.org/10.3758/s13423-024-02601-5 \u003c/li\u003e\n\u003cli\u003eLee, M. H., Siewiorek, D. P., Smailagic, A., Bernardino, A., \u0026amp; Berm\u0026uacute;dez i Badia, S. (2022, 2022). Towards efficient annotations for a human-ai collaborative, clinical decision support system: A case study on physical stroke rehabilitation assessment. IUI, Helsinki, Finland.\u003c/li\u003e\n\u003cli\u003eLee, M. K., Kusbit, D., Kahng, A., Kim, J. T., Yuan, X., Chan, A., See, D., Noothigattu, R., Lee, S., Psomas, A., \u0026amp; Procaccia, A. D. (2019). Webuildai: Participatory framework for algorithmic governance. \u003cem\u003eProceedings of the ACM on Human-Computer Interaction\u003c/em\u003e,\u003cem\u003e 3\u003c/em\u003e. https://doi.org/doi:10.1145/3359283 \u003c/li\u003e\n\u003cli\u003eLettieri, N., Guarino, A., Zaccagnino, R., \u0026amp; Malandrino, D. (2023). Keeping judges in the loop: a human\u0026ndash;machine collaboration strategy against the blind spots of AI in criminal justice. \u003cem\u003eSoft Computing\u003c/em\u003e,\u003cem\u003e 27\u003c/em\u003e(16), 11275-11293. https://doi.org/10.1007/s00500-023-08604-z \u003c/li\u003e\n\u003cli\u003eLevin, K., Cashore, B., Bernstein, S., \u0026amp; Auld, G. (2012). Overcoming the tragedy of super wicked problems: constraining our future selves to ameliorate global climate change. \u003cem\u003ePolicy Sciences\u003c/em\u003e,\u003cem\u003e 45\u003c/em\u003e(2), 123-152. \u003c/li\u003e\n\u003cli\u003eLi, X., Zong, Q., \u0026amp; Cheng, M. (2024). The Impact of Medical Explainable Artificial Intelligence on Nurses\u0026rsquo; Innovation Behaviour: A Structural Equation Modelling Approach. \u003cem\u003eJournal of Nursing Management\u003c/em\u003e,\u003cem\u003e 2024\u003c/em\u003e(1), 8885760. https://doi.org/https://doi.org/10.1155/2024/8885760 \u003c/li\u003e\n\u003cli\u003eLiao, Q. V., \u0026amp; Sundar, S. S. (2022). \u003cem\u003eDesigning for Responsible Trust in AI Systems: A Communication Perspective\u003c/em\u003e Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, Seoul, Republic of Korea. https://doi-org.tudelft.idm.oclc.org/10.1145/3531146.3533182\u003c/li\u003e\n\u003cli\u003eLu, N., Liu, L., Yu, D., \u0026amp; Fu, B. (2021). Navigating trade-offs in the social-ecological systems. \u003cem\u003eCurrent Opinion in Environmental Sustainability\u003c/em\u003e,\u003cem\u003e 48\u003c/em\u003e, 77-84. https://doi.org/https://doi.org/10.1016/j.cosust.2020.10.014 \u003c/li\u003e\n\u003cli\u003eMahajan, S., Hausladen, C. I., S\u0026aacute;nchez-Vaquerizo, J. A., Korecki, M., \u0026amp; Helbing, D. (2022). Participatory resilience: Surviving, recovering and improving together. \u003cem\u003eSustainable Cities and Society\u003c/em\u003e,\u003cem\u003e 83\u003c/em\u003e, 103942. \u003c/li\u003e\n\u003cli\u003eMaiden, N., Lockerbie, J., Zachos, K., Wolf, A., \u0026amp; Brown, A. (2023). Designing new digital tools to augment human creative thinking at work: An application in elite sports coaching. \u003cem\u003eExpert Systems\u003c/em\u003e,\u003cem\u003e 40\u003c/em\u003e(3), e13194. https://doi.org/https://doi.org/10.1111/exsy.13194 \u003c/li\u003e\n\u003cli\u003eMaitlis, S., \u0026amp; Christianson, M. (2014). Sensemaking in Organizations: Taking Stock and Moving Forward. \u003cem\u003eThe Academy of Management Annals\u003c/em\u003e,\u003cem\u003e 8\u003c/em\u003e(1), 57-125. https://doi.org/10.1080/19416520.2014.873177 \u003c/li\u003e\n\u003cli\u003eMalone, T. W., Malone, T. W., Crowston, K., \u0026amp; Crowston, K. (1994). The interdisciplinary study of coordination. \u003cem\u003eACM Computing Surveys\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(1), 87-119. https://doi.org/http://doi.acm.org/10.1145/174666.174668 \u003c/li\u003e\n\u003cli\u003eMatthias, A. (2004). The responsibility gap: Ascribing responsibility for the actions of learning automata. \u003cem\u003eEthics and Information Technology\u003c/em\u003e,\u003cem\u003e 6\u003c/em\u003e(3), 175-183. https://doi.org/10.1007/s10676-004-3422-1 \u003c/li\u003e\n\u003cli\u003eMuhren, W. J., \u0026amp; Van de Walle, B. (2010). A Call for Sensemaking Support Systems in Crisis Management. In R. Babu\u0026scaron;ka \u0026amp; F. C. A. Groen (Eds.), \u003cem\u003eInteractive Collaborative Information Systems\u003c/em\u003e (pp. 425-452). Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-11688-9_16 \u003c/li\u003e\n\u003cli\u003eMuller, J. (2018). \u003cem\u003eThe tyranny of metrics\u003c/em\u003e. Princeton University Press. \u003c/li\u003e\n\u003cli\u003eMunn, Z., Peters, M. D. J., Stern, C., Tufanaru, C., McArthur, A., \u0026amp; Aromataris, E. (2018). Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. \u003cem\u003eBMC Medical Research Methodology\u003c/em\u003e,\u003cem\u003e 18\u003c/em\u003e(1), 143. https://doi.org/10.1186/s12874-018-0611-x \u003c/li\u003e\n\u003cli\u003eNabizadeh Rafsanjani, H., \u0026amp; Nabizadeh, A. H. (2023). Towards human-centered artificial intelligence (AI) in architecture, engineering, and construction (AEC) industry. \u003cem\u003eComputers in Human Behavior Reports\u003c/em\u003e,\u003cem\u003e 11\u003c/em\u003e. https://doi.org/doi:10.1016/j.chbr.2023.100319 \u003c/li\u003e\n\u003cli\u003eNeale, D. C., Carroll, J. M., \u0026amp; Rosson, M. B. (2004). \u003cem\u003eEvaluating computer-supported cooperative work: models and frameworks\u003c/em\u003e Proceedings of the 2004 ACM conference on Computer supported cooperative work, Chicago, Illinois, USA. https://doi-org.tudelft.idm.oclc.org/10.1145/1031607.1031626\u003c/li\u003e\n\u003cli\u003eNespeca, V., Comes, T., \u0026amp; Brazier, F. (2021). A methodology to develop agent-based models for policy design in socio-technical systems based on qualitative inquiry. In M. Czupryna \u0026amp; B. Kamiński (Eds.), \u003cem\u003eSocial Simulation Conference 2021\u003c/em\u003e. Springer. \u003c/li\u003e\n\u003cli\u003eOECD. (2019). \u003cem\u003eRecommendation of the Council on Artificial Intelligence\u003c/em\u003e. https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449\u003c/li\u003e\n\u003cli\u003eOzmen Garibay, O., Winslow, B., Andolina, S., Antona, M., Bodenschatz, A., Coursaris, C., Falco, G., Fiore, S. M., Garibay, I., Grieman, K., Havens, J. C., Jirotka, M., Kacorri, H., Karwowski, W., Kider, J., Konstan, J., Koon, S., Lopez-Gonzalez, M., Maifeld-Carucci, I.,\u0026hellip;Xu, W. (2023). Six Human-Centered Artificial Intelligence Grand Challenges. \u003cem\u003eInternational Journal of Human\u0026ndash;Computer Interaction\u003c/em\u003e,\u003cem\u003e 39\u003c/em\u003e(3), 391-437. https://doi.org/10.1080/10447318.2022.2153320 \u003c/li\u003e\n\u003cli\u003eParasuraman, R., Sheridan, T. B., \u0026amp; Wickens, C. D. (2000). A model for types and levels of human interaction with automation. \u003cem\u003eIEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans\u003c/em\u003e,\u003cem\u003e 30\u003c/em\u003e(3), 286-297. https://doi.org/10.1109/3468.844354 \u003c/li\u003e\n\u003cli\u003ePasslack, N., Hammerschmidt, T., \u0026amp; Posegga, O. (2024). It was(n\u0026rsquo;t) me: vignette experiments on managers\u0026rsquo; responsibility attribution in AI-advised decision-making. \u003cem\u003eBehaviour \u0026amp; Information Technology\u003c/em\u003e, 1-30. https://doi.org/10.1080/0144929X.2024.2431050 \u003c/li\u003e\n\u003cli\u003ePaulus, D., Fathi, R., Fiedrich, F., de Walle, B. V., \u0026amp; Comes, T. (2022). On the Interplay of Data and Cognitive Bias in Crisis Information Management. \u003cem\u003eInformation Systems Frontiers\u003c/em\u003e. https://doi.org/10.1007/s10796-022-10241-0 \u003c/li\u003e\n\u003cli\u003ePloeg, J., Semsar-Kazerooni, E., Lijster, G., Van de Wouw, N., \u0026amp; Nijmeijer, H. (2014). Graceful degradation of cooperative adaptive cruise control. \u003cem\u003eIEEE Transactions on Intelligent Transportation Systems\u003c/em\u003e,\u003cem\u003e 16\u003c/em\u003e(1), 488-497. \u003c/li\u003e\n\u003cli\u003ePonti, M., \u0026amp; Seredko, A. (2022). Human-machine-learning integration and task allocation in citizen science. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e,\u003cem\u003e 9\u003c/em\u003e(1), 48. https://doi.org/10.1057/s41599-022-01049-z \u003c/li\u003e\n\u003cli\u003eRahwan, I., Cebrian, M., Obradovich, N., Bongard, J., Bonnefon, J.-F., Breazeal, C., Crandall, J. W., Christakis, N. A., Couzin, I. D., \u0026amp; Jackson, M. O. (2019). Machine behaviour. \u003cem\u003eNature\u003c/em\u003e,\u003cem\u003e 568\u003c/em\u003e(7753), 477-486. \u003c/li\u003e\n\u003cli\u003eRiedl, M. O. (2019). Human‐centered artificial intelligence and machine learning. \u003cem\u003eHuman behavior and emerging technologies\u003c/em\u003e,\u003cem\u003e 1\u003c/em\u003e(1), 33-36. \u003c/li\u003e\n\u003cli\u003eRittel, H. W. J., \u0026amp; Webber, M. M. (1973). Dilemmas in a general theory of planning. \u003cem\u003ePolicy Sciences\u003c/em\u003e,\u003cem\u003e 4\u003c/em\u003e(2), 155-169. \u003c/li\u003e\n\u003cli\u003eSadek, M., Kallina, E., Bohn\u0026eacute;, T., Mougenot, C., Calvo, R. A., \u0026amp; Cave, S. (2024). Challenges of responsible AI in practice: scoping review and recommended actions. \u003cem\u003eAI \u0026amp; society\u003c/em\u003e. https://doi.org/10.1007/s00146-024-01880-9 \u003c/li\u003e\n\u003cli\u003eS\u0026aelig;tre, A. S., \u0026amp; Van de Ven, A. (2021). Generating theory by abduction. \u003cem\u003eAcademy of Management Review\u003c/em\u003e,\u003cem\u003e 46\u003c/em\u003e(4), 684-701. \u003c/li\u003e\n\u003cli\u003eSeidel, S., Chandra Kruse, L., Sz\u0026eacute;kely, N., Gau, M., \u0026amp; Stieger, D. (2018). Design principles for sensemaking support systems in environmental sustainability transformations. \u003cem\u003eEuropean Journal of Information Systems\u003c/em\u003e,\u003cem\u003e 27\u003c/em\u003e(2), 221-247. https://doi.org/10.1057/s41303-017-0039-0 \u003c/li\u003e\n\u003cli\u003eSellen, A., \u0026amp; Horvitz, E. (2024). The Rise of the AI Co-Pilot: Lessons for Design from Aviation and Beyond. \u003cem\u003eCommun. ACM\u003c/em\u003e,\u003cem\u003e 67\u003c/em\u003e(7), 18\u0026ndash;23. https://doi.org/10.1145/3637865 \u003c/li\u003e\n\u003cli\u003eSharma, N., Liao, Q. V., \u0026amp; Xiao, Z. (2024). \u003cem\u003eGenerative Echo Chamber? Effect of LLM-Powered Search Systems on Diverse Information Seeking\u003c/em\u003e Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, Honolulu, HI, USA. https://doi-org.tudelft.idm.oclc.org/10.1145/3613904.3642459\u003c/li\u003e\n\u003cli\u003eSharoda, P. A., \u0026amp; Reddy, M. C. (2010). \u003cem\u003eUnderstanding Together: Sensemaking in Collaborative Information Seeking\u003c/em\u003e CSCW, \u003c/li\u003e\n\u003cli\u003eSheridan, T. B., Verplank, W. L., \u0026amp; Brooks, T. L. (1978, 1978). Human/computer control of undersea teleoperators. \u003c/li\u003e\n\u003cli\u003eShneiderman, B. (2020). Human-centered artificial intelligence: Reliable, safe \u0026amp; trustworthy. \u003cem\u003eInternational Journal of Human\u0026ndash;Computer Interaction\u003c/em\u003e,\u003cem\u003e 36\u003c/em\u003e(6), 495-504. \u003c/li\u003e\n\u003cli\u003eShulner-Tal, A., Kuflik, T., \u0026amp; Kliger, D. (2023). Enhancing Fairness Perception\u0026ndash;Towards Human-Centred AI and Personalized Explanations Understanding the Factors Influencing Laypeople\u0026rsquo;s Fairness Perceptions of Algorithmic Decisions. \u003cem\u003eInternational Journal of Human-Computer Interaction\u003c/em\u003e,\u003cem\u003e 39\u003c/em\u003e(7), 1455-1482. https://doi.org/doi:10.1080/10447318.2022.2095705 \u003c/li\u003e\n\u003cli\u003eSigfrids, A., Leikas, J., Salo-P\u0026ouml;ntinen, H., \u0026amp; Koskimies, E. (2023). Human-centricity in AI governance: A systemic approach [Perspective]. \u003cem\u003eFrontiers in Artificial Intelligence\u003c/em\u003e,\u003cem\u003e 6\u003c/em\u003e. https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2023.976887 \u003c/li\u003e\n\u003cli\u003eSloane, M., Solano-Kamaiko, I. R., Yuan, J., Dasgupta, A., \u0026amp; Stoyanovich, J. (2023). Introducing contextual transparency for automated decision systems. \u003cem\u003eNature Machine Intelligence\u003c/em\u003e,\u003cem\u003e 5\u003c/em\u003e(3), 187-195. https://doi.org/10.1038/s42256-023-00623-7 \u003c/li\u003e\n\u003cli\u003eS\u0026oslash;gaard J\u0026oslash;rgensen, P., Jansen, R. E. V., Avila Ortega, D. I., Wang-Erlandsson, L., Donges, J. F., \u0026Ouml;sterblom, H., Olsson, P., Nystr\u0026ouml;m, M., Lade, S. J., \u0026amp; Hahn, T. (2024). Evolution of the polycrisis: Anthropocene traps that challenge global sustainability. \u003cem\u003ePhilosophical Transactions of the Royal Society B\u003c/em\u003e,\u003cem\u003e 379\u003c/em\u003e(1893), 20220261. \u003c/li\u003e\n\u003cli\u003eSprenkamp, K., Dolata, M., Schwabe, G., \u0026amp; Zavolokina, L. (2025). Data-driven intelligence in crisis: The case of Ukrainian refugee management. \u003cem\u003eGovernment Information Quarterly\u003c/em\u003e,\u003cem\u003e 42\u003c/em\u003e(1), 101978. https://doi.org/https://doi.org/10.1016/j.giq.2024.101978 \u003c/li\u003e\n\u003cli\u003eStapleton, L., Lee, M. H., Qing, D., Wright, M., Chouldechova, A., Holstein, K., Wu, Z. S., \u0026amp; Zhu, H. (2022). \u003cem\u003eImagining new futures beyond predictive systems in child welfare: A qualitative study with impacted stakeholders\u003c/em\u003e Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, Seoul, Republic of Korea. https://doi-org.tudelft.idm.oclc.org/10.1145/3531146.3533177\u003c/li\u003e\n\u003cli\u003eStasser, G., \u0026amp; Titus, W. (1985). Pooling of unshared information in group decision making: Biased information sampling during discussion. \u003cem\u003eJournal of Personality and Social Psychology\u003c/em\u003e,\u003cem\u003e 48\u003c/em\u003e(6), 1467-1478. https://doi.org/10.1037/0022-3514.48.6.1467 \u003c/li\u003e\n\u003cli\u003eStewart, D. D., \u0026amp; Stasser, G. (1995). Expert role assignment and information sampling during collective recall and decision making. \u003cem\u003eJournal of Personality and Social Psychology\u003c/em\u003e,\u003cem\u003e 69\u003c/em\u003e(4), 619-628. \u003c/li\u003e\n\u003cli\u003eStone, P., Kaminka, G., Kraus, S., \u0026amp; Rosenschein, J. (2010, 2010). Ad hoc autonomous agent teams: Collaboration without pre-coordination. \u003c/li\u003e\n\u003cli\u003eSuchan, J., Bhatt, M., \u0026amp; Varadarajan, S. (2021). Commonsense visual sensemaking for autonomous driving \u0026ndash; On generalised neurosymbolic online abduction integrating vision and semantics. \u003cem\u003eArtificial Intelligence\u003c/em\u003e,\u003cem\u003e 299\u003c/em\u003e. https://doi.org/doi:10.1016/j.artint.2021.103522 \u003c/li\u003e\n\u003cli\u003eSun, J. (2022). Investigating Explainability of Generative AI for Code through Scenario-based Design. \u003cem\u003e27th International Conference on Intelligent User Interfaces\u003c/em\u003e. https://doi.org/info:doi/10.1145/3490099.3511119 \u003c/li\u003e\n\u003cli\u003eTausch, A., \u0026amp; Kluge, A. (2022). The best task allocation process is to decide on one\u0026rsquo;s own: effects of the allocation agent in human\u0026ndash;robot interaction on perceived work characteristics and satisfaction. \u003cem\u003eCognition, Technology \u0026amp; Work\u003c/em\u003e,\u003cem\u003e 24\u003c/em\u003e(1), 39-55. https://doi.org/10.1007/s10111-020-00656-7 \u003c/li\u003e\n\u003cli\u003eTetlock, P. E. (2003). Thinking the unthinkable: sacred values and taboo cognitions. \u003cem\u003eTrends in Cognitive Sciences\u003c/em\u003e,\u003cem\u003e 7\u003c/em\u003e(7), 320-324. https://doi.org/https://doi.org/10.1016/S1364-6613(03)00135-9 \u003c/li\u003e\n\u003cli\u003eThompson, J. (2021). Mental Models and Interpretability in AI Fairness Tools and Code Environments. In C. Stephanidis, M. Kurosu, J. Y. C. Chen, G. Fragomeni, N. Streitz, S. i. Konomi, H. Degen, \u0026amp; S. Ntoa, \u003cem\u003eHCI International 2021 - Late Breaking Papers: Multimodality, eXtended Reality, and Artificial Intelligence\u003c/em\u003e Cham.\u003c/li\u003e\n\u003cli\u003eTricco, A. C., Lillie, E., Zarin, W., O\u0026apos;Brien, K. K., Colquhoun, H., Levac, D., Moher, D., Peters, M. D. J., Horsley, T., \u0026amp; Weeks, L. (2018). PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. \u003cem\u003eAnnals of internal medicine\u003c/em\u003e,\u003cem\u003e 169\u003c/em\u003e(7), 467-473. \u003c/li\u003e\n\u003cli\u003eUNESCO. (2022). \u003cem\u003eRecommendation on the Ethics of Artificial Intelligence\u003c/em\u003e. https://unesdoc.unesco.org/ark:/48223/pf0000381137\u003c/li\u003e\n\u003cli\u003eVallor, S. (2015). Moral Deskilling and Upskilling in a New Machine Age: Reflections on the Ambiguous Future of Character. \u003cem\u003ePhilosophy \u0026amp; Technology\u003c/em\u003e,\u003cem\u003e 28\u003c/em\u003e(1), 107-124. https://doi.org/10.1007/s13347-014-0156-9 \u003c/li\u003e\n\u003cli\u003eVan Berkel, N., Bellio, M., Skov, M. B., \u0026amp; Blandford, A. (2023). Measurements, algorithms, and presentations of reality: Framing interactions with AI-enabled decision support. \u003cem\u003eACM Transactions on Computer-Human Interaction\u003c/em\u003e,\u003cem\u003e 30\u003c/em\u003e(2), 1-33. \u003c/li\u003e\n\u003cli\u003evan Berkel, N., Tag, B., Goncalves, J., \u0026amp; Hosio, S. (2022). Human-centred artificial intelligence: a contextual morality perspective. \u003cem\u003eBehaviour \u0026amp; Information Technology\u003c/em\u003e,\u003cem\u003e 41\u003c/em\u003e(3), 502-518. https://doi.org/10.1080/0144929X.2020.1818828 \u003c/li\u003e\n\u003cli\u003evan Leersum, C. M., \u0026amp; Maathuis, C. (2025). Human centred explainable AI decision-making in healthcare. \u003cem\u003eJournal of Responsible Technology\u003c/em\u003e,\u003cem\u003e 21\u003c/em\u003e, 100108. https://doi.org/https://doi.org/10.1016/j.jrt.2025.100108 \u003c/li\u003e\n\u003cli\u003eVerma, H., Mlynar, J., Schaer, R., Reichenbach, J., Jreige, M., Prior, J., Ev\u0026eacute;quoz, F., \u0026amp; Depeursinge, A. (2023, 2023). Rethinking the role of AI with physicians in oncology: revealing perspectives from clinical and research workflows. CHI Human Factors, \u003c/li\u003e\n\u003cli\u003eVolkema, R. J. (1983). Problem Formulation in Planning and Design. \u003cem\u003eManagement Science\u003c/em\u003e,\u003cem\u003e 29\u003c/em\u003e(6), 639-652. https://doi.org/10.1287/mnsc.29.6.639 \u003c/li\u003e\n\u003cli\u003eWebster, M. (2008). Incorporating path dependency into decision-analytic methods: An application to global climate-change policy. \u003cem\u003eDecision Analysis\u003c/em\u003e,\u003cem\u003e 5\u003c/em\u003e(2), 60-75. \u003c/li\u003e\n\u003cli\u003eWeick, K., \u0026amp; Sutcliffe, K. (2007). Managing the unexpected: resilient performance in an age of uncertainty, second edition. http://dl.acm.org/citation.cfm?id=1408051 \u003c/li\u003e\n\u003cli\u003eWeick, K. E. (1993). The Collapse of Sensemaking in Organizations: The Mann Gulch Disaster. \u003cem\u003eAdministrative Science Quarterly\u003c/em\u003e,\u003cem\u003e 38\u003c/em\u003e(4), 628-652. \u003c/li\u003e\n\u003cli\u003eWeick, K. E. (1995). \u003cem\u003eSensemaking in organizations\u003c/em\u003e (Vol. 3). Sage publications Thousand Oaks, CA. \u003c/li\u003e\n\u003cli\u003eWeick, K. E. (2015). Ambiguity as grasp: The reworking of sense. \u003cem\u003eJournal of Contingencies and Crisis Management\u003c/em\u003e,\u003cem\u003e 23\u003c/em\u003e(2), 117-123. \u003c/li\u003e\n\u003cli\u003eWiener, N. (1948). Cybernetics. \u003cem\u003eScientific American\u003c/em\u003e,\u003cem\u003e 179\u003c/em\u003e(5), 14-19. \u003c/li\u003e\n\u003cli\u003eYazdanpanah, V., Gerding, E. H., Stein, S., Dastani, M., Jonker, C. M., \u0026amp; Norman, T. J. (2021). \u003cem\u003eResponsibility Research for Trustworthy Autonomous Systems\u003c/em\u003e AAMAS, Online. \u003c/li\u003e\n\u003cli\u003eZhang, R., Flathmann, C., Musick, G., Schelble, B., McNeese, N. J., Knijnenburg, B., \u0026amp; Duan, W. (2024). I Know This Looks Bad, But I Can Explain: Understanding When AI Should Explain Actions In Human-AI Teams. \u003cem\u003eACM Trans. Interact. Intell. Syst.\u003c/em\u003e,\u003cem\u003e 14\u003c/em\u003e(1), Article 6. https://doi.org/10.1145/3635474 \u003c/li\u003e\n\u003cli\u003eZwitter, A. (2024). Cybernetic governance: implications of technology convergence on governance convergence. \u003cem\u003eEthics and Information Technology\u003c/em\u003e,\u003cem\u003e 26\u003c/em\u003e(2), 24. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"epj-data-science","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"epds","sideBox":"Learn more about [EPJ Data Science](https://epjdatascience.springeropen.com/)","snPcode":"13688","submissionUrl":"https://submission.springernature.com/new-submission/13688/3","title":"EPJ Data Science","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Scoping Review, Sensemaking AI, Human-AI interaction, Decision Theory, Human-Centred AI, Complex Systems, Collective Intelligence, Optimisation: Human-AI networks","lastPublishedDoi":"10.21203/rs.3.rs-7463619/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7463619/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDigital technologies and AI promise to optimise complex systems through data-driven decisions, predictive modelling, and anticipatory action. However, this optimisation imperative creates a fundamental paradox: as systems excel at achieving measurable objectives, they may erode the collective intelligence and adaptive capacity of our societies. Recognising this tension, the field of Human-Centred AI (HCAI) has emerged to ensure AI aligns with human values. However, research on HCAI often focuses on idealised interactions, neglecting the pressure, moral dilemmas, and social dynamics typical of today’s complex problems.\u003c/p\u003e\n\u003cp\u003eThis paper introduces and advocates for a paradigm shift towards \u003cem\u003eSensemaking AI\u003c/em\u003e: AI that supports collective meaning-making processes in evolving human-AI networks. This novel perspective recognises that algorithmic and AI systems actively participate in the social processes through which humans interpret information, coordinate responses, and adapt their values. Grounded in sensemaking and decision theory and informed by a scoping review of the HCAI literature, this paper identifies three connected research areas: (i) sensemaking-aware automation that preserves interpretive flexibility; (ii) collective agency for network-level control; and (iii) value-aware sensemaking that supports collective meaning-making. These principles form the basis for Sensemaking AI as a design and research agenda that prioritises collective meaning-making and democratic deliberation in networks.\u003c/p\u003e","manuscriptTitle":"Sensemaking AI: Introducing a Research and Design Agenda for Human–AI Networks","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-11 12:39:34","doi":"10.21203/rs.3.rs-7463619/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-23T15:52:09+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-21T16:03:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-14T16:07:55+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-02T09:05:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"188722241784078270883280135165080560555","date":"2025-09-22T15:11:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"58552664277813586688570793386059072884","date":"2025-09-17T12:57:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"309493492017869813273991511802453390004","date":"2025-09-11T07:23:06+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-04T13:24:56+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-02T09:36:06+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-01T06:33:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"EPJ Data Science","date":"2025-08-26T13:45:59+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"epj-data-science","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"epds","sideBox":"Learn more about [EPJ Data Science](https://epjdatascience.springeropen.com/)","snPcode":"13688","submissionUrl":"https://submission.springernature.com/new-submission/13688/3","title":"EPJ Data Science","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4b507d4f-c840-4be1-bfec-81bafe66d5f6","owner":[],"postedDate":"September 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-23T16:00:36+00:00","versionOfRecord":{"articleIdentity":"rs-7463619","link":"https://doi.org/10.1140/epjds/s13688-026-00634-5","journal":{"identity":"epj-data-science","isVorOnly":false,"title":"EPJ Data Science"},"publishedOn":"2026-03-19 15:58:04","publishedOnDateReadable":"March 19th, 2026"},"versionCreatedAt":"2025-09-11 12:39:34","video":"","vorDoi":"10.1140/epjds/s13688-026-00634-5","vorDoiUrl":"https://doi.org/10.1140/epjds/s13688-026-00634-5","workflowStages":[]},"version":"v1","identity":"rs-7463619","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7463619","identity":"rs-7463619","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}