ChatGPT as a Metacognitive Catalyst in Personalized Learning Ecosystems

ChatGPT as a Metacognitive Catalyst in Personalized Learning Ecosystems | 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 ChatGPT as a Metacognitive Catalyst in Personalized Learning Ecosystems Eugene Kwasi Gyekye This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7339599/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This qualitative study explores the role of ChatGPT in shaping students’ metacognitive development and self-regulated learning within equity-driven, personalized learning environments. It draws on data from 24 semi-structured interviews, 300 hours of classroom observation, and analysis of student work collected from a yearlong youth apprenticeship program based in the northeastern United States. The program served culturally and socioeconomically diverse high school students and combined project-based learning with blended synchronous and asynchronous online instruction designed to promote career readiness and digital fluency. Grounded in Social Constructivism and Self-Regulated Learning (SRL) theory, one of the study’s key outcomes is the development of AI-Mediated Metacognitive Development (AIMD), a new theoretical framework that conceptualizes artificial intelligence as a dynamic cognitive mediator that can either support or hinder metacognitive growth. The findings reveal that ChatGPT fosters reflection, strategic thinking, and adaptive reasoning when implemented through structured scaffolding and peer collaboration. However, in the absence of intentional guidance, students often relied on ChatGPT in ways that led to cognitive offloading and superficial engagement. The study emphasizes that the educational value of AI depends not solely on its technological capabilities but on how it is embedded within pedagogical, cultural, and social contexts. AIMD provides a critical framework for designing AI-integrated instruction that advances learner agency, equity, and reflective autonomy. Metacognition Self-Regulated Learning ChatGPT Personalized Learning AI-Mediated Metacognitive Development (AIMD) Introduction The integration of artificial intelligence (AI) tools like ChatGPT into personalized learning ecosystems is transforming contemporary education (Asy'ari & Sharov, 2024), offering adaptive pathways, real-time feedback, and scalable support for diverse student populations (Rizvi et al., 2025 ). These advancements promise greater accessibility and tailored learning experiences (Yamijala et al., 2024 ; Mahmoud & Sørensen, 2024 ). However, this shift also introduces significant pedagogical, ethical, and infrastructural challenges (Komissarov, 2025 ; Gesser-Edelsburg et al., 2024 ). Kettler and Taliaferro ( 2022 ) note that while AI has the potential to support personalized instruction, its capacity to respond to the nuanced and evolving needs of individual learners is still in question. A particularly underexplored dimension of this technological integration is the impact of AI tools like ChatGPT on students’ metacognitive and self-regulated learning (SRL) capacities (Dahri et al., 2024 ). Metacognition, which includes the ability to plan, monitor, and evaluate one’s learning strategies, is a critical element of intellectual autonomy and lifelong learning (Marantika, 2021 ; Ushioda, 2014 ). Xie ( 2023 ) intimates that as education increasingly incorporates AI tools, understanding how these tools influence learners’ metacognitive development is essential (see Yang & Xia, 2023 ; Atonte, 202). This study explores how educators integrate ChatGPT into personalized learning environments and how its use shapes students’ development of metacognitive awareness and self-regulation. The central research question guiding this inquiry is: How does the use of ChatGPT in personalized learning environments influence students' metacognitive development and self-regulated learning? Literature Review Personalized learning aims to tailor educational experiences to each learner’s unique needs, strengths, and preferences. Bernacki et al. (2021) emphasize that personalization involves more than differentiated instruction; it requires a deep understanding of individual learner profiles. These profiles are not solely academic but also include learners’ metacognitive capacities, such as how they monitor, evaluate, and regulate their own understanding (Karlen, 2016 ). As such, effective personalized learning must attend not only to content delivery but also to the learner’s ability to reflect on and control their learning processes (Bernacki et al., 2021). AI tools like ChatGPT are increasingly used to fulfill this promise by offering real-time support, adjusting content difficulty, and enabling scalable one-on-one engagement (Sharma, 2024 ). These features have the potential to scaffold students' metacognitive development by prompting self-reflection, tracking progress, and offering iterative feedback (Shekh-Abed, 2024 ). Venter et al. ( 2024 ) argue that ChatGPT’s responsiveness and accessibility make it particularly valuable in resource-limited settings where human instructional time is constrained. Tabib and Alrabeei (2024) note that in such environments, ChatGPT may act as a metacognitive nudge, helping students plan study strategies, seek clarification, and test their understanding through iterative questioning. Despite these advantages, AI’s effectiveness in truly personalizing instruction and fostering metacognitive growth is debated (Levin et al., 2025; Hutson & Plate, 2023 ). Mergen et al. ( 2025 ) point out that ChatGPT and similar AI systems are limited in their ability to interpret emotional cues, cultural nuances, and cognitive diversity. These limitations may result in the marginalization of learners whose metacognitive strategies do not align with the dominant models represented in AI training data (Li & Brooks, 2024 ). This not only affects content relevance but also impairs students’ ability to develop self-awareness and adaptive learning strategies. Kettler and Taliaferro ( 2022 ) add that while AI may efficiently deliver instructional content, it lacks the contextual awareness and relational dynamics that human educators bring to the learning process. According to Eunkyung ( 2024 ), these relational elements are essential for developing metacognitive habits such as modeling reflective practice, encouraging intellectual risk-taking, and providing feedback that prompts self-correction. Without the ability to sense learners’ confusion, hesitation, or confidence levels, ChatGPT may inadvertently flatten the reflective space needed for metacognitive development (Roozenbeek & van der Linden, 2024 ). This disconnection between content delivery and authentic personalization challenges the assumption that AI inherently enhances individualized learning. Instead, it raises concerns that AI-based platforms may present a mechanized form of education that mimics personalization without addressing learners’ holistic development (Rane et al., 2023 ). Crucially, this includes the development of self-regulation and reflective awareness, which are foundational to metacognitive learning (Merkebu et al., 2023 ). According to Chung ( 2024 ), AI models, including ChatGPT, are trained on extensive datasets that reflect societal biases, which can manifest in skewed or culturally insensitive responses (see Panarese et al., 2025 ). Wang ( 2025 ) highlights how such biases can undermine efforts toward educational equity by reinforcing dominant cultural narratives and excluding marginalized perspectives. From a metacognitive perspective, Fan et al. ( 2025 ) opine that exposure to biased content can distort students’ reflective judgments, misguide their information evaluation processes, and potentially erode trust in digital learning tools. These issues are especially problematic in personalized learning contexts where inclusivity is assumed (Davoodi, 2024 ). Bhavana et al. ( 2024 ) note that when AI systems deliver recommendations or feedback shaped by biased data, students from underrepresented backgrounds may receive inferior guidance or feel alienated from the learning process. In the view of Lee et al. ( 2025 ), such experiences can impair metacognitive efficacy, especially if learners begin to question the validity of feedback, the reliability of their strategies, or the fairness of the educational environment. This contradiction challenges the ethical foundation of AI-enhanced personalization (Hussain, 2025 ), making it critical to examine not just whether students receive differentiated instruction but whether that instruction supports reflective, self-directed learning for all. According to Regan and Jesse ( 2018 ), beyond bias, ethical concerns extend to data privacy, student dependence, and the shifting role of educators. ChatGPT’s functionality relies on the collection and processing of user inputs, raising concerns about how personal data is stored, shared, and protected (Wu et al., 2024 ). From a metacognitive standpoint, lack of transparency about how AI models use learner data can inhibit trust and hinder students’ willingness to engage in open self-reflection and questioning (Levin et al., 2025). Metacognition requires learners to be honest and vulnerable about what they know and do not know (LaVaque-Manty et al., 2023 ). If they fear surveillance or data misuse, their ability to engage in metacognitive dialogue with AI tools may be compromised (Walker et al., 2025 ). Dependence on AI-generated assistance may also lead to negative learning behaviors. Gesser-Edelsburg et al. ( 2024 ) argue that students may over-rely on ChatGPT’s convenience, bypassing the reflective and effortful processes required for deep understanding. In my experience as an urban educator, I have seen how, when learners habitually defer to AI for explanations, their critical thinking and problem-solving capacities may deteriorate. This risks diminishing metacognitive monitoring and control, as students are less likely to pause, evaluate alternative strategies, or question the validity of a given response. Viola ( 2023 ) notes that such overdependence challenges the goal of cultivating autonomous, self-regulated learners who can transfer knowledge and skills across domains. While AI can supplement instruction, it cannot replace the mentorship, empathy, and human-centered adaptability that teachers provide (El Karafli, 2025 ). Kajiwara et al. ( 2023 ) note that educators play a vital role in modeling metacognitive behaviors such as articulating their thinking, demonstrating how to revise misconceptions, and encouraging learners to reflect on their learning processes. The challenge lies in finding a balanced model where AI enhances learning without diminishing the relational aspects of education that support emotional, social, and cognitive development. According to Vo et al. ( 2025 ), the risk of diminished metacognitive engagement, underscores the importance of pedagogical scaffolding. Educators must guide students in using ChatGPT to support rather than replace metacognitive effort (Tabib & Alrabeei, 2024). Melisa et al. ( 2025 ) note that instructional strategies might include requiring students to compare their reasoning with the AI’s output, explain their choices, or revise their thinking after receiving feedback. Such practices can help ensure that ChatGPT serves as a metacognitive catalyst rather than a cognitive crutch. Research Gap and Theoretical Contribution While AI tools like ChatGPT are increasingly present in education, there is limited empirical research examining their impact on students’ metacognitive development and self-regulated learning. Most existing studies focus on outcomes such as performance, user satisfaction, or system efficiency. These areas, though important, largely overlook the internal cognitive and affective processes that shape learners' ability to think critically, plan strategically, and reflect on their learning. Without understanding how AI affects these higher-order skills, educators risk deploying technology in ways that compromise long-term educational goals. To address this gap, this study investigates the research question: How does the use of ChatGPT in personalized learning environments influence students’ metacognitive development and self-regulated learning? Rather than viewing students as passive recipients of AI-generated content, this study emphasizes their role as active learners capable of self-monitoring, reflection, and adaptation. In response to the gap, this research introduces a new theoretical perspective, AI-Mediated Metacognitive Development (AIMD), that builds on Social Constructivism and Self-Regulated Learning Theory. AIMD conceptualizes AI not as a neutral tool, but as a dynamic influence that can either support or hinder students’ metacognitive growth, depending on how it is integrated into learning environments. Theoretical Framework The integration of artificial intelligence (AI) into personalized learning environments calls for a robust theoretical foundation to evaluate its impact on students' cognitive and metacognitive development. This study draws upon two foundational educational theories: Social Constructivism and Self-Regulated Learning (SRL) Theory to examine how ChatGPT supports or hinders the development of learner autonomy, critical thinking, and self-awareness. Social Constructivism: AI as a Mediator of Interaction Grounded in Vygotsky’s (1978) Social Constructivism, learning is understood as a socially mediated process where knowledge is co-constructed through interaction. A central concept in this theory is the Zone of Proximal Development (ZPD), which suggests learners can achieve higher levels of understanding with appropriate scaffolding from more knowledgeable others. In AI-integrated classrooms, tools like ChatGPT can function as digital scaffolds, offering context-specific feedback, explanations, and prompts aligned with learners’ needs. However, while ChatGPT may facilitate access to knowledge, it lacks the emotional intelligence and cultural sensitivity inherent in human interaction (Rawat et al., 2024 ). Because AI cannot engage in reciprocal dialogue or perceive socio-emotional cues, overreliance may reduce opportunities for rich, collaborative learning experiences. Vygotsky emphasized that authentic learning occurs within socially meaningful contexts - an element that AI, in its current form, cannot fully replicate (Vygotsky, 1978; Mercer & Howe, 2012 ). Self-Regulated Learning: AI and Metacognitive Support Complementing constructivist theory, SRL Theory (Zimmerman, 2002 ) provides a lens for understanding how learners manage their own learning processes. SRL encompasses three cyclical phases: forethought (goal-setting and planning), performance (strategy use and monitoring), and self-reflection (evaluating and adjusting approaches). Wang et al. ( 2025 ) posit that ChatGPT can assist across these phases by prompting goal clarification, offering strategy suggestions, and providing feedback that supports real-time reflection and metacognitive monitoring. Despite these affordances, AI may unintentionally undermine SRL by encouraging cognitive offloading ( see Kim et al., 2025 ), where students outsource thinking to the tool instead of developing independent reasoning. Unlike human mentors who can gradually reduce support to foster autonomy, AI lacks the capacity to calibrate its assistance based on developmental cues or encourage productive struggle (Winne, 2018; Roll & Winne, 2015 ). When students become passive recipients of AI outputs, their capacity for self-regulation and critical thinking may diminish. AI-Mediated Metacognitive Development (AIMD): A Theoretical Construct I construe the AI-Mediated Metacognitive Development (AIMD) as the process through which artificial intelligence, particularly generative AI tools like ChatGPT, mediates, supports, or impedes the development of learners’ metacognitive capacities within personalized learning environments. Rather than conceptualizing AI as a passive instructional aid, I position AI as an active agent that significantly shapes learners' abilities to monitor, evaluate, and regulate their thinking. I ground AIMD in two complementary theoretical traditions: Social Constructivism, which emphasizes the role of mediated social and cultural tools in cognitive development, and Self-Regulated Learning (SRL) Theory, which focuses on learners’ capacity for planning, monitoring, and evaluating their own learning processes. Together, these perspectives provide a robust foundation for understanding how AI can enhance or hinder metacognitive development depending on the conditions of its use. From a Social Constructivist perspective, learning is inherently mediated by tools and artifacts situated within social contexts. AIMD extends this concept to artificial intelligence, treating AI as a digital cognitive artifact that co-constructs knowledge with learners. Through interactive dialogue, feedback loops, and modeled reasoning, AI tools like ChatGPT engage with learners in ways that can influence their conceptual understanding and thought patterns. I argue that these AI systems do not merely transmit information; they actively shape the discourse through which learners explore and refine their thinking. In parallel, Self-Regulated Learning theory positions metacognition, defined as awareness and regulation of one’s thinking, as a foundational element of effective learning. AIMD builds on this theory by analyzing how AI affects each phase of self-regulation: from planning and goal-setting, to real-time monitoring, to post-task evaluation and reflection. AIMD identifies the mediating function of AI as central to its impact on metacognitive development. Thus, AI can mediate learning in both facilitative and inhibitive ways. When integrated intentionally into pedagogically sound learning environments, AI can play a facilitative role by prompting learners to reflect on their reasoning, offering iterative feedback, scaffolding goal setting, and encouraging strategic decision-making. For instance, students using ChatGPT can be guided to justify their responses, pose follow-up questions, or test alternative solutions. These interactions foster deeper metacognitive engagement and enhance learners’ awareness of how they approach and solve problems. However, when AI is used in an unstructured, convenience-driven manner (such as asking for direct answers without engaging with the process), it can inhibit metacognitive growth. In such cases, learners may bypass cognitive effort, develop dependency on AI-generated outputs, and engage in cognitive offloading, thereby reducing opportunities for reflective thought and strategic learning. To support sustained development, AIMD introduces the concept of dynamic scaffolding , which sees AI as a temporary support structure. Effective integration of AI into learning follows a developmental trajectory: support, fade, and internalize . Initially, AI offers structured prompts and guidance to help learners navigate new tasks or concepts. As learners build confidence and competence, the level of AI assistance should gradually fade. Eventually, learners are expected to internalize metacognitive strategies and apply them independently, without relying on the tool. This model aligns with cognitive apprenticeship frameworks, where learners move from guided participation to autonomous mastery, absorbing expert behaviors through repeated exposure and practice. Another core dimension of AIMD is learner-AI interaction as cognitive apprenticeship . In this view, AI tools serve as cognitive partners that model expert-like thinking. ChatGPT, for example, can demonstrate how to evaluate competing arguments, reason through ambiguity, or reflect on past errors. Through repeated interaction, learners absorb these metacognitive behaviors and begin to emulate them independently. Over time, the AI transitions from an active guide to a reflective sounding board, supporting the learner’s journey toward cognitive autonomy. AIMD also emphasizes the importance of metacognitive intentionality in instructional design. I argue that AI should not merely function as an answer generator. Rather, its use must be aligned with explicit metacognitive goals that encourage students to justify their reasoning, compare multiple solutions, identify knowledge gaps, and establish learning objectives. This intentional design promotes deeper cognitive engagement and transforms AI from a static information source into a dynamic thinking partner. The ultimate aim is not simply to help students complete tasks more efficiently, but to cultivate learners who are more reflective, strategic, and aware of their own learning processes. Importantly, AIMD acknowledges that the benefits of AI-mediated metacognitive development are not equally distributed. In view of this, I foreground ethical and equity considerations to cater for learners from underserved communities. Learners from underserved communities may face systemic barriers that hinder their ability to benefit fully from AI-enhanced learning environments. These include limited access to high-quality digital tools, insufficient AI literacy, and prior educational experiences that undermine academic confidence. As a result, some students may over-rely on AI for quick answers rather than engage it as a reflective tool. AIMD, therefore, demands equity-conscious pedagogy that provides targeted support, models effective AI use, and fosters reflective agency among all learners, regardless of background. Only through intentional and inclusive design can AI become a tool for cognitive empowerment rather than another mechanism of stratification. As a contribution to educational theory, this framework reframes the role of AI in learning from that of content delivery to that of cognitive mediation. It bridges human–machine interaction with established theories of learning by showing how AI influences metacognitive development in both constructive and constraining ways. The framework introduces a dialectical view of AI as both catalyst and constraint, depending on its use context. For practitioners, AIMD provides a set of guiding principles to design, implement, and evaluate AI-integrated instruction. It helps educators create learning tasks that provoke strategic thinking, monitor and assess learners’ metacognitive engagement with AI, and balance the use of technological support with opportunities for independent reflection. Research Methodology This qualitative study drew on constructivist and interpretive research traditions to examine how students and educators engaged with ChatGPT in a blended online learning environment. Guided by the frameworks of AI-Mediated Metacognitive Development (AIMD), Self-Regulated Learning (SRL), and Self-Regulated Learning (SRL) Theory, the study focused on how ChatGPT shaped metacognitive processes and equitable learning experiences among high school students from underserved communities. The research was conducted during the 2023–2024 academic year within a youth apprenticeship program facilitated by the Innovation Learning Collective (a pseudonym), a nonprofit based in the American Northeast. The program served a diverse cohort of 100 students: 40% Black, 30% Hispanic, 20% White, and 10% Asian, roughly half of whom qualified for free or reduced-price lunch. Students participated in synchronous workshops and asynchronous activities designed to build career readiness skills through project-based learning and responsible AI use. Using purposive sampling (Patton, 1990; Maxwell, 2022), participants were selected from three stakeholder groups: students, facilitators, and program managers. Five high school students were chosen to reflect variability in race, socioeconomic status, and digital fluency. Their perspectives provided insight into how ChatGPT supported identity, agency, and self-regulation. Five facilitators representing STEM, humanities, and career development disciplines contributed instructional perspectives on AI implementation. Two program managers offered institutional context regarding curriculum design and equity goals. Data collection included 12 semi-structured interviews with students, 9 with facilitators, and 3 with program managers. These were complemented by ongoing ethnographic interviews, 300 hours of field observations during synchronous sessions, and analysis of student-generated artifacts such as reflective essays and ChatGPT-influenced assignments. All data were transcribed and thematically coded to identify patterns in how ChatGPT was used, interpreted, and evaluated by participants. Cross-case comparisons and triangulation ensured credibility and analytic depth. This multi-dimensional approach enabled a rich examination of the central question: How does the use of ChatGPT in personalized learning environments influence students’ metacognitive development and self-regulated learning? Research Context This study was conducted during the 2023–2024 academic year within a youth apprenticeship program hosted by the Innovation Learning Collective (a pseudonym), a nonprofit in the American Northeast. Designed to build career readiness skills through personalized and equity-centered learning, the program served 100 high school students from culturally and socioeconomically diverse backgrounds: 40% Black, 30% Hispanic, 20% White, and 10% Asian, with about half qualifying for free or reduced-price lunch. Students were organized into five groups, each supported by two facilitators and overseen by program managers. Instruction followed a blended synchronous and asynchronous online learning model. During synchronous sessions, students engaged in guided discussions and interactive demonstrations of ChatGPT’s capabilities, with an emphasis on ethical use, critical analysis of AI outputs, and learning equity. In asynchronous settings, students used ChatGPT to support project-based assignments, reflective writing, and independent research, positioning the tool as a metacognitive aid in goal-setting, self-monitoring, and strategic thinking. Participants A purposive sampling strategy (Patton, 1990; Maxwell, 2022) was used to select participants from three stakeholder groups; students, facilitators, and program managers, ensuring diverse perspectives on ChatGPT’s role in personalized, AI-integrated learning. Student Participants Five students were selected to reflect variation in race, socioeconomic status, gender, and digital fluency. All were actively enrolled in the AI-assisted course and used ChatGPT across structured and exploratory learning tasks. Table 1 Demographic Characteristics of Student Participants Name Age Gender Ethnicity Socioeconic Status Digital Fluency Alex 18 Male African American Economically Disadvantaged Fluent Jordan 17 Male African American Economically Advantaged Semi-Fluent Taylor 17 Female White Economically Advantaged Fluent Morghan 18 Male Asian Economically Advantaged Fluent Casey 17 Female Latinx Economically Disadvantaged Semi-Fluent These students were the focal point for analyzing how ChatGPT shaped metacognitive engagement within a sociotechnical and culturally relevant pedagogical framework. Their voices illuminated how AI tools either fostered or hindered agency, identity affirmation, and learning autonomy. Facilitator Participants Five educators contributed instructional perspectives from STEM, humanities, and career development disciplines. Their experiences informed how ChatGPT was implemented pedagogically and how it supported or constrained culturally responsive and metacognitive learning. Table 2 Demographic Characteristics of Facilitator Participants Name Age Gender Ethnicity Class Taught Ms. Hailey 28 Female African American Workforce Readiness Mr. Goneday 24 Male African American STEM Ms. Goods 26 Female Latinx STEM Mr. Griffiths 29 Male White Humanities Mr. Andrew 25 Male Latinx Humanities Their inclusion enabled an examination of how facilitators mediated students’ metacognitive development through instructional choices, tool design, and equitable access Program Manager Participants Two program managers; Mr. Gobble and Ms. Green offered strategic insights into the design, goals, and equity considerations underpinning ChatGPT integration in the learning ecosystem. Table 2 Demographic Characteristics of Program Manager Participants Name Age Gender Ethnicity Designation Mr. Gobble 30 Male White Program Manager Ms. Green 28 Female White Program Manager Their perspectives were essential for understanding the institutional context in which ChatGPT was positioned as a tool for promoting student agency, critical reflection, and equitable AI use. Data Sources Data collection included 12 semi-structured interviews (see Appendix A) with students, 9 with facilitators, and 3 with program managers. These were augmented by ongoing ethnographic interviews, analysis of student-generated artifacts (e.g., reflective essays, ChatGPT-assisted work products), and approximately 300 hours of synchronous observations (see Appendix B) and generated field notes capturing real-time student interaction with AI. All data were transcribed and coded using thematic analysis to explore how ChatGPT functioned as a metacognitive catalyst. Triangulation and multi-stakeholder perspectives supported a robust understanding of how students navigated personalized learning with AI under varying structural and social conditions. Researcher Positionality As the sole author of this study, I identify as an African American educator-researcher with a longstanding commitment to culturally responsive and equity-centered pedagogies. This research was conducted within a youth apprenticeship program where I previously served as an instructor, and while I did not teach the student participants during the study period, my ongoing relationship with the site and shared cultural heritage with many of the learners fostered an atmosphere of trust, openness, and mutual respect. My proximity to the community allowed for deeper insight into the sociocultural contexts that shaped students’ interactions with AI tools like ChatGPT, particularly in terms of access, agency, and reflective learning. Recognizing that my positionality could influence data interpretation, I employed reflexive practices throughout the research process, including memoing, peer debriefing, and journaling, to surface and bracket my assumptions. I was particularly attuned to the ways that my advocacy for metacognitive learning and digital equity could shape the thematic emphasis of the analysis. To mitigate this, I foregrounded participant voice, allowed for disconfirming evidence, and remained anchored in the study’s theoretical frameworks of AI-Mediated Metacognitive Development (AIMD), Social Constructivism, and Self-Regulated Learning. Transparency with participants about the research goals further ensured an ethical, dialogic approach, reinforcing a shared commitment to understanding how AI can serve not merely as a technological tool, but as a catalyst for deeper, more equitable learning. Control Measures in Data Collection To ensure theoretical coherence and methodological rigor, this study employed carefully designed control measures across all data collection activities. These measures were grounded in the study’s theoretical framework, Social Constructivism, Self-Regulated Learning (SRL), and AI-Mediated Metacognitive Development (AIMD) to examine how ChatGPT mediates students’ metacognitive development within personalized learning environments. Each data instrument was intentionally designed to assess how learners interacted with AI as a cognitive partner, a source of scaffolding, and a potential site of cognitive offloading. Interviews Semi-structured interview protocols were developed to explore how students and educators perceived ChatGPT’s role in metacognitive processes such as planning, monitoring, and reflecting on learning. Drawing from SRL theory (Zimmerman, 2002 ), interview prompts were designed to trace each phase of the self-regulatory cycle. Questions focused on learners’ strategies before, during, and after AI use, and invited reflection on goal-setting, decision-making, and adaptation. From a Social Constructivist lens, interview questions also probed how learners engaged with ChatGPT as a digital scaffold; what kinds of dialogue they co-constructed with the tool, and whether these interactions reflected internalized learning. The AIMD framework informed questions regarding dependency, autonomy, and strategic fading of AI support, to assess whether learners were progressing toward internalization or becoming reliant on the tool. All interview protocols were piloted with high school students and educators to ensure alignment with the theoretical constructs, clarity of language, and cultural responsiveness. Interviews were audio-recorded, transcribed verbatim, and verified through member-checking to maintain fidelity to participants’ voices and cognitive frames. Observations Observational data were gathered across 300 hours of classroom instruction using a structured protocol derived from the AIMD framework. The observation form captured indicators of students’ metacognitive engagement, strategic use of AI, collaborative meaning-making, and awareness of bias or inequity in AI outputs. Social Constructivism informed the documentation of peer-to-peer AI interaction and the social contexts in which AI use occurred. Observers noted whether learners engaged in co-construction of knowledge through discussion, critique, or collaborative prompting with ChatGPT. The SRL framework guided attention to strategic behavior, such as whether students refined prompts, monitored outputs, and evaluated ChatGPT’s feedback in real time. AIMD shaped the observational categories to reflect whether AI was functioning as a scaffold, a mentor, or a crutch. The protocol enabled researchers to document both productive AI-mediated metacognitive engagement and signs of cognitive offloading or passive consumption. T his observation protocol ensured consistent data collection across contexts, reduced observer bias, and operationalized theoretical constructs in concrete classroom behavior. Expert Validation To ensure analytical validity, two colleagues of mine who are academics in qualitative research and metacognitive learning reviewed a subset of coded transcripts, observation notes, and emergent themes. Their task was to examine whether the data interpretations were consistent with the theoretical framework, particularly the core claims of the AIMD model. Their review confirmed that the data coding procedures and thematic interpretations faithfully represented Social Constructivist principles of mediated learning, SRL concepts of metacognitive strategy use, and AIMD’s emphasis on AI as both scaffold and constraint. Expert validation further strengthened the credibility of the findings by confirming that themes such as strategic fading, critical reflection, and AI-assisted autonomy were not only emergent from the data but theoretically coherent. Data Analysis This study’s data analysis was conducted through a qualitative, inductive lens guided by the AI-Mediated Metacognitive Development (AIMD) framework, which integrates Social Constructivism and Self-Regulated Learning (SRL) theories. To capture the nuanced ways learners engaged with AI tools like ChatGPT in personalized learning environments, the analysis followed a three-phase coding process: open coding, axial coding, and selective coding. Each phase progressively distilled the data into more abstract and conceptual themes. Open Coding: Raw Data Exploration The initial phase employed open coding to identify discrete actions, perceptions, and reflections expressed by students regarding their interactions with ChatGPT. During this phase, verbatim quotes were extracted and interpreted to uncover learners’ immediate responses to AI’s affordances and constraints. Table 1 below exemplifies the open codes, paired with representative quotes and interpretive comments that revealed emergent metacognitive and behavioral patterns. Table 1 Data Chart Excerpt for Open Coding with Illustrative Quotes Open Code Illustrative Quote Interpretation Distrust of AI “I don’t trust everything it gives me, so I check it with other sources.” – Morgan Indicates evaluative thinking and digital skepticism—student questions AI authority. Rewriting for Tone “When I asked ChatGPT to help with an email, it sounded too chill. I had to change it so it looked right.” - Casey Student demonstrates critical literacy and revision, not accepting AI at face value. Recognizing AI Limits “I’ve seen ChatGPT mess up when it tries to do too much.” – Jordan Student shows awareness of AI overreach; reflects analytical monitoring. Independent Look-Up “Sometimes it just doesn’t make sense. I look stuff up on my own after.” - Alex Reflects independent learning and cross-checking strategies. Copy-Pasting AI Outputs “If I’m stuck, I just type it in and copy what looks good.” – Jordan Surface-level use of AI; cognitive offloading without self-evaluation. Passive Use “I usually just go with the first thing it gives me, unless it sounds really weird.” – Casey Minimal engagement or reflection; trust in AI defaults. Curiosity About AI “It looks cool when they use it right. Maybe I’d try if I knew more about how it works.” – Alex Student expresses interest but lacks access or understanding—entry point to AI literacy. Desire for Explanation “I wish it told them why it said that.” – Mr. Griffiths Craving for AI transparency; signals desire to engage more deeply. Educator Scaffolding “We had to explain what we changed in the ChatGPT draft. That part made me think harder.” – Ms. Hailey Assignment promotes metacognitive reflection and ownership of learning. Peer Critique Activities “They debated which AI answers were more persuasive and why.” - Mr. Gobble AI used as a social object for dialogue, promoting co-regulation and reflective evaluation. These initial codes revealed a complex spectrum of learner interactions with AI: from critical evaluation and revision to passive acceptance and over-reliance. Several learners expressed curiosity and a desire for transparency in AI’s reasoning, which suggested nascent forms of metacognitive intentionality and agency. Axial Coding: Thematic Grouping Building on the open codes, axial coding grouped related concepts into broader thematic categories that described how students engaged cognitively and socially with ChatGPT. This phase connected discrete behaviors to conceptual patterns of metacognitive engagement or disengagement and identified the role of scaffolding and co-design in AI use. Table 2 presents these thematic categories alongside representative quotes and interpretive commentary. Table 2 Data Chart Excerpt for Axial Coding with Thematic Categories Thematic Category Illustrative Quote Interpretation Critical AI Engagement “I don’t trust everything it gives me, so I check it with other sources.” - Morgan Students display active metacognition: evaluating, editing, and cross-referencing AI output. “It sounded too chill. I had to change it.” - Alex Reflects critical literacy and deliberate revision of AI-generated content. “I look stuff up on my own after.” - Taylor Indicates independent verification and strategic monitoring of AI. Cognitive Offloading & Over-Reliance “I just type it in and copy what looks good.” - Casey Shows superficial engagement and reliance on AI as a shortcut, limiting deeper cognitive processing. “I usually just go with the first thing it gives me.” Morghan Reveals passive trust and minimal metacognitive effort. Structured AI Scaffolding “They had to explain what they changed.” - Ms. Goods Educator scaffolding promotes reflection and metacognitive awareness through active revision. “We debated which AI answers were better.” - Mr. Andrew Teacher initiated peer discourse activates co-regulation and critical evaluation. Co-Designing Digital Learning “I wish it told me why it said that.” - Casey Expresses learner desire for transparency and control, signaling emerging metacognitive agency. “Maybe I’d try if I knew more about how it works.” - Alex Highlights the need for accessible AI literacy and inclusive design to foster learner engagement. This phase illustrated how metacognitive outcomes depended heavily on the context of AI use: when learners were scaffolded and engaged socially, they demonstrated critical evaluation and strategic thinking. Conversely, unsupported use often led to cognitive offloading, reducing opportunities for metacognitive growth. The aspiration for transparency and deeper understanding suggested fertile ground for designing AI systems that promoted learner autonomy and reflection. Selective Coding: Core Themes, Illustrative Code and Interpretation In the final phase, selective coding synthesized the axial themes into core theoretical constructs aligned with the AIMD framework. This phase distilled the dualistic influence of AI on metacognitive development into facilitative and inhibitive roles, while highlighting the importance of learner agency in AI integration. Table 3 summarizes these core themes with supporting quotes and interpretations. Table 3 Data Chart Excerpt for Selective Coding Thematic Category Illustrative Quote Interpretation Facilitative Role of AI “I don’t trust everything it gives me…” - Morgan ChatGPT supports metacognitive growth when paired with critical thinking and scaffolding. “I look stuff up on my own.” - Taylor Learners develop independent verification strategies alongside AI use. “We debated which AI answers were more persuasive.” - Casey Structured peer dialogue fosters analytical evaluation and reflective reasoning. Inhibitive Role of AI “I just copy what looks good.” - Alex AI can encourage passive learning and cognitive offloading without metacognitive reflection. “I just go with the first thing.” - Jordan Lack of scaffolding leads to diminished learner autonomy and reflective capacity. Student Agency in AI Integration “I wish it told me why it said that.” - Casey Learners desire AI transparency and control, crucial for metacognitive engagement and empowerment. “Maybe I’d try if I knew more…” - Alex Highlights importance of equitable and inclusive AI literacy to foster learner agency and curiosity. The selective coding phase crystallized the dynamic tension inherent in AI-mediated learning environments. AI functioned as a double-edged sword: it facilitated metacognitive development when thoughtfully integrated with pedagogical scaffolds, but it also inhibited reflective learning when used passively. Importantly, learners’ calls for transparency and understanding emphasized the need for AI tools designed to empower rather than replace metacognitive processes. Findings This study examined how ChatGPT shaped high school students’ metacognitive development within a culturally diverse, personalized learning environment. Drawing on the AI-Mediated Metacognitive Development (AIMD) framework, findings illuminated three overarching themes: (1) AI as a scaffold for metacognitive growth, (2) risks of cognitive offloading, and (3) student agency as a mediating force. These themes were grounded in Social Constructivist theory and Self-Regulated Learning (SRL) principles, revealing how AI’s influence depended on context, instructional design, and learners’ cultural and cognitive positioning. In environments where ChatGPT was used with explicit pedagogical scaffolding, students demonstrated increased metacognitive engagement. Learners reflected on their thought processes, questioned AI outputs, and revised their work with a growing awareness of their cognitive strategies. Taylor, described how ChatGPT prompted her to reconsider her reasoning: “I was working on an argument for my history project. ChatGPT gave me one perspective, but then I asked, ‘what’s the counterargument?’ and that made me rethink everything. I hadn’t even considered that angle until it brought it up.” Taylor’s interaction exemplified strategic monitoring, a core SRL process, where learners engage in evaluating alternative approaches. Her use of AI moved beyond passive acceptance toward self-questioning and adaptive thinking which are hallmarks of metacognitive maturity. Similarly, Casey, emphasized how facilitator-designed assignments elevated her reflection: “We had to explain what we changed in the ChatGPT draft. That part made me think harder. Like, why did I reword it? What didn’t feel right?” Here, the facilitator’s prompt functioned as a “metacognitive nudge,” consistent with AIMD’s dynamic scaffolding concept, where AI support was framed as temporary and strategically faded to promote autonomy. Facilitators played a central role in activating this scaffolding. Mr. Andrew, a Latinx humanities instructor, reported: “We structured reflection logs where students had to describe how they edited AI content. It wasn’t about right or wrong… it was about their thinking. That’s when you saw growth.” Such strategies not only modeled reflective practice but also shifted learners’ attention from product to process, promoting metacognitive internalization. This mirrored Vygotsky’s (1978) ZPD theory, where knowledge is co-constructed through guided participation with more capable others, including AI as a cultural artifact. In contrast, when ChatGPT was used without structured reflection or social mediation, many students exhibited cognitive offloading, bypassing planning and evaluation in favor of convenience. This uncritical use often occurred during independent or asynchronous tasks. Zoe, a Black female student with high digital fluency, admitted: “If I’m stuck, I just type it in and copy what looks good. I don’t always read it carefully.” Her comment reflected surface-level engagement. Rather than prompting metacognitive strategy use, ChatGPT functioned here as a cognitive shortcut, undermining the learner’s self-monitoring capacity. Darius, an African American male student, echoed this sentiment: “I usually just go with the first thing it gives me, unless it sounds really weird.” These patterns aligned with Kim et al.’s ( 2025 ) warning that AI tools can promote “passive alignment,” where learners abdicate responsibility for evaluating outputs, thus weakening metacognitive control. Observational data confirmed these findings. In one STEM class, students pasted ChatGPT-generated code into a lab report without modification or explanation. The facilitator, Ms. Goods, noted: “They didn’t analyze or troubleshoot the code-they assumed it was correct because it came from ChatGPT.” This revealed a critical shift: instead of AI acting as a thinking partner, it became an authoritative source, reinforcing dependence and reducing productive struggle; an essential component of SRL. From an AIMD perspective, such use disrupted the strategic fading process. Rather than scaffolding learners toward independence, AI became an intellectual crutch. Without teacher mediation or peer dialogue, learners failed to activate reflection or revision, jeopardizing long-term cognitive autonomy. Amidst these divergent usage patterns, students expressed a clear desire for agency in navigating their AI interactions. Casey voiced frustration with AI’s opacity: “I wish it told me why it said that. Like, how did it come up with that answer?” Her comment underscored the need for AI transparency; what AIMD identifies as a condition for metacognitive intentionality. When learners understand the logic behind AI responses, they are more likely to engage in strategic evaluation and adaptive learning. Alex, an African American male with fluent digital literacy, reflected: “It looks cool when they use it right. Maybe I’d try it more if I knew how it works under the hood.” Alex’s curiosity pointed to a broader theme: digital fluency gaps constrained some students’ ability to engage metacognitively. Without access to AI literacy instruction, students from underserved communities were at risk of using ChatGPT as consumers rather than co-designers of knowledge as echoed in Bhavana et al. ( 2024 ). Program manager Ms. Green articulated this challenge: “We had students with so much potential, but they didn’t know how to push ChatGPT to be more than a search engine. Equity in this work means teaching them to interrogate, not just receive.” This highlighted the critical interplay between student agency and institutional scaffolding. Even students who lacked initial fluency demonstrated reflective intent when given the tools. As Jordan, a semi-fluent African American student, put it: “At first I didn’t get it. But once we started comparing what ChatGPT said to what we found on our own, I started seeing where it was helpful and where it was off.” These instances of strategic evaluation reflected the beginnings of metacognitive co-regulation, here learners engaged AI as a dynamic thinking partner, consistent with AIMD’s model of internalization through scaffolded apprenticeship. Peer interaction emerged as a powerful amplifier of metacognitive growth. Collaborative critique and discussion transformed ChatGPT from a static tool into a shared object of inquiry. Morghan, described this dynamic saying: “Sometimes I think ChatGPT gave me a good response, but when I shared it with my group, someone would ask, ‘Does that even make sense?’ Then I’d go back and rethink it.” This cycle mirrored Vygotsky’s (1978) emphasis on social mediation, where higher-order thinking emerges through interaction. By externalizing their ideas and receiving critique, learners engaged in co-regulation, strengthening their metacognitive strategies. Educators leveraged this through practices like “AI audit circles,” where students reviewed each other’s ChatGPT-assisted work. Mr. Griffiths, a white humanities instructor, explained: “When students debate which AI answer is stronger, they’re not just evaluating content… they’re thinking about thinking.” These activities positioned ChatGPT as a dialogic stimulus rather than a solution engine. Students were not passive recipients but active co-constructors of knowledge, using AI as a catalyst for metacognitive discourse. The data revealed a dialectical tension within AIMD: ChatGPT’s metacognitive impact was neither inherently positive nor negative but contingent upon how it was used. This duality echoed Rane et al.’s ( 2023 ) critique that AI often “mimics” personalization without promoting deeper learning unless integrated intentionally. In structured settings, ChatGPT modeled critical thinking, offered scaffolded prompts, and supported metacognitive dialogue. In unstructured contexts, it invited dependency, discouraged effort, and limited reflection. This mirrors SRL literature that emphasizes the importance of feedback loops, goal clarity, and iterative evaluation for metacognitive development (Zimmerman, 2002 ). Importantly, this finding challenges deterministic narratives of AI as either savior or threat. Instead, it positions educators, and their pedagogical choices, as the true mediators of metacognitive outcomes. Across multiple classrooms, a new pedagogical pattern emerged: metacognitive co-creation, where students and ChatGPT engaged in iterative dialogue, challenging assumptions and refining strategies. Rather than reflecting alone, learners used ChatGPT to simulate Socratic questioning, test hypotheses, and explore counterarguments. Jordan recounted:“I asked ChatGPT for a take on utilitarianism. Then I tried to argue against it. It was weird but helpful. It made me think deeper about my own view.” This reciprocal engagement aligned with AIMD’s model of AI as cognitive apprentice: modeling reasoning, provoking revision, and eventually fading into the background as learners internalized reflective routines. The co-creation of metacognitive space also extended to ethics. Students like Casey raised concerns about biased outputs, asking facilitators why ChatGPT seemed to “prefer certain kinds of examples.” These critiques not only reflected epistemic awareness but also digital criticality which is a metacognitive disposition essential in AI-mediated environments. The findings of this study reveal that ChatGPT plays a dual and dynamic role in students’ metacognitive development, acting as both a scaffold that fosters strategic thinking and a crutch that may hinder reflective learning, depending on the context of its use. When embedded within structured, socially mediated, and equity-conscious learning environments, ChatGPT catalyzed metacognitive growth by prompting self-questioning, enabling peer dialogue, and supporting iterative evaluation. Conversely, in the absence of instructional scaffolding and digital literacy supports, students often engaged in cognitive offloading, demonstrating passive reliance on AI outputs. Central to these outcomes was the role of educators in designing intentional interactions that positioned AI not as a source of answers but as a thinking partner, and the agency of learners in leveraging AI to regulate, critique, and expand their own understanding. Ultimately, the study underscores that the impact of AI on metacognition is not intrinsic to the technology itself, but emerges from the pedagogical, social, and cultural ecologies in which it is situated. Discussion This study demonstrated that ChatGPT influenced students’ metacognitive development in personalized learning environments in both facilitative and inhibitive ways. Using the AI-Mediated Metacognitive Development (AIMD) framework, the findings showed that ChatGPT supported metacognitive processes such as planning, monitoring, and evaluating, when paired with intentional instructional design and structured reflection. In these contexts, students used the tool to critique ideas, explore alternatives, and refine their thinking, aligning with Self-Regulated Learning theory. Peer dialogue and collaborative critique further enhanced this engagement, transforming ChatGPT into a socially shared scaffold rather than an isolated digital tool. This reflected Social Constructivist principles, where metacognitive growth was co-constructed through interaction. When educators embedded prompts for critique and required justification of AI use, students demonstrated deeper cognitive control and self-awareness. However, in unstructured or unsupported contexts, students often defaulted to accepting ChatGPT’s responses without reflection, revealing patterns of cognitive offloading. This inhibited metacognitive engagement, especially among students with limited digital fluency or prior experience in reflective learning. These patterns highlighted the risk of overdependence on AI, particularly when it was introduced without sufficient scaffolding or clarity of purpose. The findings also revealed inequities in metacognitive outcomes. Students from historically underserved backgrounds, despite showing interest and curiosity, faced greater challenges in using ChatGPT strategically. Gaps in access, AI literacy, and instructional support limited their opportunities to engage the tool reflectively, underscoring how inequitable integration could reinforce rather than disrupt educational disparities. Importantly, the study identified the emergence of metacognitive co-creation - a process in which students and AI collaboratively shaped learning through iterative dialogue. When students actively revised, questioned, and challenged ChatGPT outputs, they demonstrated distributed metacognitive regulation. However, this potential only emerged under conditions that supported inquiry, transparency, and learner agency. Overall, the findings challenged the notion that AI inherently promotes personalization. Without deliberate guidance, AI use often remained superficial and performative. ChatGPT’s value lay not in generating answers but in prompting deeper cognitive effort, when educators positioned it as a tool for thinking, not a replacement for it. The study concluded that metacognitive development in AI-mediated environments depended less on technological sophistication and more on pedagogical intentionality and equity-driven design. Implications for Educators and Students The findings highlight that the educational value of AI tools like ChatGPT depends entirely on how they are introduced, guided, and contextualized. For educators, this underscores the need to move beyond using AI for task completion and instead design learning experiences that position AI as a scaffold for metacognitive engagement. This includes embedding structured reflection prompts, peer critique protocols, and “AI audit” tasks that require students to explain, revise, or evaluate AI-generated content. Educators must also build students’ AI literacy, helping them understand not only how to use these tools, but how to question their outputs, recognize their limitations, and maintain ownership of their thinking. For students, the challenge is to resist passive use and develop habits of inquiry, revision, and self-monitoring when interacting with AI. Metacognitive growth requires effortful engagement, and students must learn to treat AI not as a source of answers, but as a tool for enhancing their reasoning. Equipping students with the mindset and skills to think critically with AI, rather than through it, will be essential to building agency, adaptability, and reflective autonomy in AI-mediated learning environments. Directions for Future Research Future research should explore how different instructional models influence students’ long-term metacognitive growth in AI-integrated classrooms, particularly across diverse cultural and socioeconomic contexts. Longitudinal studies could examine whether structured AI use leads to sustained improvements in self-regulation, critical thinking, and learning transfer beyond a single course or program. Additional inquiry is needed into how AI tools can be adapted to support culturally responsive pedagogy, recognizing and valuing the diverse cognitive approaches students bring to learning. As AI continues to evolve, studies must address how learners’ roles shift within these ecosystems and how educational institutions can support equitable access to meaningful, reflective, and empowering AI use. Limitations of the Study This study has several limitations. First, it was conducted within a single apprenticeship program in the American Northeast, which may limit the generalizability of findings to other educational contexts. Second, the small sample size restricts the breadth of conclusions that can be drawn. Third, the study relied on self-reported data and researcher interpretation, which, despite triangulation and reflexive practices, may still carry subjectivity. Fourth, the exclusive focus on ChatGPT limits applicability to other AI tools with different functionalities or interfaces. Finally, the study examined short-term metacognitive outcomes and did not assess the long-term impact of AI use on sustained self-regulated learning. Declarations Declaration Statements The data that support the findings of this study are embedded in the study. The authors declare no funding associated with this study. Permission to conduct research with human subjects was granted by the Institutional Research Board of a University in the United States. Informed consents were obtained from all participants and all names are pseudonyms Consent to publish and participate was obtained from the legal guardians of the parents. The authors declare no conflicts of interest associated with this study. The authors declare no funding associated with this study. The ethics board of a public University in the United States granted permission to conduct research with human subjects. Informed consent was obtained from all participants. Conflict of Interest: None Author Contribution EKG wrote the manuscript solo References Asy’ari, M., & Sharov, S. (2024). Transforming education with CHATGPT: Advancing Personalized Learning, accessibility, and ethical AI integration. International Journal of Essential Competencies in Education , 3 (2), 119–157. https://doi.org/10.36312/ijece.v3i2.2424 Atonte, B. F. (2025). Personalized AI-Driven Education: An AI Framework for Enhanced Learning Outcomes Among School-Age Children. https://doi.org/10.36227/techrxiv.174561166.67234998/v1 Bernacki, M. L., Greene, M. J., & Lobczowski, N. G. (2021a). A systematic review of research on Personalized Learning: Personalized by whom, to what, how, and for what purpose(s)? Educational Psychology Review , 33 (4), 1675–1715. https://doi.org/10.1007/s10648-021-09615-8 Bernacki, M. L., Greene, M. J., & Lobczowski, N. G. (2021b). A systematic review of research on Personalized Learning: Personalized by whom, to what, how, and for what purpose(s)? Educational Psychology Review , 33 (4), 1675–1715. https://doi.org/10.1007/s10648-021-09615-8 Bhavana, S., Jayashree, K., & Rao, T. V. (2024). Navigating AI biases in education. Advances in Educational Marketing, Administration, and Leadership, 135–160. https://doi.org/10.4018/979-8-3693-5443-8.ch005 Chung, C. J. (2024). Preservice teachers’ perceptions of AI in Education. AI-EDU Arxiv, 1–5. https://doi.org/10.36851/ai-edu.vi0.4155 Dahri, N. A., Yahaya, N., Al-Rahmi, W. M., Aldraiweesh, A., Alturki, U., Almutairy, S., Shutaleva, A., & Soomro, R. B. (2024). Extended Tam based acceptance of ai-powered chatgpt for supporting metacognitive self-regulated learning in education: A mixed-methods study. Heliyon , 10 (8). https://doi.org/10.1016/j.heliyon.2024.e29317 Davoodi, A. (2024). Equal ai: A framework for enhancing equity, quality, understanding and accessibility in liberal arts through AI for Multilingual Learners. Language Technology and Social Media , 2 (2), 178–203. https://doi.org/10.70211/ltsm.v2i2.139 El Karafli, K. (2025). Empowering educators with AI. Advances in Computational Intelligence and Robotics, 369–400. https://doi.org/10.4018/979-8-3373-1827-1.ch014 Eunkyung, H. (2024). AI-based Edutech for adaptive teaching and learning. Artificial Intelligence . https://doi.org/10.5772/intechopen.1004952 Fan, L., Kim, J., Detrick, R., & Li, N. (2025). Developing students’ higher-order thinking skills with Generative AI insights and strategies from a case study. Advances in Computational Intelligence and Robotics , 411–428. https://doi.org/10.4018/979-8-3693-7332-3.ch024 Gesser-Edelsburg, A., Hijazi, R., Eliyahu, E., & Tal, A. (2024). Bridging the divide: An empirical investigation of artificial intelligence and generative artificial intelligence integration across genders, disciplines and academic roles. European Journal of Open Distance and E-Learning , 26 (s1), 51–69. https://doi.org/10.2478/eurodl-2024-0008 Hussain, S. R. (2025). The Personalization Paradox: When AI Stops Helping and Starts Overwhelming. https://doi.org/10.2139/ssrn.5223317 Hutson, J., & Plate, D. (2023). Human-AI collaboration for Smart Education: Reframing applied learning to support metacognition. Advanced Virtual Assistants - A Window to the Virtual Future [Working Title]. https://doi.org/10.5772/intechopen.1001832 Kajiwara, Y., Matsuoka, A., & Shinbo, F. (2023). Machine learning role playing game: Instructional design of AI education for age-appropriate in K-12 and beyond. Computers and Education: Artificial Intelligence, 5, 100162. https://doi.org/10.1016/j.caeai.2023.100162 Karlen, Y. (2016). Differences in students’ metacognitive strategy knowledge, motivation, and strategy use: A typology of self-regulated learners. The Journal of Educational Research , 109 (3), 253–265. https://doi.org/10.1080/00220671.2014.942895 Kettler, T., & Taliaferro, C. (2022). The role of technology in personalized learning. Personalized Learning in Gifted Education , 216–242. https://doi.org/10.4324/9781003237136-13 Kim, J., Detrick, R., Yu, S., Song, Y., Bol, L., & Li, N. (2025). Socially shared regulation of learning and Artificial Intelligence: Opportunities to support socially shared regulation. Education and Information Technologies , 30 (9), 11483–11521. https://doi.org/10.1007/s10639-024-13187-9 Komissarov, A. (2025). Artificial Intelligence in education: Challenges and opportunities. ARTIFICIAL INTELLIGENCE IN EDUCATION: CHALLENGES AND OPPORTUNITIES, 151. https://doi.org/10.20319/ictel.2025.151 LaVaque-Manty, D., Meizlish, D., Silver, N., Kaplan, M., & Rhem, J. (2023). Using Reflection and Metacognition to Improve Student Learning. https://doi.org/10.4324/9781003448570 Lee, H., Stinar, F., Zong, R., Valdiviejas, H., Wang, D., & Bosch, N. (2025). Learning behaviors mediate the effect of AI-powered support for metacognitive calibration on learning outcomes. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, 1–18. https://doi.org/10.1145/3706598.3713960 Levin, I., Marom, M., & Kojukhov, A. (2025a). Rethinking AI in education: Highlighting the metacognitive challenge. BRAIN. Broad Research in Artificial Intelligence and Neuroscience , 16 (1 Sup1), 250. https://doi.org/10.70594/brain/16.s1/21 Levin, I., Marom, M., & Kojukhov, A. (2025b). Rethinking AI in education: Highlighting the metacognitive challenge. BRAIN. Broad Research in Artificial Intelligence and Neuroscience , 16 (1 Sup1), 250. https://doi.org/10.70594/brain/16.s1/21 Li, W., & Brooks, C. (2024). Influence on judgements of learning given perceived AI annotations. Proceedings of the Eleventh ACM Conference on Learning @ Scale, 221–231. https://doi.org/10.1145/3657604.3662044 Lopez-Gazpio, I. (2025). Integrating large language models into accessible and Inclusive Education: Access Democratization and individualized learning enhancement supported by Generative Artificial Intelligence. Information , 16 (6), 473. https://doi.org/10.3390/info16060473 Mahmoud, C. F., & Sørensen, J. T. (2024). Artificial Intelligence in personalized learning with a focus on current developments and future prospects. Research and Advances in Education , 3 (8), 25–31. https://doi.org/10.56397/rae.2024.08.04 Marantika, J. E. (2021). Metacognitive ability and autonomous learning strategy in improving learning outcomes. Journal of Education and Learning (EduLearn) , 15 (1), 88–96. https://doi.org/10.11591/edulearn.v15i1.17392 Melisa, R., Ashadi, A., Triastuti, A., Hidayati, S., Salido, A., Ero, P. E., Marlini, C., Zefrin, Z., & Fuad, A., Z (2025). Critical thinking in the age of AI: A systematic review of Ai’s effects on Higher Education. Educational Process International Journal , 14 (1). https://doi.org/10.22521/edupij.2025.14.31 Mercer, N., & Howe, C. (2012). Explaining the dialogic processes of teaching and learning: The value and potential of sociocultural theory. Learning Culture and Social Interaction , 1 (1), 12–21. https://doi.org/10.1016/j.lcsi.2012.03.001 Mergen, A., Çetin-Kılıç, N., & Özbilgin, M. F. (2025). Artificial Intelligence and bias towards marginalised groups: Theoretical roots and challenges. International Perspectives on Equality, Diversity and Inclusion, 17–38. https://doi.org/10.1108/s2051-233320250000012004 Merkebu, J., Kitsantas, A., Durning, S. J., & Ma, T. (2023). What is metacognitive reflection? the moderating role of metacognition on emotional regulation and Reflection. Frontiers in Education . https://doi.org/10.3389/feduc.2023.1166195 . 8. Panarese, P., Grasso, M. M., & Solinas, C. (2025). Algorithmic bias, fairness, and inclusivity: A Multilevel Framework for justice-oriented AI . AI & SOCIETY. https://doi.org/10.1007/s00146-025-02451-2 Rane, N., Choudhary, S., & Rane, J. (2023). Education 4.0 and 5.0: Integrating Artificial Intelligence (AI) for personalized and adaptive learning. SSRN Electronic Journal . https://doi.org/10.2139/ssrn.4638365 Rawat, P., Bajaj, M., Vats, S., & Sharma, V. (2024). Redefining human-AI interactions: Unveiling CHATGPT’s profound emotional understanding. A Practitioner’s Approach to Problem-Solving Using AI, 1–19. https://doi.org/10.2174/9789815305364124010003 Regan, P. M., & Jesse, J. (2018). Ethical challenges of edtech, Big Data and personalized learning: Twenty-first century student sorting and tracking. Ethics and Information Technology , 21(3), 167–179. https://doi.org/10.1007/s10676-018-9492-2 . Rizvi, I., Bose, C., & Tripathi, N. (2025). Transforming education: Adaptive learning, AI, and online platforms for Personalization. Technology for Societal Transformation, 45–62. https://doi.org/10.1007/978-981-96-1721-0_4 Roll, I., & Winne, P. H. (2015). Understanding, evaluating, and supporting self-regulated learning using learning analytics. Journal of Learning Analytics , 2 (1), 7–12. https://doi.org/10.18608/jla.2015.21.2 Roozenbeek, J., & van der Linden, S. (2024). The Psychology of Misinformation. https://doi.org/10.1017/9781009214414 Sharma, D. (2024). Smart Education and sustainable learning environments in smart cities. Advances in Educational Technologies and Instructional Design, 403–430. https://doi.org/10.4018/979-8-3693-7723-9.ch024 Shekh-Abed, A. (2024). Metacognitive self-knowledge and cognitive skills in project-based learning of High School Electronics Students. European Journal of Engineering Education , 1–16. https://doi.org/10.1080/03043797.2024.2374479 Shumaila Farheen, A. A., & Cheema Rai Samee Ullah, & Dr. Marium Minhas Bandeali. (2025). Equity and bias in AI Educational Tools: A Critical Examination of algorithmic decision-making in classrooms. The Critical Review of Social Sciences Studies , 3(3), 67–85. https://doi.org/10.59075/zqmnpa62 Tabib, F. M., & Alrabeei, M. M. (2024a). Can guided CHATGPT use enhance students’ cognitive and metacognitive skills? Studies in Big Data, 143–154. https://doi.org/10.1007/978-3-031-52280-2_10 Tabib, F. M., & Alrabeei, M. M. (2024b). Can guided CHATGPT use enhance students’ cognitive and metacognitive skills? Studies in Big Data, 143–154. https://doi.org/10.1007/978-3-031-52280-2_10 Ushioda, E. (2014). Motivation, autonomy and metacognition. Language Learning & Language Teaching , 31–49. https://doi.org/10.1075/lllt.40.02ush Venter, J., Coetzee, S. A., & Schmulian, A. (2024). Exploring the use of artificial intelligence (AI) in the delivery of effective feedback. Assessment & Evaluation in Higher Education , 50 (4), 516–536. https://doi.org/10.1080/02602938.2024.2415649 Viola, L. (2023). The Humanities in the Digital: Beyond Critical Digital Humanities. https://doi.org/10.1007/978-3-031-16950-2 Vo, K., Sarkar, M., White, P. J., & Yuriev, E. (2025). Metacognitive problem solving: Exploration of students’ perspectives through the lens of multi-dimensional engagement. Chemistry Education Research and Practice , 26 (1), 141–157. https://doi.org/10.1039/d4rp00096j Walker, P. B., Haase, J. J., Mehalick, M. L., Steele, C. T., Russell, D. W., & Davidson, I. N. (2025). Harnessing metacognition for safe and responsible AI. Technologies , 13 (3), 107. https://doi.org/10.3390/technologies13030107 Wang, V., & Design (2025). 231–254. https://doi.org/10.4018/979-8-3373-0502-8.ch010 Wang, W. S., Lin, C. J., Lee, H. Y., Huang, Y. M., & Wu, T. T. (2025). Enhancing self-regulated learning and higher-order thinking skills in virtual reality: The impact of CHATGPT-integrated feedback aids. Education and Information Technologies . https://doi.org/10.1007/s10639-025-13557-x Winne, P. H. (2017). Theorizing and researching levels of processing in self-regulated learning. British Journal of Educational Psychology , 88 (1), 9–20. https://doi.org/10.1111/bjep.12173 Wu, X., Duan, R., & Ni, J. (2024). Unveiling security, privacy, and ethical concerns of chatgpt. Journal of Information and Intelligence , 2 (2), 102–115. https://doi.org/10.1016/j.jiixd.2023.10.007 Xie, X. (2023). Influence of ai-driven inquiry teaching on learning outcomes. International Journal of Emerging Technologies in Learning (iJET) , 18 (23), 59–70. https://doi.org/10.3991/ijet.v18i23.45473 Yamijala, S. M., Chodisetty, R. S., Chakravorty, C., & Sai, K. P. (2024). AI-Powered Learning Revolutionizing Smart Education with personalized learning styles. Advances in Educational Technologies and Instructional Design , 191–212. https://doi.org/10.4018/979-8-3693-8151-9.ch007 Yang, Y., & Xia, N. (2023). Enhancing students’ metacognition via AI-Driven Educational Support Systems. International Journal of Emerging Technologies in Learning (iJET) , 18 (24), 133–148. https://doi.org/10.3991/ijet.v18i24.45647 Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. Theory Into Practice , 41 (2), 64–70. https://doi.org/10.1207/s15430421tip4102_2 Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted 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-7339599","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":498596679,"identity":"ca12f43f-5822-4fa0-b5cc-62c5a876e806","order_by":0,"name":"Eugene Kwasi Gyekye","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYJACxgYGBjkGdh4o9wCRWowZmEnVkthAtBZ+/jWGD2e22aT3N/Me/FzYxiDHdyMBvxbJGW+MDTe2peXOOMyXLD2zjcFYkpAWgxtnzCQfth3ObTjMYyDN28aQuIGQFvsbZ8x/Pmz7ny5/mMf4N1BLPUEtBvw9Zowb2w4kGBzmMQPZkmBASIvEDbZiyRnnkg03HuZLs+Y5J2E488wD/Fr4+w9v/NhTZicvd7z38G2eMht5vuMEbGGQyDBgYGSDchjZJAgoB1tzHOiOPzDeH9wKR8EoGAWjYOQCAJHvSMEW5T22AAAAAElFTkSuQmCC","orcid":"","institution":"New York City Department of Education","correspondingAuthor":true,"prefix":"","firstName":"Eugene","middleName":"Kwasi","lastName":"Gyekye","suffix":""}],"badges":[],"createdAt":"2025-08-10 15:08:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7339599/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7339599/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96246392,"identity":"6417b434-f1c2-449d-830f-0fc5d26b586c","added_by":"auto","created_at":"2025-11-19 07:25:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1040941,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7339599/v1/5c3ab9d6-b24d-49a1-aeb8-653f12b39696.pdf"},{"id":91549697,"identity":"ae611503-2990-4094-ad72-523a230aa82d","added_by":"auto","created_at":"2025-09-17 15:29:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":18187,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7339599/v1/5537699780a2fc8e54c25d87.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"ChatGPT as a Metacognitive Catalyst in Personalized Learning Ecosystems","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe integration of artificial intelligence (AI) tools like ChatGPT into personalized learning ecosystems is transforming contemporary education (Asy'ari \u0026amp; Sharov, 2024), offering adaptive pathways, real-time feedback, and scalable support for diverse student populations (Rizvi et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These advancements promise greater accessibility and tailored learning experiences (Yamijala et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mahmoud \u0026amp; S\u0026oslash;rensen, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, this shift also introduces significant pedagogical, ethical, and infrastructural challenges (Komissarov, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Gesser-Edelsburg et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Kettler and Taliaferro (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) note that while AI has the potential to support personalized instruction, its capacity to respond to the nuanced and evolving needs of individual learners is still in question.\u003c/p\u003e\u003cp\u003eA particularly underexplored dimension of this technological integration is the impact of AI tools like ChatGPT on students\u0026rsquo; metacognitive and self-regulated learning (SRL) capacities (Dahri et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Metacognition, which includes the ability to plan, monitor, and evaluate one\u0026rsquo;s learning strategies, is a critical element of intellectual autonomy and lifelong learning (Marantika, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ushioda, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Xie (\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) intimates that as education increasingly incorporates AI tools, understanding how these tools influence learners\u0026rsquo; metacognitive development is essential (see Yang \u0026amp; Xia, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Atonte, 202). This study explores how educators integrate ChatGPT into personalized learning environments and how its use shapes students\u0026rsquo; development of metacognitive awareness and self-regulation. The central research question guiding this inquiry is:\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eHow does the use of ChatGPT in personalized learning environments influence students' metacognitive development and self-regulated learning?\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Literature Review","content":"\u003cp\u003ePersonalized learning aims to tailor educational experiences to each learner\u0026rsquo;s unique needs, strengths, and preferences. Bernacki et al. (2021) emphasize that personalization involves more than differentiated instruction; it requires a deep understanding of individual learner profiles. These profiles are not solely academic but also include learners\u0026rsquo; metacognitive capacities, such as how they monitor, evaluate, and regulate their own understanding (Karlen, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). As such, effective personalized learning must attend not only to content delivery but also to the learner\u0026rsquo;s ability to reflect on and control their learning processes (Bernacki et al., 2021).\u003c/p\u003e\u003cp\u003eAI tools like ChatGPT are increasingly used to fulfill this promise by offering real-time support, adjusting content difficulty, and enabling scalable one-on-one engagement (Sharma, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). These features have the potential to scaffold students' metacognitive development by prompting self-reflection, tracking progress, and offering iterative feedback (Shekh-Abed, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Venter et al. (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) argue that ChatGPT\u0026rsquo;s responsiveness and accessibility make it particularly valuable in resource-limited settings where human instructional time is constrained. Tabib and Alrabeei (2024) note that in such environments, ChatGPT may act as a metacognitive nudge, helping students plan study strategies, seek clarification, and test their understanding through iterative questioning.\u003c/p\u003e\u003cp\u003eDespite these advantages, AI\u0026rsquo;s effectiveness in truly personalizing instruction and fostering metacognitive growth is debated (Levin et al., 2025; Hutson \u0026amp; Plate, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Mergen et al. (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) point out that ChatGPT and similar AI systems are limited in their ability to interpret emotional cues, cultural nuances, and cognitive diversity. These limitations may result in the marginalization of learners whose metacognitive strategies do not align with the dominant models represented in AI training data (Li \u0026amp; Brooks, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This not only affects content relevance but also impairs students\u0026rsquo; ability to develop self-awareness and adaptive learning strategies.\u003c/p\u003e\u003cp\u003eKettler and Taliaferro (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) add that while AI may efficiently deliver instructional content, it lacks the contextual awareness and relational dynamics that human educators bring to the learning process. According to Eunkyung (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), these relational elements are essential for developing metacognitive habits such as modeling reflective practice, encouraging intellectual risk-taking, and providing feedback that prompts self-correction. Without the ability to sense learners\u0026rsquo; confusion, hesitation, or confidence levels, ChatGPT may inadvertently flatten the reflective space needed for metacognitive development (Roozenbeek \u0026amp; van der Linden, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis disconnection between content delivery and authentic personalization challenges the assumption that AI inherently enhances individualized learning. Instead, it raises concerns that AI-based platforms may present a mechanized form of education that mimics personalization without addressing learners\u0026rsquo; holistic development (Rane et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Crucially, this includes the development of self-regulation and reflective awareness, which are foundational to metacognitive learning (Merkebu et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAccording to Chung (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), AI models, including ChatGPT, are trained on extensive datasets that reflect societal biases, which can manifest in skewed or culturally insensitive responses (see Panarese et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Wang (\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) highlights how such biases can undermine efforts toward educational equity by reinforcing dominant cultural narratives and excluding marginalized perspectives. From a metacognitive perspective, Fan et al. (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) opine that exposure to biased content can distort students\u0026rsquo; reflective judgments, misguide their information evaluation processes, and potentially erode trust in digital learning tools.\u003c/p\u003e\u003cp\u003eThese issues are especially problematic in personalized learning contexts where inclusivity is assumed (Davoodi, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Bhavana et al. (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) note that when AI systems deliver recommendations or feedback shaped by biased data, students from underrepresented backgrounds may receive inferior guidance or feel alienated from the learning process. In the view of Lee et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), such experiences can impair metacognitive efficacy, especially if learners begin to question the validity of feedback, the reliability of their strategies, or the fairness of the educational environment. This contradiction challenges the ethical foundation of AI-enhanced personalization (Hussain, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), making it critical to examine not just whether students receive differentiated instruction but whether that instruction supports reflective, self-directed learning for all.\u003c/p\u003e\u003cp\u003eAccording to Regan and Jesse (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), beyond bias, ethical concerns extend to data privacy, student dependence, and the shifting role of educators. ChatGPT\u0026rsquo;s functionality relies on the collection and processing of user inputs, raising concerns about how personal data is stored, shared, and protected (Wu et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). From a metacognitive standpoint, lack of transparency about how AI models use learner data can inhibit trust and hinder students\u0026rsquo; willingness to engage in open self-reflection and questioning (Levin et al., 2025). Metacognition requires learners to be honest and vulnerable about what they know and do not know (LaVaque-Manty et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). If they fear surveillance or data misuse, their ability to engage in metacognitive dialogue with AI tools may be compromised (Walker et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDependence on AI-generated assistance may also lead to negative learning behaviors. Gesser-Edelsburg et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) argue that students may over-rely on ChatGPT\u0026rsquo;s convenience, bypassing the reflective and effortful processes required for deep understanding. In my experience as an urban educator, I have seen how, when learners habitually defer to AI for explanations, their critical thinking and problem-solving capacities may deteriorate. This risks diminishing metacognitive monitoring and control, as students are less likely to pause, evaluate alternative strategies, or question the validity of a given response. Viola (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) notes that such overdependence challenges the goal of cultivating autonomous, self-regulated learners who can transfer knowledge and skills across domains.\u003c/p\u003e\u003cp\u003eWhile AI can supplement instruction, it cannot replace the mentorship, empathy, and human-centered adaptability that teachers provide (El Karafli, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Kajiwara et al. (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) note that educators play a vital role in modeling metacognitive behaviors such as articulating their thinking, demonstrating how to revise misconceptions, and encouraging learners to reflect on their learning processes. The challenge lies in finding a balanced model where AI enhances learning without diminishing the relational aspects of education that support emotional, social, and cognitive development.\u003c/p\u003e\u003cp\u003eAccording to Vo et al. (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), the risk of diminished metacognitive engagement, underscores the importance of pedagogical scaffolding. Educators must guide students in using ChatGPT to support rather than replace metacognitive effort (Tabib \u0026amp; Alrabeei, 2024). Melisa et al. (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) note that instructional strategies might include requiring students to compare their reasoning with the AI\u0026rsquo;s output, explain their choices, or revise their thinking after receiving feedback. Such practices can help ensure that ChatGPT serves as a metacognitive catalyst rather than a cognitive crutch.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eResearch Gap and Theoretical Contribution\u003c/h2\u003e\u003cp\u003eWhile AI tools like ChatGPT are increasingly present in education, there is limited empirical research examining their impact on students\u0026rsquo; metacognitive development and self-regulated learning. Most existing studies focus on outcomes such as performance, user satisfaction, or system efficiency. These areas, though important, largely overlook the internal cognitive and affective processes that shape learners' ability to think critically, plan strategically, and reflect on their learning. Without understanding how AI affects these higher-order skills, educators risk deploying technology in ways that compromise long-term educational goals.\u003c/p\u003e\u003cp\u003eTo address this gap, this study investigates the research question: How does the use of ChatGPT in personalized learning environments influence students\u0026rsquo; metacognitive development and self-regulated learning? Rather than viewing students as passive recipients of AI-generated content, this study emphasizes their role as active learners capable of self-monitoring, reflection, and adaptation.\u003c/p\u003e\u003cp\u003eIn response to the gap, this research introduces a new theoretical perspective, AI-Mediated Metacognitive Development (AIMD), that builds on Social Constructivism and Self-Regulated Learning Theory. AIMD conceptualizes AI not as a neutral tool, but as a dynamic influence that can either support or hinder students\u0026rsquo; metacognitive growth, depending on how it is integrated into learning environments.\u003c/p\u003e\u003c/div\u003e"},{"header":"Theoretical Framework","content":"\u003cp\u003eThe integration of artificial intelligence (AI) into personalized learning environments calls for a robust theoretical foundation to evaluate its impact on students' cognitive and metacognitive development. This study draws upon two foundational educational theories: Social Constructivism and Self-Regulated Learning (SRL) Theory to examine how ChatGPT supports or hinders the development of learner autonomy, critical thinking, and self-awareness.\u003c/p\u003e\n\u003ch3\u003eSocial Constructivism: AI as a Mediator of Interaction\u003c/h3\u003e\n\u003cp\u003eGrounded in Vygotsky\u0026rsquo;s (1978) Social Constructivism, learning is understood as a socially mediated process where knowledge is co-constructed through interaction. A central concept in this theory is the Zone of Proximal Development (ZPD), which suggests learners can achieve higher levels of understanding with appropriate scaffolding from more knowledgeable others. In AI-integrated classrooms, tools like ChatGPT can function as digital scaffolds, offering context-specific feedback, explanations, and prompts aligned with learners\u0026rsquo; needs.\u003c/p\u003e\u003cp\u003eHowever, while ChatGPT may facilitate access to knowledge, it lacks the emotional intelligence and cultural sensitivity inherent in human interaction (Rawat et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Because AI cannot engage in reciprocal dialogue or perceive socio-emotional cues, overreliance may reduce opportunities for rich, collaborative learning experiences. Vygotsky emphasized that authentic learning occurs within socially meaningful contexts - an element that AI, in its current form, cannot fully replicate (Vygotsky, 1978; Mercer \u0026amp; Howe, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eSelf-Regulated Learning: AI and Metacognitive Support\u003c/h3\u003e\n\u003cp\u003eComplementing constructivist theory, SRL Theory (Zimmerman, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2002\u003c/span\u003e) provides a lens for understanding how learners manage their own learning processes. SRL encompasses three cyclical phases: forethought (goal-setting and planning), performance (strategy use and monitoring), and self-reflection (evaluating and adjusting approaches). Wang et al. (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) posit that ChatGPT can assist across these phases by prompting goal clarification, offering strategy suggestions, and providing feedback that supports real-time reflection and metacognitive monitoring.\u003c/p\u003e\u003cp\u003eDespite these affordances, AI may unintentionally undermine SRL by encouraging cognitive offloading ( see Kim et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), where students outsource thinking to the tool instead of developing independent reasoning. Unlike human mentors who can gradually reduce support to foster autonomy, AI lacks the capacity to calibrate its assistance based on developmental cues or encourage productive struggle (Winne, 2018; Roll \u0026amp; Winne, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). When students become passive recipients of AI outputs, their capacity for self-regulation and critical thinking may diminish.\u003c/p\u003e\n\u003ch3\u003eAI-Mediated Metacognitive Development (AIMD): A Theoretical Construct\u003c/h3\u003e\n\u003cp\u003eI construe the AI-Mediated Metacognitive Development (AIMD) as the process through which artificial intelligence, particularly generative AI tools like ChatGPT, mediates, supports, or impedes the development of learners\u0026rsquo; metacognitive capacities within personalized learning environments. Rather than conceptualizing AI as a passive instructional aid, I position AI as an active agent that significantly shapes learners' abilities to monitor, evaluate, and regulate their thinking.\u003c/p\u003e\u003cp\u003eI ground AIMD in two complementary theoretical traditions: Social Constructivism, which emphasizes the role of mediated social and cultural tools in cognitive development, and Self-Regulated Learning (SRL) Theory, which focuses on learners\u0026rsquo; capacity for planning, monitoring, and evaluating their own learning processes. Together, these perspectives provide a robust foundation for understanding how AI can enhance or hinder metacognitive development depending on the conditions of its use.\u003c/p\u003e\u003cp\u003eFrom a Social Constructivist perspective, learning is inherently mediated by tools and artifacts situated within social contexts. AIMD extends this concept to artificial intelligence, treating AI as a digital cognitive artifact that co-constructs knowledge with learners. Through interactive dialogue, feedback loops, and modeled reasoning, AI tools like ChatGPT engage with learners in ways that can influence their conceptual understanding and thought patterns.\u003c/p\u003e\u003cp\u003eI argue that these AI systems do not merely transmit information; they actively shape the discourse through which learners explore and refine their thinking. In parallel, Self-Regulated Learning theory positions metacognition, defined as awareness and regulation of one\u0026rsquo;s thinking, as a foundational element of effective learning. AIMD builds on this theory by analyzing how AI affects each phase of self-regulation: from planning and goal-setting, to real-time monitoring, to post-task evaluation and reflection.\u003c/p\u003e\u003cp\u003eAIMD identifies the mediating function of AI as central to its impact on metacognitive development. Thus, AI can mediate learning in both facilitative and inhibitive ways. When integrated intentionally into pedagogically sound learning environments, AI can play a facilitative role by prompting learners to reflect on their reasoning, offering iterative feedback, scaffolding goal setting, and encouraging strategic decision-making. For instance, students using ChatGPT can be guided to justify their responses, pose follow-up questions, or test alternative solutions. These interactions foster deeper metacognitive engagement and enhance learners\u0026rsquo; awareness of how they approach and solve problems.\u003c/p\u003e\u003cp\u003eHowever, when AI is used in an unstructured, convenience-driven manner (such as asking for direct answers without engaging with the process), it can inhibit metacognitive growth. In such cases, learners may bypass cognitive effort, develop dependency on AI-generated outputs, and engage in cognitive offloading, thereby reducing opportunities for reflective thought and strategic learning.\u003c/p\u003e\u003cp\u003eTo support sustained development, AIMD introduces the concept of \u003cem\u003edynamic scaffolding\u003c/em\u003e, which sees AI as a temporary support structure. Effective integration of AI into learning follows a developmental trajectory: \u003cem\u003esupport, fade, and internalize\u003c/em\u003e. Initially, AI offers structured prompts and guidance to help learners navigate new tasks or concepts. As learners build confidence and competence, the level of AI assistance should gradually fade. Eventually, learners are expected to internalize metacognitive strategies and apply them independently, without relying on the tool. This model aligns with cognitive apprenticeship frameworks, where learners move from guided participation to autonomous mastery, absorbing expert behaviors through repeated exposure and practice.\u003c/p\u003e\u003cp\u003eAnother core dimension of AIMD is \u003cem\u003elearner-AI interaction as cognitive apprenticeship\u003c/em\u003e. In this view, AI tools serve as cognitive partners that model expert-like thinking. ChatGPT, for example, can demonstrate how to evaluate competing arguments, reason through ambiguity, or reflect on past errors. Through repeated interaction, learners absorb these metacognitive behaviors and begin to emulate them independently. Over time, the AI transitions from an active guide to a reflective sounding board, supporting the learner\u0026rsquo;s journey toward cognitive autonomy.\u003c/p\u003e\u003cp\u003eAIMD also emphasizes the importance of \u003cem\u003emetacognitive intentionality\u003c/em\u003e in instructional design. I argue that AI should not merely function as an answer generator. Rather, its use must be aligned with explicit metacognitive goals that encourage students to justify their reasoning, compare multiple solutions, identify knowledge gaps, and establish learning objectives. This intentional design promotes deeper cognitive engagement and transforms AI from a static information source into a dynamic thinking partner. The ultimate aim is not simply to help students complete tasks more efficiently, but to cultivate learners who are more reflective, strategic, and aware of their own learning processes.\u003c/p\u003e\u003cp\u003eImportantly, AIMD acknowledges that the benefits of AI-mediated metacognitive development are not equally distributed. In view of this, I foreground ethical and equity considerations to cater for learners from underserved communities. Learners from underserved communities may face systemic barriers that hinder their ability to benefit fully from AI-enhanced learning environments. These include limited access to high-quality digital tools, insufficient AI literacy, and prior educational experiences that undermine academic confidence. As a result, some students may over-rely on AI for quick answers rather than engage it as a reflective tool. AIMD, therefore, demands \u003cem\u003eequity-conscious pedagogy\u003c/em\u003e that provides targeted support, models effective AI use, and fosters reflective agency among all learners, regardless of background. Only through intentional and inclusive design can AI become a tool for cognitive empowerment rather than another mechanism of stratification.\u003c/p\u003e\u003cp\u003eAs a contribution to educational theory, this framework reframes the role of AI in learning from that of content delivery to that of cognitive mediation. It bridges human\u0026ndash;machine interaction with established theories of learning by showing how AI influences metacognitive development in both constructive and constraining ways. The framework introduces a dialectical view of AI as both catalyst and constraint, depending on its use context. For practitioners, AIMD provides a set of guiding principles to design, implement, and evaluate AI-integrated instruction. It helps educators create learning tasks that provoke strategic thinking, monitor and assess learners\u0026rsquo; metacognitive engagement with AI, and balance the use of technological support with opportunities for independent reflection.\u003c/p\u003e"},{"header":"Research Methodology","content":"\u003cp\u003eThis qualitative study drew on constructivist and interpretive research traditions to examine how students and educators engaged with ChatGPT in a blended online learning environment. Guided by the frameworks of AI-Mediated Metacognitive Development (AIMD), Self-Regulated Learning (SRL), and Self-Regulated Learning (SRL) Theory, the study focused on how ChatGPT shaped metacognitive processes and equitable learning experiences among high school students from underserved communities.\u003c/p\u003e\u003cp\u003eThe research was conducted during the 2023\u0026ndash;2024 academic year within a youth apprenticeship program facilitated by the Innovation Learning Collective (a pseudonym), a nonprofit based in the American Northeast. The program served a diverse cohort of 100 students: 40% Black, 30% Hispanic, 20% White, and 10% Asian, roughly half of whom qualified for free or reduced-price lunch. Students participated in synchronous workshops and asynchronous activities designed to build career readiness skills through project-based learning and responsible AI use.\u003c/p\u003e\u003cp\u003eUsing purposive sampling (Patton, 1990; Maxwell, 2022), participants were selected from three stakeholder groups: students, facilitators, and program managers. Five high school students were chosen to reflect variability in race, socioeconomic status, and digital fluency. Their perspectives provided insight into how ChatGPT supported identity, agency, and self-regulation. Five facilitators representing STEM, humanities, and career development disciplines contributed instructional perspectives on AI implementation. Two program managers offered institutional context regarding curriculum design and equity goals.\u003c/p\u003e\u003cp\u003eData collection included 12 semi-structured interviews with students, 9 with facilitators, and 3 with program managers. These were complemented by ongoing ethnographic interviews, 300 hours of field observations during synchronous sessions, and analysis of student-generated artifacts such as reflective essays and ChatGPT-influenced assignments. All data were transcribed and thematically coded to identify patterns in how ChatGPT was used, interpreted, and evaluated by participants. Cross-case comparisons and triangulation ensured credibility and analytic depth. This multi-dimensional approach enabled a rich examination of the central question: How does the use of ChatGPT in personalized learning environments influence students\u0026rsquo; metacognitive development and self-regulated learning?\u003c/p\u003e\u003c/div\u003e"},{"header":"Research Context","content":"\u003cp\u003eThis study was conducted during the 2023\u0026ndash;2024 academic year within a youth apprenticeship program hosted by the Innovation Learning Collective (a pseudonym), a nonprofit in the American Northeast. Designed to build career readiness skills through personalized and equity-centered learning, the program served 100 high school students from culturally and socioeconomically diverse backgrounds: 40% Black, 30% Hispanic, 20% White, and 10% Asian, with about half qualifying for free or reduced-price lunch.\u003c/p\u003e\u003cp\u003eStudents were organized into five groups, each supported by two facilitators and overseen by program managers. Instruction followed a blended synchronous and asynchronous online learning model. During synchronous sessions, students engaged in guided discussions and interactive demonstrations of ChatGPT\u0026rsquo;s capabilities, with an emphasis on ethical use, critical analysis of AI outputs, and learning equity. In asynchronous settings, students used ChatGPT to support project-based assignments, reflective writing, and independent research, positioning the tool as a metacognitive aid in goal-setting, self-monitoring, and strategic thinking.\u003c/p\u003e\n\u003ch3\u003eParticipants\u003c/h3\u003e\n\u003cp\u003eA purposive sampling strategy (Patton, 1990; Maxwell, 2022) was used to select participants from three stakeholder groups; students, facilitators, and program managers, ensuring diverse perspectives on ChatGPT\u0026rsquo;s role in personalized, AI-integrated learning.\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eStudent Participants\u003c/h2\u003e\u003cp\u003eFive students were selected to reflect variation in race, socioeconomic status, gender, and digital fluency. All were actively enrolled in the AI-assisted course and used ChatGPT across structured and exploratory learning tasks.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eDemographic Characteristics of Student Participants\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eName\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEthnicity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSocioeconic Status\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eDigital Fluency\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAlex\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e18\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eAfrican American\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eEconomically Disadvantaged\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eFluent\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eJordan\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e17\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eAfrican American\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eEconomically Advantaged\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eSemi-Fluent\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTaylor\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e17\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eWhite\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eEconomically Advantaged\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eFluent\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMorghan\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e18\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eAsian\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eEconomically Advantaged\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eFluent\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCasey\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e17\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eLatinx\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eEconomically Disadvantaged\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eSemi-Fluent\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThese students were the focal point for analyzing how ChatGPT shaped metacognitive engagement within a sociotechnical and culturally relevant pedagogical framework. Their voices illuminated how AI tools either fostered or hindered agency, identity affirmation, and learning autonomy.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eFacilitator Participants\u003c/h2\u003e\u003cp\u003eFive educators contributed instructional perspectives from STEM, humanities, and career development disciplines. Their experiences informed how ChatGPT was implemented pedagogically and how it supported or constrained culturally responsive and metacognitive learning.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eDemographic Characteristics of Facilitator Participants\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eName\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEthnicity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eClass Taught\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMs. Hailey\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e28\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eAfrican American\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eWorkforce Readiness\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMr. Goneday\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e24\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eAfrican American\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eSTEM\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMs. Goods\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e26\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eLatinx\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eSTEM\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMr. Griffiths\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e29\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eWhite\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eHumanities\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMr. Andrew\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e25\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eLatinx\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eHumanities\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTheir inclusion enabled an examination of how facilitators mediated students\u0026rsquo; metacognitive development through instructional choices, tool design, and equitable access\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eProgram Manager Participants\u003c/h2\u003e\u003cp\u003eTwo program managers; Mr. Gobble and Ms. Green offered strategic insights into the design, goals, and equity considerations underpinning ChatGPT integration in the learning ecosystem.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eDemographic Characteristics of Program Manager Participants\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eName\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEthnicity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDesignation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMr. Gobble\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e30\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eWhite\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eProgram Manager\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMs. Green\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cem\u003e28\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eWhite\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eProgram Manager\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTheir perspectives were essential for understanding the institutional context in which ChatGPT was positioned as a tool for promoting student agency, critical reflection, and equitable AI use.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eData Sources\u003c/h2\u003e\u003cp\u003eData collection included 12 semi-structured interviews (see Appendix A) with students, 9 with facilitators, and 3 with program managers. These were augmented by ongoing ethnographic interviews, analysis of student-generated artifacts (e.g., reflective essays, ChatGPT-assisted work products), and approximately 300 hours of synchronous observations (see Appendix B) and generated field notes capturing real-time student interaction with AI. All data were transcribed and coded using thematic analysis to explore how ChatGPT functioned as a metacognitive catalyst. Triangulation and multi-stakeholder perspectives supported a robust understanding of how students navigated personalized learning with AI under varying structural and social conditions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eResearcher Positionality\u003c/h2\u003e\u003cp\u003eAs the sole author of this study, I identify as an African American educator-researcher with a longstanding commitment to culturally responsive and equity-centered pedagogies. This research was conducted within a youth apprenticeship program where I previously served as an instructor, and while I did not teach the student participants during the study period, my ongoing relationship with the site and shared cultural heritage with many of the learners fostered an atmosphere of trust, openness, and mutual respect. My proximity to the community allowed for deeper insight into the sociocultural contexts that shaped students\u0026rsquo; interactions with AI tools like ChatGPT, particularly in terms of access, agency, and reflective learning.\u003c/p\u003e\u003cp\u003eRecognizing that my positionality could influence data interpretation, I employed reflexive practices throughout the research process, including memoing, peer debriefing, and journaling, to surface and bracket my assumptions. I was particularly attuned to the ways that my advocacy for metacognitive learning and digital equity could shape the thematic emphasis of the analysis. To mitigate this, I foregrounded participant voice, allowed for disconfirming evidence, and remained anchored in the study\u0026rsquo;s theoretical frameworks of AI-Mediated Metacognitive Development (AIMD), Social Constructivism, and Self-Regulated Learning. Transparency with participants about the research goals further ensured an ethical, dialogic approach, reinforcing a shared commitment to understanding how AI can serve not merely as a technological tool, but as a catalyst for deeper, more equitable learning.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eControl Measures in Data Collection\u003c/h2\u003e\u003cp\u003eTo ensure theoretical coherence and methodological rigor, this study employed carefully designed control measures across all data collection activities. These measures were grounded in the study\u0026rsquo;s theoretical framework, Social Constructivism, Self-Regulated Learning (SRL), and AI-Mediated Metacognitive Development (AIMD) to examine how ChatGPT mediates students\u0026rsquo; metacognitive development within personalized learning environments. Each data instrument was intentionally designed to assess how learners interacted with AI as a cognitive partner, a source of scaffolding, and a potential site of cognitive offloading.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eInterviews\u003c/h2\u003e\u003cp\u003eSemi-structured interview protocols were developed to explore how students and educators perceived ChatGPT\u0026rsquo;s role in metacognitive processes such as planning, monitoring, and reflecting on learning. Drawing from SRL theory (Zimmerman, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2002\u003c/span\u003e), interview prompts were designed to trace each phase of the self-regulatory cycle. Questions focused on learners\u0026rsquo; strategies before, during, and after AI use, and invited reflection on goal-setting, decision-making, and adaptation.\u003c/p\u003e\u003cp\u003eFrom a Social Constructivist lens, interview questions also probed how learners engaged with ChatGPT as a digital scaffold; what kinds of dialogue they co-constructed with the tool, and whether these interactions reflected internalized learning. The AIMD framework informed questions regarding dependency, autonomy, and strategic fading of AI support, to assess whether learners were progressing toward internalization or becoming reliant on the tool. All interview protocols were piloted with high school students and educators to ensure alignment with the theoretical constructs, clarity of language, and cultural responsiveness. Interviews were audio-recorded, transcribed verbatim, and verified through member-checking to maintain fidelity to participants\u0026rsquo; voices and cognitive frames.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eObservations\u003c/h2\u003e\u003cp\u003eObservational data were gathered across 300 hours of classroom instruction using a structured protocol derived from the AIMD framework. The observation form captured indicators of students\u0026rsquo; metacognitive engagement, strategic use of AI, collaborative meaning-making, and awareness of bias or inequity in AI outputs.\u003c/p\u003e\u003cp\u003eSocial Constructivism informed the documentation of peer-to-peer AI interaction and the social contexts in which AI use occurred. Observers noted whether learners engaged in co-construction of knowledge through discussion, critique, or collaborative prompting with ChatGPT. The SRL framework guided attention to strategic behavior, such as whether students refined prompts, monitored outputs, and evaluated ChatGPT\u0026rsquo;s feedback in real time.\u003c/p\u003e\u003cp\u003eAIMD shaped the observational categories to reflect whether AI was functioning as a scaffold, a mentor, or a crutch. The protocol enabled researchers to document both productive AI-mediated metacognitive engagement and signs of cognitive offloading or passive consumption. \u003cb\u003eT\u003c/b\u003ehis observation protocol ensured consistent data collection across contexts, reduced observer bias, and operationalized theoretical constructs in concrete classroom behavior.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eExpert Validation\u003c/h2\u003e\u003cp\u003eTo ensure analytical validity, two colleagues of mine who are academics in qualitative research and metacognitive learning reviewed a subset of coded transcripts, observation notes, and emergent themes. Their task was to examine whether the data interpretations were consistent with the theoretical framework, particularly the core claims of the AIMD model. Their review confirmed that the data coding procedures and thematic interpretations faithfully represented Social Constructivist principles of mediated learning, SRL concepts of metacognitive strategy use, and AIMD\u0026rsquo;s emphasis on AI as both scaffold and constraint. Expert validation further strengthened the credibility of the findings by confirming that themes such as strategic fading, critical reflection, and AI-assisted autonomy were not only emergent from the data but theoretically coherent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003eData Analysis\u003c/h2\u003e\u003cp\u003e This study\u0026rsquo;s data analysis was conducted through a qualitative, inductive lens guided by the AI-Mediated Metacognitive Development (AIMD) framework, which integrates Social Constructivism and Self-Regulated Learning (SRL) theories. To capture the nuanced ways learners engaged with AI tools like ChatGPT in personalized learning environments, the analysis followed a three-phase coding process: open coding, axial coding, and selective coding. Each phase progressively distilled the data into more abstract and conceptual themes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003eOpen Coding: Raw Data Exploration\u003c/h2\u003e\u003cp\u003eThe initial phase employed open coding to identify discrete actions, perceptions, and reflections expressed by students regarding their interactions with ChatGPT. During this phase, verbatim quotes were extracted and interpreted to uncover learners\u0026rsquo; immediate responses to AI\u0026rsquo;s affordances and constraints. Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e1\u003c/span\u003e below exemplifies the open codes, paired with representative quotes and interpretive comments that revealed emergent metacognitive and behavioral patterns.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eData Chart Excerpt for Open Coding with Illustrative Quotes\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOpen Code\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIllustrative Quote\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInterpretation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDistrust of AI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I don\u0026rsquo;t trust everything it gives me, so I check it with other sources.\u0026rdquo; \u0026ndash; Morgan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIndicates evaluative thinking and digital skepticism\u0026mdash;student questions AI authority.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRewriting for Tone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;When I asked ChatGPT to help with an email, it sounded too chill. I had to change it so it looked right.\u0026rdquo; - Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudent demonstrates critical literacy and revision, not accepting AI at face value.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRecognizing AI Limits\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I\u0026rsquo;ve seen ChatGPT mess up when it tries to do too much.\u0026rdquo; \u0026ndash; Jordan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudent shows awareness of AI overreach; reflects analytical monitoring.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIndependent Look-Up\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;Sometimes it just doesn\u0026rsquo;t make sense. I look stuff up on my own after.\u0026rdquo; - Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eReflects independent learning and cross-checking strategies.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCopy-Pasting AI Outputs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;If I\u0026rsquo;m stuck, I just type it in and copy what looks good.\u0026rdquo; \u0026ndash; Jordan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSurface-level use of AI; cognitive offloading without self-evaluation.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePassive Use\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I usually just go with the first thing it gives me, unless it sounds really weird.\u0026rdquo; \u0026ndash; Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMinimal engagement or reflection; trust in AI defaults.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCuriosity About AI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;It looks cool when they use it right. Maybe I\u0026rsquo;d try if I knew more about how it works.\u0026rdquo; \u0026ndash; Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudent expresses interest but lacks access or understanding\u0026mdash;entry point to AI literacy.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDesire for Explanation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I wish it told them why it said that.\u0026rdquo; \u0026ndash; Mr. Griffiths\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCraving for AI transparency; signals desire to engage more deeply.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducator Scaffolding\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;We had to explain what we changed in the ChatGPT draft. That part made me think harder.\u0026rdquo; \u0026ndash; Ms. Hailey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAssignment promotes metacognitive reflection and ownership of learning.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePeer Critique Activities\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;They debated which AI answers were more persuasive and why.\u0026rdquo; - Mr. Gobble\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAI used as a social object for dialogue, promoting co-regulation and reflective evaluation.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThese initial codes revealed a complex spectrum of learner interactions with AI: from critical evaluation and revision to passive acceptance and over-reliance. Several learners expressed curiosity and a desire for transparency in AI\u0026rsquo;s reasoning, which suggested nascent forms of metacognitive intentionality and agency.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003eAxial Coding: Thematic Grouping\u003c/h2\u003e\u003cp\u003eBuilding on the open codes, axial coding grouped related concepts into broader thematic categories that described how students engaged cognitively and socially with ChatGPT. This phase connected discrete behaviors to conceptual patterns of metacognitive engagement or disengagement and identified the role of scaffolding and co-design in AI use. Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents these thematic categories alongside representative quotes and interpretive commentary.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eData Chart Excerpt for Axial Coding with Thematic Categories\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThematic Category\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIllustrative Quote\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInterpretation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eCritical AI Engagement\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I don\u0026rsquo;t trust everything it gives me, so I check it with other sources.\u0026rdquo; - Morgan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudents display active metacognition: evaluating, editing, and cross-referencing AI output.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;It sounded too chill. I had to change it.\u0026rdquo; - Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eReflects critical literacy and deliberate revision of AI-generated content.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I look stuff up on my own after.\u0026rdquo; - Taylor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIndicates independent verification and strategic monitoring of AI.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCognitive Offloading \u0026amp; Over-Reliance\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I just type it in and copy what looks good.\u0026rdquo; - Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eShows superficial engagement and reliance on AI as a shortcut, limiting deeper cognitive processing.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I usually just go with the first thing it gives me.\u0026rdquo; Morghan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eReveals passive trust and minimal metacognitive effort.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eStructured AI Scaffolding\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;They had to explain what they changed.\u0026rdquo; - Ms. Goods\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEducator scaffolding promotes reflection and metacognitive awareness through active revision.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;We debated which AI answers were better.\u0026rdquo; - Mr. Andrew\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTeacher initiated peer discourse activates co-regulation and critical evaluation.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCo-Designing Digital Learning\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I wish it told me why it said that.\u0026rdquo; - Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eExpresses learner desire for transparency and control, signaling emerging metacognitive agency.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;Maybe I\u0026rsquo;d try if I knew more about how it works.\u0026rdquo; - Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHighlights the need for accessible AI literacy and inclusive design to foster learner engagement.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThis phase illustrated how metacognitive outcomes depended heavily on the context of AI use: when learners were scaffolded and engaged socially, they demonstrated critical evaluation and strategic thinking. Conversely, unsupported use often led to cognitive offloading, reducing opportunities for metacognitive growth. The aspiration for transparency and deeper understanding suggested fertile ground for designing AI systems that promoted learner autonomy and reflection.\u003c/p\u003e\u003cdiv id=\"Sec23\" class=\"Section3\"\u003e\u003ch2\u003eSelective Coding: Core Themes, Illustrative Code and Interpretation\u003c/h2\u003e\u003cp\u003eIn the final phase, selective coding synthesized the axial themes into core theoretical constructs aligned with the AIMD framework. This phase distilled the dualistic influence of AI on metacognitive development into facilitative and inhibitive roles, while highlighting the importance of learner agency in AI integration. Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e3\u003c/span\u003e summarizes these core themes with supporting quotes and interpretations.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cem\u003eData Chart Excerpt for Selective Coding\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThematic Category\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIllustrative Quote\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInterpretation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eFacilitative Role of AI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I don\u0026rsquo;t trust everything it gives me\u0026hellip;\u0026rdquo; - Morgan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eChatGPT supports metacognitive growth when paired with critical thinking and scaffolding.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I look stuff up on my own.\u0026rdquo; - Taylor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLearners develop independent verification strategies alongside AI use.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;We debated which AI answers were more persuasive.\u0026rdquo; - Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStructured peer dialogue fosters analytical evaluation and reflective reasoning.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eInhibitive Role of AI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I just copy what looks good.\u0026rdquo; - Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAI can encourage passive learning and cognitive offloading without metacognitive reflection.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I just go with the first thing.\u0026rdquo; - Jordan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLack of scaffolding leads to diminished learner autonomy and reflective capacity.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eStudent Agency in AI Integration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;I wish it told me why it said that.\u0026rdquo; - Casey\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLearners desire AI transparency and control, crucial for metacognitive engagement and empowerment.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026ldquo;Maybe I\u0026rsquo;d try if I knew more\u0026hellip;\u0026rdquo; - Alex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHighlights importance of equitable and inclusive AI literacy to foster learner agency and curiosity.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe selective coding phase crystallized the dynamic tension inherent in AI-mediated learning environments. AI functioned as a double-edged sword: it facilitated metacognitive development when thoughtfully integrated with pedagogical scaffolds, but it also inhibited reflective learning when used passively. Importantly, learners\u0026rsquo; calls for transparency and understanding emphasized the need for AI tools designed to empower rather than replace metacognitive processes.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Findings","content":"\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003cp\u003eThis study examined how ChatGPT shaped high school students\u0026rsquo; metacognitive development within a culturally diverse, personalized learning environment. Drawing on the AI-Mediated Metacognitive Development (AIMD) framework, findings illuminated three overarching themes: (1) AI as a scaffold for metacognitive growth, (2) risks of cognitive offloading, and (3) student agency as a mediating force. These themes were grounded in Social Constructivist theory and Self-Regulated Learning (SRL) principles, revealing how AI\u0026rsquo;s influence depended on context, instructional design, and learners\u0026rsquo; cultural and cognitive positioning.\u003c/p\u003e\u003cp\u003eIn environments where ChatGPT was used with explicit pedagogical scaffolding, students demonstrated increased metacognitive engagement. Learners reflected on their thought processes, questioned AI outputs, and revised their work with a growing awareness of their cognitive strategies. Taylor, described how ChatGPT prompted her to reconsider her reasoning: \u0026ldquo;I was working on an argument for my history project. ChatGPT gave me one perspective, but then I asked, \u0026lsquo;what\u0026rsquo;s the counterargument?\u0026rsquo; and that made me rethink everything. I hadn\u0026rsquo;t even considered that angle until it brought it up.\u0026rdquo; Taylor\u0026rsquo;s interaction exemplified strategic monitoring, a core SRL process, where learners engage in evaluating alternative approaches. Her use of AI moved beyond passive acceptance toward self-questioning and adaptive thinking which are hallmarks of metacognitive maturity.\u003c/p\u003e\u003cp\u003eSimilarly, Casey, emphasized how facilitator-designed assignments elevated her reflection: \u0026ldquo;We had to explain what we changed in the ChatGPT draft. That part made me think harder. Like, why did I reword it? What didn\u0026rsquo;t feel right?\u0026rdquo; Here, the facilitator\u0026rsquo;s prompt functioned as a \u0026ldquo;metacognitive nudge,\u0026rdquo; consistent with AIMD\u0026rsquo;s dynamic scaffolding concept, where AI support was framed as temporary and strategically faded to promote autonomy.\u003c/p\u003e\u003cp\u003eFacilitators played a central role in activating this scaffolding. Mr. Andrew, a Latinx humanities instructor, reported: \u0026ldquo;We structured reflection logs where students had to describe how they edited AI content. It wasn\u0026rsquo;t about right or wrong\u0026hellip; it was about their thinking. That\u0026rsquo;s when you saw growth.\u0026rdquo; Such strategies not only modeled reflective practice but also shifted learners\u0026rsquo; attention from product to process, promoting metacognitive internalization. This mirrored Vygotsky\u0026rsquo;s (1978) ZPD theory, where knowledge is co-constructed through guided participation with more capable others, including AI as a cultural artifact.\u003c/p\u003e\u003cp\u003eIn contrast, when ChatGPT was used without structured reflection or social mediation, many students exhibited cognitive offloading, bypassing planning and evaluation in favor of convenience. This uncritical use often occurred during independent or asynchronous tasks. Zoe, a Black female student with high digital fluency, admitted: \u0026ldquo;If I\u0026rsquo;m stuck, I just type it in and copy what looks good. I don\u0026rsquo;t always read it carefully.\u0026rdquo; Her comment reflected surface-level engagement. Rather than prompting metacognitive strategy use, ChatGPT functioned here as a cognitive shortcut, undermining the learner\u0026rsquo;s self-monitoring capacity.\u003c/p\u003e\u003cp\u003eDarius, an African American male student, echoed this sentiment: \u0026ldquo;I usually just go with the first thing it gives me, unless it sounds really weird.\u0026rdquo; These patterns aligned with Kim et al.\u0026rsquo;s (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) warning that AI tools can promote \u0026ldquo;passive alignment,\u0026rdquo; where learners abdicate responsibility for evaluating outputs, thus weakening metacognitive control.\u003c/p\u003e\u003cp\u003eObservational data confirmed these findings. In one STEM class, students pasted ChatGPT-generated code into a lab report without modification or explanation. The facilitator, Ms. Goods, noted: \u0026ldquo;They didn\u0026rsquo;t analyze or troubleshoot the code-they assumed it was correct because it came from ChatGPT.\u0026rdquo; This revealed a critical shift: instead of AI acting as a thinking partner, it became an authoritative source, reinforcing dependence and reducing productive struggle; an essential component of SRL. From an AIMD perspective, such use disrupted the strategic fading process. Rather than scaffolding learners toward independence, AI became an intellectual crutch. Without teacher mediation or peer dialogue, learners failed to activate reflection or revision, jeopardizing long-term cognitive autonomy.\u003c/p\u003e\u003cp\u003eAmidst these divergent usage patterns, students expressed a clear desire for agency in navigating their AI interactions. Casey voiced frustration with AI\u0026rsquo;s opacity: \u0026ldquo;I wish it told me why it said that. Like, how did it come up with that answer?\u0026rdquo; Her comment underscored the need for AI transparency; what AIMD identifies as a condition for metacognitive intentionality. When learners understand the logic behind AI responses, they are more likely to engage in strategic evaluation and adaptive learning.\u003c/p\u003e\u003cp\u003eAlex, an African American male with fluent digital literacy, reflected: \u0026ldquo;It looks cool when they use it right. Maybe I\u0026rsquo;d try it more if I knew how it works under the hood.\u0026rdquo; Alex\u0026rsquo;s curiosity pointed to a broader theme: digital fluency gaps constrained some students\u0026rsquo; ability to engage metacognitively. Without access to AI literacy instruction, students from underserved communities were at risk of using ChatGPT as consumers rather than co-designers of knowledge as echoed in Bhavana et al. (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eProgram manager Ms. Green articulated this challenge: \u0026ldquo;We had students with so much potential, but they didn\u0026rsquo;t know how to push ChatGPT to be more than a search engine. Equity in this work means teaching them to interrogate, not just receive.\u0026rdquo; This highlighted the critical interplay between student agency and institutional scaffolding. Even students who lacked initial fluency demonstrated reflective intent when given the tools. As Jordan, a semi-fluent African American student, put it: \u0026ldquo;At first I didn\u0026rsquo;t get it. But once we started comparing what ChatGPT said to what we found on our own, I started seeing where it was helpful and where it was off.\u0026rdquo; These instances of strategic evaluation reflected the beginnings of metacognitive co-regulation, here learners engaged AI as a dynamic thinking partner, consistent with AIMD\u0026rsquo;s model of internalization through scaffolded apprenticeship.\u003c/p\u003e\u003cp\u003ePeer interaction emerged as a powerful amplifier of metacognitive growth. Collaborative critique and discussion transformed ChatGPT from a static tool into a shared object of inquiry. Morghan, described this dynamic saying: \u0026ldquo;Sometimes I think ChatGPT gave me a good response, but when I shared it with my group, someone would ask, \u0026lsquo;Does that even make sense?\u0026rsquo; Then I\u0026rsquo;d go back and rethink it.\u0026rdquo; This cycle mirrored Vygotsky\u0026rsquo;s (1978) emphasis on social mediation, where higher-order thinking emerges through interaction. By externalizing their ideas and receiving critique, learners engaged in co-regulation, strengthening their metacognitive strategies.\u003c/p\u003e\u003cp\u003eEducators leveraged this through practices like \u0026ldquo;AI audit circles,\u0026rdquo; where students reviewed each other\u0026rsquo;s ChatGPT-assisted work. Mr. Griffiths, a white humanities instructor, explained: \u0026ldquo;When students debate which AI answer is stronger, they\u0026rsquo;re not just evaluating content\u0026hellip; they\u0026rsquo;re thinking about thinking.\u0026rdquo; These activities positioned ChatGPT as a dialogic stimulus rather than a solution engine. Students were not passive recipients but active co-constructors of knowledge, using AI as a catalyst for metacognitive discourse.\u003c/p\u003e\u003cp\u003eThe data revealed a dialectical tension within AIMD: ChatGPT\u0026rsquo;s metacognitive impact was neither inherently positive nor negative but contingent upon how it was used. This duality echoed Rane et al.\u0026rsquo;s (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) critique that AI often \u0026ldquo;mimics\u0026rdquo; personalization without promoting deeper learning unless integrated intentionally. In structured settings, ChatGPT modeled critical thinking, offered scaffolded prompts, and supported metacognitive dialogue. In unstructured contexts, it invited dependency, discouraged effort, and limited reflection. This mirrors SRL literature that emphasizes the importance of feedback loops, goal clarity, and iterative evaluation for metacognitive development (Zimmerman, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2002\u003c/span\u003e). Importantly, this finding challenges deterministic narratives of AI as either savior or threat. Instead, it positions educators, and their pedagogical choices, as the true mediators of metacognitive outcomes.\u003c/p\u003e\u003cp\u003eAcross multiple classrooms, a new pedagogical pattern emerged: metacognitive co-creation, where students and ChatGPT engaged in iterative dialogue, challenging assumptions and refining strategies. Rather than reflecting alone, learners used ChatGPT to simulate Socratic questioning, test hypotheses, and explore counterarguments. Jordan recounted:\u0026ldquo;I asked ChatGPT for a take on utilitarianism. Then I tried to argue against it. It was weird but helpful. It made me think deeper about my own view.\u0026rdquo; This reciprocal engagement aligned with AIMD\u0026rsquo;s model of AI as cognitive apprentice: modeling reasoning, provoking revision, and eventually fading into the background as learners internalized reflective routines.\u003c/p\u003e\u003cp\u003eThe co-creation of metacognitive space also extended to ethics. Students like Casey raised concerns about biased outputs, asking facilitators why ChatGPT seemed to \u0026ldquo;prefer certain kinds of examples.\u0026rdquo; These critiques not only reflected epistemic awareness but also digital criticality which is a metacognitive disposition essential in AI-mediated environments.\u003c/p\u003e\u003cp\u003eThe findings of this study reveal that ChatGPT plays a dual and dynamic role in students\u0026rsquo; metacognitive development, acting as both a scaffold that fosters strategic thinking and a crutch that may hinder reflective learning, depending on the context of its use. When embedded within structured, socially mediated, and equity-conscious learning environments, ChatGPT catalyzed metacognitive growth by prompting self-questioning, enabling peer dialogue, and supporting iterative evaluation.\u003c/p\u003e\u003cp\u003eConversely, in the absence of instructional scaffolding and digital literacy supports, students often engaged in cognitive offloading, demonstrating passive reliance on AI outputs. Central to these outcomes was the role of educators in designing intentional interactions that positioned AI not as a source of answers but as a thinking partner, and the agency of learners in leveraging AI to regulate, critique, and expand their own understanding. Ultimately, the study underscores that the impact of AI on metacognition is not intrinsic to the technology itself, but emerges from the pedagogical, social, and cultural ecologies in which it is situated.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study demonstrated that ChatGPT influenced students\u0026rsquo; metacognitive development in personalized learning environments in both facilitative and inhibitive ways. Using the AI-Mediated Metacognitive Development (AIMD) framework, the findings showed that ChatGPT supported metacognitive processes such as planning, monitoring, and evaluating, when paired with intentional instructional design and structured reflection. In these contexts, students used the tool to critique ideas, explore alternatives, and refine their thinking, aligning with Self-Regulated Learning theory.\u003c/p\u003e\u003cp\u003ePeer dialogue and collaborative critique further enhanced this engagement, transforming ChatGPT into a socially shared scaffold rather than an isolated digital tool. This reflected Social Constructivist principles, where metacognitive growth was co-constructed through interaction. When educators embedded prompts for critique and required justification of AI use, students demonstrated deeper cognitive control and self-awareness.\u003c/p\u003e\u003cp\u003eHowever, in unstructured or unsupported contexts, students often defaulted to accepting ChatGPT\u0026rsquo;s responses without reflection, revealing patterns of cognitive offloading. This inhibited metacognitive engagement, especially among students with limited digital fluency or prior experience in reflective learning. These patterns highlighted the risk of overdependence on AI, particularly when it was introduced without sufficient scaffolding or clarity of purpose.\u003c/p\u003e\u003cp\u003eThe findings also revealed inequities in metacognitive outcomes. Students from historically underserved backgrounds, despite showing interest and curiosity, faced greater challenges in using ChatGPT strategically. Gaps in access, AI literacy, and instructional support limited their opportunities to engage the tool reflectively, underscoring how inequitable integration could reinforce rather than disrupt educational disparities.\u003c/p\u003e\u003cp\u003eImportantly, the study identified the emergence of \u003cem\u003emetacognitive co-creation\u003c/em\u003e - a process in which students and AI collaboratively shaped learning through iterative dialogue. When students actively revised, questioned, and challenged ChatGPT outputs, they demonstrated distributed metacognitive regulation. However, this potential only emerged under conditions that supported inquiry, transparency, and learner agency.\u003c/p\u003e\u003cp\u003eOverall, the findings challenged the notion that AI inherently promotes personalization. Without deliberate guidance, AI use often remained superficial and performative. ChatGPT\u0026rsquo;s value lay not in generating answers but in prompting deeper cognitive effort, when educators positioned it as a tool for thinking, not a replacement for it. The study concluded that metacognitive development in AI-mediated environments depended less on technological sophistication and more on pedagogical intentionality and equity-driven design.\u003c/p\u003e\u003cdiv id=\"Sec26\" class=\"Section2\"\u003e\u003ch2\u003eImplications for Educators and Students\u003c/h2\u003e\u003cp\u003eThe findings highlight that the educational value of AI tools like ChatGPT depends entirely on how they are introduced, guided, and contextualized. For educators, this underscores the need to move beyond using AI for task completion and instead design learning experiences that position AI as a scaffold for metacognitive engagement. This includes embedding structured reflection prompts, peer critique protocols, and \u0026ldquo;AI audit\u0026rdquo; tasks that require students to explain, revise, or evaluate AI-generated content. Educators must also build students\u0026rsquo; AI literacy, helping them understand not only how to use these tools, but how to question their outputs, recognize their limitations, and maintain ownership of their thinking.\u003c/p\u003e\u003cp\u003eFor students, the challenge is to resist passive use and develop habits of inquiry, revision, and self-monitoring when interacting with AI. Metacognitive growth requires effortful engagement, and students must learn to treat AI not as a source of answers, but as a tool for enhancing their reasoning. Equipping students with the mindset and skills to think critically with AI, rather than through it, will be essential to building agency, adaptability, and reflective autonomy in AI-mediated learning environments.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec27\" class=\"Section2\"\u003e\u003ch2\u003eDirections for Future Research\u003c/h2\u003e\u003cp\u003eFuture research should explore how different instructional models influence students\u0026rsquo; long-term metacognitive growth in AI-integrated classrooms, particularly across diverse cultural and socioeconomic contexts. Longitudinal studies could examine whether structured AI use leads to sustained improvements in self-regulation, critical thinking, and learning transfer beyond a single course or program.\u003c/p\u003e\u003cp\u003eAdditional inquiry is needed into how AI tools can be adapted to support culturally responsive pedagogy, recognizing and valuing the diverse cognitive approaches students bring to learning. As AI continues to evolve, studies must address how learners\u0026rsquo; roles shift within these ecosystems and how educational institutions can support equitable access to meaningful, reflective, and empowering AI use.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec28\" class=\"Section2\"\u003e\u003ch2\u003eLimitations of the Study\u003c/h2\u003e\u003cp\u003eThis study has several limitations. First, it was conducted within a single apprenticeship program in the American Northeast, which may limit the generalizability of findings to other educational contexts. Second, the small sample size restricts the breadth of conclusions that can be drawn. Third, the study relied on self-reported data and researcher interpretation, which, despite triangulation and reflexive practices, may still carry subjectivity. Fourth, the exclusive focus on ChatGPT limits applicability to other AI tools with different functionalities or interfaces. Finally, the study examined short-term metacognitive outcomes and did not assess the long-term impact of AI use on sustained self-regulated learning.\u003c/p\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eDeclaration Statements The data that support the findings of this study are embedded in the study. The authors declare no funding associated with this study. Permission to conduct research with human subjects was granted by the Institutional Research Board of a University in the United States. Informed consents were obtained from all participants and all names are pseudonyms Consent to publish and participate was obtained from the legal guardians of the parents.\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflicts of interest associated with this study.\u003c/p\u003e\n\u003cp\u003eThe authors declare no funding associated with this study.\u003c/p\u003e\n\u003cp\u003eThe ethics board of a public University in the United States granted permission to conduct research with human subjects.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eInformed consent was obtained from all participants.\u003c/p\u003e\n\u003ch2\u003eConflict of Interest:\u003c/h2\u003e\n\u003cp\u003eNone\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eEKG wrote the manuscript solo\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAsy\u0026rsquo;ari, M., \u0026amp; Sharov, S. (2024). Transforming education with CHATGPT: Advancing Personalized Learning, accessibility, and ethical AI integration. \u003cem\u003eInternational Journal of Essential Competencies in Education\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(2), 119\u0026ndash;157. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.36312/ijece.v3i2.2424\u003c/span\u003e\u003cspan address=\"10.36312/ijece.v3i2.2424\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAtonte, B. F. (2025). Personalized AI-Driven Education: An AI Framework for Enhanced Learning Outcomes Among School-Age Children. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.36227/techrxiv.174561166.67234998/v1\u003c/span\u003e\u003cspan address=\"10.36227/techrxiv.174561166.67234998/v1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBernacki, M. L., Greene, M. J., \u0026amp; Lobczowski, N. G. (2021a). A systematic review of research on Personalized Learning: Personalized by whom, to what, how, and for what purpose(s)? \u003cem\u003eEducational Psychology Review\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(4), 1675\u0026ndash;1715. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-021-09615-8\u003c/span\u003e\u003cspan address=\"10.1007/s10648-021-09615-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBernacki, M. L., Greene, M. J., \u0026amp; Lobczowski, N. G. (2021b). A systematic review of research on Personalized Learning: Personalized by whom, to what, how, and for what purpose(s)? \u003cem\u003eEducational Psychology Review\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(4), 1675\u0026ndash;1715. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-021-09615-8\u003c/span\u003e\u003cspan address=\"10.1007/s10648-021-09615-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBhavana, S., Jayashree, K., \u0026amp; Rao, T. V. (2024). Navigating AI biases in education. Advances in Educational Marketing, Administration, and Leadership, 135\u0026ndash;160. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3693-5443-8.ch005\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3693-5443-8.ch005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChung, C. J. (2024). Preservice teachers\u0026rsquo; perceptions of AI in Education. AI-EDU Arxiv, 1\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.36851/ai-edu.vi0.4155\u003c/span\u003e\u003cspan address=\"10.36851/ai-edu.vi0.4155\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDahri, N. A., Yahaya, N., Al-Rahmi, W. M., Aldraiweesh, A., Alturki, U., Almutairy, S., Shutaleva, A., \u0026amp; Soomro, R. B. (2024). Extended Tam based acceptance of ai-powered chatgpt for supporting metacognitive self-regulated learning in education: A mixed-methods study. \u003cem\u003eHeliyon\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(8). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.heliyon.2024.e29317\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2024.e29317\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDavoodi, A. (2024). Equal ai: A framework for enhancing equity, quality, understanding and accessibility in liberal arts through AI for Multilingual Learners. \u003cem\u003eLanguage Technology and Social Media\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(2), 178\u0026ndash;203. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.70211/ltsm.v2i2.139\u003c/span\u003e\u003cspan address=\"10.70211/ltsm.v2i2.139\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEl Karafli, K. (2025). Empowering educators with AI. Advances in Computational Intelligence and Robotics, 369\u0026ndash;400. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3373-1827-1.ch014\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3373-1827-1.ch014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEunkyung, H. (2024). AI-based Edutech for adaptive teaching and learning. \u003cem\u003eArtificial Intelligence\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5772/intechopen.1004952\u003c/span\u003e\u003cspan address=\"10.5772/intechopen.1004952\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFan, L., Kim, J., Detrick, R., \u0026amp; Li, N. (2025). Developing students\u0026rsquo; higher-order thinking skills with Generative AI insights and strategies from a case study. \u003cem\u003eAdvances in Computational Intelligence and Robotics\u003c/em\u003e, 411\u0026ndash;428. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3693-7332-3.ch024\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3693-7332-3.ch024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGesser-Edelsburg, A., Hijazi, R., Eliyahu, E., \u0026amp; Tal, A. (2024). Bridging the divide: An empirical investigation of artificial intelligence and generative artificial intelligence integration across genders, disciplines and academic roles. \u003cem\u003eEuropean Journal of Open Distance and E-Learning\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(s1), 51\u0026ndash;69. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2478/eurodl-2024-0008\u003c/span\u003e\u003cspan address=\"10.2478/eurodl-2024-0008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHussain, S. R. (2025). The Personalization Paradox: When AI Stops Helping and Starts Overwhelming. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.5223317\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.5223317\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHutson, J., \u0026amp; Plate, D. (2023). Human-AI collaboration for Smart Education: Reframing applied learning to support metacognition. Advanced Virtual Assistants - A Window to the Virtual Future [Working Title]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5772/intechopen.1001832\u003c/span\u003e\u003cspan address=\"10.5772/intechopen.1001832\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKajiwara, Y., Matsuoka, A., \u0026amp; Shinbo, F. (2023). Machine learning role playing game: Instructional design of AI education for age-appropriate in K-12 and beyond. Computers and Education: Artificial Intelligence, 5, 100162. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.caeai.2023.100162\u003c/span\u003e\u003cspan address=\"10.1016/j.caeai.2023.100162\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKarlen, Y. (2016). Differences in students\u0026rsquo; metacognitive strategy knowledge, motivation, and strategy use: A typology of self-regulated learners. \u003cem\u003eThe Journal of Educational Research\u003c/em\u003e, \u003cem\u003e109\u003c/em\u003e(3), 253\u0026ndash;265. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/00220671.2014.942895\u003c/span\u003e\u003cspan address=\"10.1080/00220671.2014.942895\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKettler, T., \u0026amp; Taliaferro, C. (2022). The role of technology in personalized learning. \u003cem\u003ePersonalized Learning in Gifted Education\u003c/em\u003e, 216\u0026ndash;242. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4324/9781003237136-13\u003c/span\u003e\u003cspan address=\"10.4324/9781003237136-13\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim, J., Detrick, R., Yu, S., Song, Y., Bol, L., \u0026amp; Li, N. (2025). Socially shared regulation of learning and Artificial Intelligence: Opportunities to support socially shared regulation. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(9), 11483\u0026ndash;11521. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-024-13187-9\u003c/span\u003e\u003cspan address=\"10.1007/s10639-024-13187-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKomissarov, A. (2025). Artificial Intelligence in education: Challenges and opportunities. ARTIFICIAL INTELLIGENCE IN EDUCATION: CHALLENGES AND OPPORTUNITIES, 151. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.20319/ictel.2025.151\u003c/span\u003e\u003cspan address=\"10.20319/ictel.2025.151\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLaVaque-Manty, D., Meizlish, D., Silver, N., Kaplan, M., \u0026amp; Rhem, J. (2023). Using Reflection and Metacognition to Improve Student Learning. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4324/9781003448570\u003c/span\u003e\u003cspan address=\"10.4324/9781003448570\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee, H., Stinar, F., Zong, R., Valdiviejas, H., Wang, D., \u0026amp; Bosch, N. (2025). Learning behaviors mediate the effect of AI-powered support for metacognitive calibration on learning outcomes. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, 1\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3706598.3713960\u003c/span\u003e\u003cspan address=\"10.1145/3706598.3713960\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLevin, I., Marom, M., \u0026amp; Kojukhov, A. (2025a). Rethinking AI in education: Highlighting the metacognitive challenge. BRAIN. \u003cem\u003eBroad Research in Artificial Intelligence and Neuroscience\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(1 Sup1), 250. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.70594/brain/16.s1/21\u003c/span\u003e\u003cspan address=\"10.70594/brain/16.s1/21\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLevin, I., Marom, M., \u0026amp; Kojukhov, A. (2025b). Rethinking AI in education: Highlighting the metacognitive challenge. BRAIN. \u003cem\u003eBroad Research in Artificial Intelligence and Neuroscience\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(1 Sup1), 250. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.70594/brain/16.s1/21\u003c/span\u003e\u003cspan address=\"10.70594/brain/16.s1/21\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, W., \u0026amp; Brooks, C. (2024). Influence on judgements of learning given perceived AI annotations. Proceedings of the Eleventh ACM Conference on Learning @ Scale, 221\u0026ndash;231. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3657604.3662044\u003c/span\u003e\u003cspan address=\"10.1145/3657604.3662044\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLopez-Gazpio, I. (2025). Integrating large language models into accessible and Inclusive Education: Access Democratization and individualized learning enhancement supported by Generative Artificial Intelligence. \u003cem\u003eInformation\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(6), 473. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/info16060473\u003c/span\u003e\u003cspan address=\"10.3390/info16060473\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMahmoud, C. F., \u0026amp; S\u0026oslash;rensen, J. T. (2024). Artificial Intelligence in personalized learning with a focus on current developments and future prospects. \u003cem\u003eResearch and Advances in Education\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(8), 25\u0026ndash;31. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.56397/rae.2024.08.04\u003c/span\u003e\u003cspan address=\"10.56397/rae.2024.08.04\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarantika, J. E. (2021). Metacognitive ability and autonomous learning strategy in improving learning outcomes. \u003cem\u003eJournal of Education and Learning (EduLearn)\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(1), 88\u0026ndash;96. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.11591/edulearn.v15i1.17392\u003c/span\u003e\u003cspan address=\"10.11591/edulearn.v15i1.17392\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMelisa, R., Ashadi, A., Triastuti, A., Hidayati, S., Salido, A., Ero, P. E., Marlini, C., Zefrin, Z., \u0026amp; Fuad, A., Z (2025). Critical thinking in the age of AI: A systematic review of Ai\u0026rsquo;s effects on Higher Education. \u003cem\u003eEducational Process International Journal\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.22521/edupij.2025.14.31\u003c/span\u003e\u003cspan address=\"10.22521/edupij.2025.14.31\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMercer, N., \u0026amp; Howe, C. (2012). Explaining the dialogic processes of teaching and learning: The value and potential of sociocultural theory. \u003cem\u003eLearning Culture and Social Interaction\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(1), 12\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.lcsi.2012.03.001\u003c/span\u003e\u003cspan address=\"10.1016/j.lcsi.2012.03.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMergen, A., \u0026Ccedil;etin-Kılı\u0026ccedil;, N., \u0026amp; \u0026Ouml;zbilgin, M. F. (2025). Artificial Intelligence and bias towards marginalised groups: Theoretical roots and challenges. International Perspectives on Equality, Diversity and Inclusion, 17\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/s2051-233320250000012004\u003c/span\u003e\u003cspan address=\"10.1108/s2051-233320250000012004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMerkebu, J., Kitsantas, A., Durning, S. J., \u0026amp; Ma, T. (2023). What is metacognitive reflection? the moderating role of metacognition on emotional regulation and Reflection. \u003cem\u003eFrontiers in Education\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/feduc.2023.1166195\u003c/span\u003e\u003cspan address=\"10.3389/feduc.2023.1166195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePanarese, P., Grasso, M. M., \u0026amp; Solinas, C. (2025). \u003cem\u003eAlgorithmic bias, fairness, and inclusivity: A Multilevel Framework for justice-oriented AI\u003c/em\u003e. AI \u0026amp; SOCIETY. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-025-02451-2\u003c/span\u003e\u003cspan address=\"10.1007/s00146-025-02451-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRane, N., Choudhary, S., \u0026amp; Rane, J. (2023). Education 4.0 and 5.0: Integrating Artificial Intelligence (AI) for personalized and adaptive learning. \u003cem\u003eSSRN Electronic Journal\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.4638365\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.4638365\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRawat, P., Bajaj, M., Vats, S., \u0026amp; Sharma, V. (2024). Redefining human-AI interactions: Unveiling CHATGPT\u0026rsquo;s profound emotional understanding. A Practitioner\u0026rsquo;s Approach to Problem-Solving Using AI, 1\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2174/9789815305364124010003\u003c/span\u003e\u003cspan address=\"10.2174/9789815305364124010003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRegan, P. M., \u0026amp; Jesse, J. (2018). Ethical challenges of edtech, Big Data and personalized learning: Twenty-first century student sorting and tracking. \u003cem\u003eEthics and Information Technology\u003c/em\u003e, 21(3), 167\u0026ndash;179. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10676-018-9492-2\u003c/span\u003e\u003cspan address=\"10.1007/s10676-018-9492-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRizvi, I., Bose, C., \u0026amp; Tripathi, N. (2025). Transforming education: Adaptive learning, AI, and online platforms for Personalization. Technology for Societal Transformation, 45\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-981-96-1721-0_4\u003c/span\u003e\u003cspan address=\"10.1007/978-981-96-1721-0_4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoll, I., \u0026amp; Winne, P. H. (2015). Understanding, evaluating, and supporting self-regulated learning using learning analytics. \u003cem\u003eJournal of Learning Analytics\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 7\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18608/jla.2015.21.2\u003c/span\u003e\u003cspan address=\"10.18608/jla.2015.21.2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoozenbeek, J., \u0026amp; van der Linden, S. (2024). The Psychology of Misinformation. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/9781009214414\u003c/span\u003e\u003cspan address=\"10.1017/9781009214414\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSharma, D. (2024). Smart Education and sustainable learning environments in smart cities. Advances in Educational Technologies and Instructional Design, 403\u0026ndash;430. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3693-7723-9.ch024\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3693-7723-9.ch024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShekh-Abed, A. (2024). Metacognitive self-knowledge and cognitive skills in project-based learning of High School Electronics Students. \u003cem\u003eEuropean Journal of Engineering Education\u003c/em\u003e, 1\u0026ndash;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/03043797.2024.2374479\u003c/span\u003e\u003cspan address=\"10.1080/03043797.2024.2374479\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShumaila Farheen, A. A., \u0026amp; Cheema Rai Samee Ullah, \u0026amp; Dr. Marium Minhas Bandeali. (2025). Equity and bias in AI Educational Tools: A Critical Examination of algorithmic decision-making in classrooms. \u003cem\u003eThe Critical Review of Social Sciences Studies\u003c/em\u003e, 3(3), 67\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.59075/zqmnpa62\u003c/span\u003e\u003cspan address=\"10.59075/zqmnpa62\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTabib, F. M., \u0026amp; Alrabeei, M. M. (2024a). Can guided CHATGPT use enhance students\u0026rsquo; cognitive and metacognitive skills? Studies in Big Data, 143\u0026ndash;154. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-031-52280-2_10\u003c/span\u003e\u003cspan address=\"10.1007/978-3-031-52280-2_10\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTabib, F. M., \u0026amp; Alrabeei, M. M. (2024b). Can guided CHATGPT use enhance students\u0026rsquo; cognitive and metacognitive skills? Studies in Big Data, 143\u0026ndash;154. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-031-52280-2_10\u003c/span\u003e\u003cspan address=\"10.1007/978-3-031-52280-2_10\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUshioda, E. (2014). Motivation, autonomy and metacognition. \u003cem\u003eLanguage Learning \u0026amp; Language Teaching\u003c/em\u003e, 31\u0026ndash;49. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1075/lllt.40.02ush\u003c/span\u003e\u003cspan address=\"10.1075/lllt.40.02ush\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVenter, J., Coetzee, S. A., \u0026amp; Schmulian, A. (2024). Exploring the use of artificial intelligence (AI) in the delivery of effective feedback. \u003cem\u003eAssessment \u0026amp; Evaluation in Higher Education\u003c/em\u003e, \u003cem\u003e50\u003c/em\u003e(4), 516\u0026ndash;536. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/02602938.2024.2415649\u003c/span\u003e\u003cspan address=\"10.1080/02602938.2024.2415649\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eViola, L. (2023). The Humanities in the Digital: Beyond Critical Digital Humanities. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-031-16950-2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-031-16950-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVo, K., Sarkar, M., White, P. J., \u0026amp; Yuriev, E. (2025). Metacognitive problem solving: Exploration of students\u0026rsquo; perspectives through the lens of multi-dimensional engagement. \u003cem\u003eChemistry Education Research and Practice\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(1), 141\u0026ndash;157. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/d4rp00096j\u003c/span\u003e\u003cspan address=\"10.1039/d4rp00096j\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWalker, P. B., Haase, J. J., Mehalick, M. L., Steele, C. T., Russell, D. W., \u0026amp; Davidson, I. N. (2025). Harnessing metacognition for safe and responsible AI. \u003cem\u003eTechnologies\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(3), 107. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/technologies13030107\u003c/span\u003e\u003cspan address=\"10.3390/technologies13030107\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, V., \u0026amp; Design (2025). 231\u0026ndash;254. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3373-0502-8.ch010\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3373-0502-8.ch010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, W. S., Lin, C. J., Lee, H. Y., Huang, Y. M., \u0026amp; Wu, T. T. (2025). Enhancing self-regulated learning and higher-order thinking skills in virtual reality: The impact of CHATGPT-integrated feedback aids. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-025-13557-x\u003c/span\u003e\u003cspan address=\"10.1007/s10639-025-13557-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWinne, P. H. (2017). Theorizing and researching levels of processing in self-regulated learning. \u003cem\u003eBritish Journal of Educational Psychology\u003c/em\u003e, \u003cem\u003e88\u003c/em\u003e(1), 9\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/bjep.12173\u003c/span\u003e\u003cspan address=\"10.1111/bjep.12173\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, X., Duan, R., \u0026amp; Ni, J. (2024). Unveiling security, privacy, and ethical concerns of chatgpt. \u003cem\u003eJournal of Information and Intelligence\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(2), 102\u0026ndash;115. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jiixd.2023.10.007\u003c/span\u003e\u003cspan address=\"10.1016/j.jiixd.2023.10.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXie, X. (2023). Influence of ai-driven inquiry teaching on learning outcomes. \u003cem\u003eInternational Journal of Emerging Technologies in Learning (iJET)\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(23), 59\u0026ndash;70. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3991/ijet.v18i23.45473\u003c/span\u003e\u003cspan address=\"10.3991/ijet.v18i23.45473\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYamijala, S. M., Chodisetty, R. S., Chakravorty, C., \u0026amp; Sai, K. P. (2024). AI-Powered Learning Revolutionizing Smart Education with personalized learning styles. \u003cem\u003eAdvances in Educational Technologies and Instructional Design\u003c/em\u003e, 191\u0026ndash;212. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3693-8151-9.ch007\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3693-8151-9.ch007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, Y., \u0026amp; Xia, N. (2023). Enhancing students\u0026rsquo; metacognition via AI-Driven Educational Support Systems. \u003cem\u003eInternational Journal of Emerging Technologies in Learning (iJET)\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(24), 133\u0026ndash;148. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3991/ijet.v18i24.45647\u003c/span\u003e\u003cspan address=\"10.3991/ijet.v18i24.45647\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. \u003cem\u003eTheory Into Practice\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(2), 64\u0026ndash;70. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/s15430421tip4102_2\u003c/span\u003e\u003cspan address=\"10.1207/s15430421tip4102_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Metacognition, Self-Regulated Learning, ChatGPT, Personalized Learning, AI-Mediated Metacognitive Development (AIMD)","lastPublishedDoi":"10.21203/rs.3.rs-7339599/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7339599/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis qualitative study explores the role of ChatGPT in shaping students\u0026rsquo; metacognitive development and self-regulated learning within equity-driven, personalized learning environments. It draws on data from 24 semi-structured interviews, 300 hours of classroom observation, and analysis of student work collected from a yearlong youth apprenticeship program based in the northeastern United States. The program served culturally and socioeconomically diverse high school students and combined project-based learning with blended synchronous and asynchronous online instruction designed to promote career readiness and digital fluency. Grounded in Social Constructivism and Self-Regulated Learning (SRL) theory, one of the study\u0026rsquo;s key outcomes is the development of AI-Mediated Metacognitive Development (AIMD), a new theoretical framework that conceptualizes artificial intelligence as a dynamic cognitive mediator that can either support or hinder metacognitive growth. The findings reveal that ChatGPT fosters reflection, strategic thinking, and adaptive reasoning when implemented through structured scaffolding and peer collaboration. However, in the absence of intentional guidance, students often relied on ChatGPT in ways that led to cognitive offloading and superficial engagement. The study emphasizes that the educational value of AI depends not solely on its technological capabilities but on how it is embedded within pedagogical, cultural, and social contexts. AIMD provides a critical framework for designing AI-integrated instruction that advances learner agency, equity, and reflective autonomy.\u003c/p\u003e","manuscriptTitle":"ChatGPT as a Metacognitive Catalyst in Personalized Learning Ecosystems","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-17 15:29:01","doi":"10.21203/rs.3.rs-7339599/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b5d08496-8629-4526-9d0b-96e281ecaa47","owner":[],"postedDate":"September 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-11-17T01:23:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-17 15:29:01","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7339599","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7339599","identity":"rs-7339599","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}