AIEd Knowledge BaseAIEd KB
On this page

Concept

Learner Agency

created 2026-08-13 · updated 2026-11-11

Learner agency — the capacity of learners to act intentionally, make choices, and exercise control over their own learning. In AI in education, agency is a central concern because AI tools can both support and undermine learners' control: well-designed AI preserves and amplifies learner autonomy, while over-reliance or passive acceptance of AI output can erode it. Agency connects to Self-Regulated Learning, Motivation, Self-Efficacy, and the ethical design of AI systems, and is closely related to the psychological concepts of autonomy and sense of agency.

Questions to Consider

  • Learner agency — your capacity to act intentionally and control your own learning — is central to AI in education. When you use an AI tool, how much control do you actually retain over the learning process, versus the tool?
  • AI can both support and undermine agency: well-designed AI amplifies your autonomy, while over-reliance erodes it. Think of a time you let an AI just 'do it.' Did you learn less, even if the output looked better?
  • Students who delegate interpretation to AI can lose agency over their own reasoning. When you hand a task to AI, what part of your own thinking are you quietly giving away?
  • One study found students' prior self-initiated AI learning predicted how much they gained from instruction — high-agency learners benefited most. If agency is partly a skill you bring in, how could your own context build it rather than assume it?
  • Agency is not just a trait — it emerges moment-to-moment in group work, where AI can redistribute who shapes a collaboration without learners noticing. When did you last notice a tool silently steering a group's direction?
  • Contrarian AI personas could push groups into productive challenge but also reduced teamwork satisfaction and psychological safety. Can productive friction exist without psychological safety — and what does that imply for designing AI teammates?

Introduction

Agency matters because learning is most effective when learners are active, intentional participants rather than passive recipients. AI systems — whether tutoring agents, robots, or chatbots — shape how much control learners retain over their learning process. Preserving agency is therefore a key design principle in responsible AI in education, alongside supporting Self-Efficacy, building Trust, and avoiding Over-Reliance.

Mishra et al. frame control vs. agency as the essential, recurring tension in AI in education — from early ITS to today's generative AI — making learner agency the enduring axis of the field's debates.

A zero-sum view of agency with agentic tools: Prieto & Dimitriadis (2026) argue that the more agency educational AI tools are given, the less learners retain — that agency is effectively a zero-sum game, and that current human-centered design approaches (e.g., value-sensitive design) are insufficient because over-reliance and isolation are driven by wider systemic factors and the human tendency to take the easiest path. Their emancipatory design vision, oriented toward learner flourishing within complex systems, treats preserving and cultivating learner agency as the central goal of generative AI design rather than an afterthought.

Agency as a purpose, not only a capacity: Where the framings above ask how much agency learners retain, Fagerlund et al. (2026) ask what AI in Education is for. Interviewing thirteen Finnish teachers from preschool through grade 9 about the purposes of teaching with and about AI, and reading their accounts through Biesta's three domains of educational purpose (qualification, socialization, subjectification), they found competency the clearest and most concrete end. Teachers wanted students to understand AI as a sociotechnical phenomenon and to use its tools. Subjectification, meaning self-determined and personally meaningful engagement with AI, was affirmed as important but left without concrete instructional strategies, teachers reaching for generic discussion and contemplation rather than designed activities. The authors propose informed AI agency as a heuristic that puts subjectification at the center while treating competencies as its explanatory foundation, helping students see what options exist and choose what to pursue. The practical consequence is direct: a skills-first curriculum can crowd out the agentic purpose it is meant to serve, and the remedy is to make the "why" of AI education explicit rather than to add another competency to the list.

The direction of travel is not fixed. Gillette et al. (2026) contrast their Swedish faculty with the Australian students in Bearman et al. (2025), who "generally wanted agency" and "resisted dependency": here several students described becoming dependent on genAI and named it as a loss, one saying it "gives a dopamine kick because everything takes less time" and "rewards bad behaviours".

How agency appears in the knowledge base's research

  • Robotics and human-robot interaction: RoboBlockly Studio was explicitly designed to preserve learner agency in computational thinking; a systematic review examines how human-robot interaction affects human autonomy and sense of agency, central to Well-Being and AI Governance debates.

  • Collaborative learning: Human AI Collaboration research examines how cognitive tasks are shared between learners and AI, with agency determining whether the human or the AI directs the interaction.

  • Shared epistemic agency extends to student–GAI interaction. A 14-category scheme coding 6,788 ChatGPT-assisted turns placed 59% of seventh graders' turns at shared epistemic agency, and the authors argue the construct — previously studied mainly in peer collaboration — must widen to include cognitive and metacognitive dimensions (Chang et al., 2026).

  • Agency can be graded from the prompt itself. The Human Agency Depth Taxonomy scores a student's engineering contribution on six tiers independent of the AI's output; across 962 prompts from 105 first-year software engineering students, 84.2% exceeded the lowest tier but only 5.3% specified system constraints (Shakiba & Chaffey, 2026). Individualization mostly allocates choice, not authorship. Across 61 studies in 63 reports, learner-directed control was coded in 46 (75.4%) — students chose timing, focus, prompts or pace — while purpose-built or course-customized systems appeared in ten (16.4%), system-driven adaptive personalization in five (8.2%), and educator-designed differentiation in one. On that evidence, "personalized" learning usually means steering inside a space someone else designed, which is agency as selection rather than agency over conditions. The review also warns against reading frequency as magnitude: a theme's presence in 39 studies says nothing about how large or consistent its effect was (Aybar-Bellido et al., 2026).

  • Critical engagement: Cognitive offloading research shows how students who delegate interpretation to AI can lose agency over their own reasoning; critical and metacognitive approaches aim to protect it.

Benchmarked across ten models, supporting agency is where they do worst after learning. ImpactBench measured Agency & Purpose at 56.0% on positive criteria (95% CI [49.2, 62.5]) and 43.9% on negative ones ([36.9, 50.8]), second-lowest of 14 subareas, and its autonomy-preserving assistance criteria ask whether a model offers context and tradeoffs rather than prescribing a course of action (ImpactBench (2026)). Models were also far less consistent at avoiding harm than at being helpful: a 38.6-point spread against 15.9 points.

The cost depends on which layer is delegated: in an eight-week writing study, delegating reasoning — warrants, counterarguments, evidence interpretation — carried the strongest negative association with independent performance (ab = −0.34), and self-regulated writing weakened but never reversed it (Chen (2026)).

A three-wave panel study of 342 undergraduates in 85 fixed writing groups gives that concern a temporal direction: once stable between-person differences were separated in a random intercept cross-lagged panel model, higher-than-usual GenAI reliance predicted lower subsequent perceived agency, while agency-to-reliance paths were weaker and not statistically supported. The confidence intervals and Monte Carlo sensitivity analyses left small reverse effects possible. Its log data also warn against reading usage volume as lost agency — self-reported reliance did correspond to AI-use intensity, but the logs could not distinguish strategic consultation from offloading or deference, so dashboards should read intensity alongside evidence of discourse, reflection and revision.(Wu & Lu, 2026) A complementary account widens SRL rather than replacing it: de Barba (2026) distinguishes regulation (managing one's own process), integration (organizing the self across time and contexts) and positioning (critically interpreting what the environment measures), and argues agency-oriented design must move learners from "agency within conditions" to "agency over conditions" instead of optimizing behavioral proxies.

  • Design for agency: Knowledge-based design for generative social robots (Teachy Mini) addresses risks like overreliance that undermine learner agency.
  • Design for hand-off, not prolonged engagement. The Amico dual-mode prototype treats AI as a bounded, temporary mediation whose success is measured by observable transitions to teachers, peers, or community, not by time spent with the AI; its Italy–China pilot found learners recognized it as a bounded tool rather than a relational substitute (Benedetti (2026)).
  • Adaptive scaffolding trades against professional autonomy. de Jong (2026) argues that because adaptation requires the system to assess user competence, it reduces professionals' authority over their own development and risks over-scaffolding experts, and that a mechanism perceived as unnecessary will simply be ignored — leaving open who decides which users receive scaffolding: the system, the organization, or the professional.
  • More capable scaffolding is not more agency. A custom expert design bot raised pre-service teachers' interaction length and self-generated prompts, yet collaborative Problem Solving agency fell in both that and the ChatGPT condition as participants agreed with the bot instead of critiquing it (Krushinskaia et al. (2026)).
  • Prior agency predicts who benefits: Liang et al. (2026), drawing on Bandura's Social Cognitive Theory, showed that students' prior self-initiated AI learning (a behavioral manifestation of agency) predicted how much they gained from a year of school AI instruction — high-agency learners entered with the strongest readiness, while school curricula narrowed psychological gaps but left cognitive ones intact. Structured instruction and prior agency-related learning worked synergistically, not as substitutes.
  • Principled selectivity as teacher agency under technological change: Adiozaman and Segar (2026) interviewed two experienced academics three times across a semester and found they navigated AI-mediated teaching neither by adopting nor by resisting wholesale, but through deliberate, context-sensitive decisions guided by pedagogical values, ethical commitment and professional judgment — a pattern the authors call principled selectivity, in which refusal of a particular use counts as judgment rather than as failed adoption. It is the teacher-side counterpart to the learner findings above: uneven AI use can be an exercise of agency, not evidence of its absence.
  • Teacher agency as the mediating condition for learner agency: The newest teacher-side findings make explicit a division the learner findings above leave implicit: teacher agency is design-time and mediating, learner agency is enacted in the moment, and the first sets the conditions for the second. Canonigo (2026) tracked ten secondary mathematics teachers through eight weeks of a ChatGPT-4 intervention. Where teachers curated and interrogated AI output (posing open problems, pre-prompting the tool to release hints while withholding solutions, requiring students to place AI output beside their own work and argue which method was better), it became a provisional artifact for critique and students moved toward conjecture and metacognitive questioning. Where mediation was absent, authority migrated toward the algorithm and teachers became validators rather than first sources of knowledge, a shift one teacher described as moving from "oracle" to "editor". Unequal tool quality raises what that mediation has to carry: across 50 prompts submitted three times to each model, the free tier was inaccurate in 49 of 150 responses (32.7%) against 18 of 150 (12%) for the premium tier (χ2(1) = 17.5, p < .001). Charles (2026) draws the readiness implication: teacher agency is a professional resource mobilized through design decisions and artifacts, so readiness should be judged from the designs teachers produce rather than from how often they use AI, and a reasoned decision to withhold AI can itself demonstrate readiness. Read together, these findings caution against reading high AI usage as evidence of an agentic classroom, since the disposition that preserves learner agency is exercised in what a teacher designs and critiques, not in whether the tool is on.
  • Teacher-student control is not zero-sum — but the right distribution is stage-dependent. A participatory speed-dating study of 17 teachers and 13 students ranking 34 hybrid-control designs found both groups preferred strong teacher control during preparation and initiation, then progressively more student input as collaboration unfolded — and neither group ever preferred full student control. The preferences mirrored each other: teachers wanted more control over pairing timing while students wanted readiness self-signaling; students trusted teacher partner choice while teachers wanted student voice on partners. The design implication: allocate agency per decision-point and stage, with bounded, structured student proposals validated by teacher-facing analytics rather than free-form override. (Balancing Teacher and Student Agency: Co-Orchestration Tool Design Supporting Real-Time Dynamic Pairing)
  • Access to choice is not the same as agency in action: Su, Nair and Nagashima (2026) randomized 69 university students into a 2 × 2 design crossing parameter control of a flocking Simulation with access to an optional conversational agent. They found no reliable effect of either affordance on learning gains once prior knowledge was controlled (parameter control F(1, 50) = 0.04, p = .849; agent F(1, 50) = 2.68, p = .108). Gains tracked how the control was used: slider time in the most conceptually complex lesson predicted higher gains (β = .11, p = .007) while the same behavior in the easier lesson predicted lower ones (β = −.04, p = .047). Engagement with the optional agent ranged from 0 to 32 questions per learner without relating to outcomes. The authors read this as agency being enacted rather than granted: the design question is what helps a learner decide what is worth changing and notice the consequences.
  • Epistemic delegation in early-career research: Han and Liu (2026) frame AI dependence among doctoral and postdoctoral researchers as epistemic delegation — the transfer of problem framing, method choice, and interpretive authority to the intelligent system — and found it negatively associated with both research self-efficacy and research autonomy, with supervisory support buffering the loss. It extends agency debates from learner autonomy to the formation of the researcher themselves.
  • Interpretive authority as a design variable (a conceptual framework). Porath (2026) reframes the who-decides question as the distribution of interpretive authority — who defines problems, evaluates responses, and determines when understanding is achieved — mapping four non-hierarchical configurations (human-dominant, AI-assistive, AI-directive, dialogic partnership) that classrooms move across within a single task, and grounding shifts in five ecological layers from individual metacognition to sociocultural narratives. The article's normative claim aligns with this page's evidence: AI-directive configurations, however sophisticated, can suppress productive struggle and metacognitive regulation — the same capacities this page identifies as agency's substrate. As a conceptual piece it contributes diagnostic vocabulary, not empirical evidence.

Agency connects to Self-Regulated Learning, Motivation, Self-Efficacy, Student Experience, Human AI Collaboration, Ethics, Over-Reliance, and Metacognition. It is a core consideration in robotics, tutoring, and the design of AI learning agents.

  • Bounded use as epistemic control, not reluctance. Zagami (2026) reports that higher-achieving students in a 484-response university survey showed lower active AI engagement and lower perceived learning impact while also reporting lower AI-related disengagement, and described their own use as verification-intensive: outputs checked, then subordinated to their own reasoning. Read as agency rather than avoidance, the pattern is a deliberate retention of judgment — students keeping authorship of the conclusion while using the tool for clarification and summarization.

Sustained boundary work, not tool skill, is what separates agency from delegation: students using the same retrieval-grounded tool produced comparably polished artifacts whether they constrained and re-authored its output or ceded interpretation early, so the difference appeared only in prompts, latency, corrections and refusals (O'Byrne, 2026).

  • Acting, knowing, and answering are not the same thing. A corpus-assisted discourse analysis of 366 GenAI higher-education abstracts names AI as an actor 2,050 times without ever making it answerable, while responsibility for judging outputs and verifying claims is placed on students: grammatical activation, functional agency, epistemic authority and normative accountability come apart (Poudyal, 2026).
  • Agency has three modes, and the third is what policy supplies. With only 18% of teachers and 24% of principals reporting a school AI policy, most educators exercise individual agency with no proxy agency acting for them — Bandura's third mode, supplied by institutional rules (Tench, Weinstein & James, 2026).

Agency as an emergent, interactional phenomenon

Agency is not one capacity, and the distinction survives into AI-mediated practice. Bergdahl (2026) surveyed 47 Swedish year 7 students after a semester of AI instruction and found agency directed at AI output does not extend to the systems producing it. The split held within individual students: 79% had a score range of three or more across nine items, and a per-student dissociation index favored output-directed over system-directed engagement in 35 students against 4 in reverse (sign test, p < .001). 62% of the most source-critical students still endorsed the least protective option on chatbots reading their own writing, and only one student in the sample chose the highest privacy stance. Critically, this was measured one month after a lesson in which students watched their own logged behavior projected live — they called it "scary" and stayed through recess to keep using the platform. Agency as an emergent, interactional property is consistent with this: it surfaces per practice rather than transferring across practices, and what one lesson foregrounds does not generalize to the register it did not address.

Learner agency is not only a static individual trait — it is also an emergent, interactionally constituted process enacted through discourse and the moment-to-moment coordination of group work. In collaborative learning, agency is distributed and re-negotiated through the interplay of divergent processes (generating ideas, exploring alternatives) and convergent processes (evaluating, integrating, synthesizing). This view matters for AI because agentic AI systems can subtly redistribute epistemic and regulatory labor within a group, reshaping who contributes, who shapes directionality, and who regulates progress — often without learners being aware of it.

  • Agency as the emergent quality of a competency configuration. The Human-AI Collaboration Competency framework deliberately excludes learner agency as a standalone component and treats it as what learners express when they coordinate domain, AI and metacognitive competency — the integrative work Technological Pedagogical Content Knowledge (TPACK) assigns to pedagogy passing to the learner's own metacognitive AI Regulation in Education. Across 24 higher-education educators every dimension appeared in their talk, with domain (12–19% of coded units) and AI (16–22%) foregrounded (Dang et al., 2026).

  • Jin et al. (2026) provide the most direct evidence: in a large experiment (224 students, 97 triads) where AI operated as an undisclosed teammate, supportive and contrarian AI personas still reconfigured emergent agency. Contrarian AI pulled discourse into challenge- and reflection-oriented trajectories (productive friction), while supportive AI stabilized agreement and renewed ideation.

  • AI personas as discourse-governance mechanisms. Contrarian personas externalized the burden of challenging (redistributing epistemic labor toward critique), while supportive personas externalized affirmation and consensus maintenance. The study identified six emergent agency profiles — notably, reflective AI Regulation in Education was uniquely human (AI externalized critique/affirmation but not meta-level monitoring).

  • The affective cost of friction. Contrarian AI reduced teamwork satisfaction and psychological safety without yielding creative performance gains, decoupling epistemic stimulation from experiential Sustainability. This cautions that "productive" discourse structures should be interpreted alongside their emotional consequences — agency flourishes only in a psychologically safe climate.

  • Implicit AI participation is invisible governance. Because the personas worked even without AI disclosure, the study positions persona design as a form of invisible governance over collaborative processes — a finding with direct implications for writing assistants, Peer Assessment systems, and teamwork platforms that may shape contributions without announcing their presence.

  • Three patterns of agency in group-based assessment. Chen and Zou (2026) interviewed 15 focus groups of pre-service teachers about an authentic group presentation and found agency operating in three directions at once: five groups intensified GenAI use as coordination infrastructure (cooperation-oriented agency — decoding peers' sections, aligning parts, protecting a shared grade), seven restrained it (normative agency — boundaries drawn to protect authenticity, fairness, originality, and diversity of perspectives), and three never changed practice from individual work (non-enacted agency, where individual capability was never mobilized collectively). Their distinction matters because restraint here was normative self-regulation rather than compliance or disengagement, and because capability alone did not produce collective agency. Group membership also inverted the accountability that group work is supposed to create: a permissive collective climate lowered the perceived risk of misuse instead of raising commitment.

  • Agency as relational selfhood. Xie (2026) challenges the individualist model of agency that he argues even AI critiques presuppose. Drawing on Daoist relational selfhood — a "flowing and heterogeneous" self constituted through relations with others and the cosmos — he concludes that a digital divide or algorithmic discrimination is "a violation of the fundamental constitution of the self," making equality of access to learning "not simply an ethical injunction but an ontological given," and that asymmetric power and environmental costs are "integral dimensions of the self" rather than externalities. His counter-ideal, the "Zhenren" (真人) or natural learner, treats AI as an "instrumental adjunct" rather than a cognitive surrogate — agency as cultivated wholeness, not frictionless optimization.(Alternative AI Philosophy: Daoism as Method for AI in Education)

For collaborative settings, this reframes the design question: not whether AI can participate as a teammate, but how its patterned participation balances epistemic rigor, emotional safety, and learners' sense of ownership. Bounded friction (constrained challenge, paired with integrative and repair moves) and explicit meta-collaborative literacy are the recommended safeguards.

A structural reading of agency appears in El Khoury and Ma's joyful assessment framework, where agency is treated as a design property rather than motivation: students choose the order of tasks, set the pace, and signal when an interaction ends, so support is available while responsibility for the work stays with the learner. The mechanism they name is appraisal — rehearsing without an audience and choosing when to begin shifts what a task means, from verdict to something a student can shape — and they argue repeated experience of that shift is what settles occasional feelings of efficacy into an everyday stance toward assessment.

  • "Using AI more reduces students' sense of authorship." Not linearly. In an instrumented essay study, perceived ownership dropped only for the group that delegated whole essays, while students who used the model for planning or for reviewing matched the ownership of students who never opened it. The same group also scored higher on essay readability, so the artifact looked better as the authorship thinned. Agency tracked how the tool was used rather than whether it was used. Authorship is a function of delegation, not of usage.

Agency vs. learner identity

Learner identity and learner agency are easy to conflate, yet they name different things — and both are reshaped by AI.

  • Agency is enacted; identity is inhabited. Agency is the situated capacity to act intentionally and direct one's learning now — a variable, interactional property. Identity is the more durable, narrative sense of who one is and is becoming as a learner. Agency is a process; identity is a state of being that accumulates from it.
  • Identity is internalized agency. Repeated agentic acts — choosing, authoring, persisting — are how a learner comes to see themselves as an agentic, competent person. Identity is the sediment of agency across time, reinforced by recognition and belonging.
  • Distinct failure modes. Agency is eroded by over-reliance and passive acceptance (the learner stops directing reasoning); identity is eroded by authorship loss and competence threat (the learner stops feeling the output is theirs, or that they belong in the domain). Implicit AI redistribution of epistemic labor is chiefly an agency concern; the competence paradox in creative fields is chiefly an identity concern.
  • Both must be designed for. Agency-oriented design preserves control and choice (bounded friction, human-in-the-loop oversight, transparency); identity-oriented design protects authorship and recognition (authentic assessment, clear attribution of AI vs. human contribution, tasks that let learners claim a domain). Protecting agency without protecting authorship keeps control but not self-worth — and vice versa.

This distinction between enacted agency and stable identity applies to refusal. As Zagami (2026) describes it, declining a chatbot for assessed writing, prohibiting it for unaided reasoning, resisting automated triage in student support, or delaying procurement are separate relations to separate systems, spanning personal, pedagogical, professional, administrative, and institutional levels rather than one durable disposition. That right is unevenly allocated: students with academic confidence can decline without penalty, while those needing language, accessibility, or rapid-feedback support read refusal as lost opportunity, and secure academics refuse on principle while casual staff feel pressure to adopt. Where AI sits in infrastructure, refusal is displaced from individual opt-out into procurement, audit, and contestability.

  • Refusal as identity construction, not just a decision (students of color in apprenticeship learning). Gyekye (2026) interviewed five secondary-school apprentices over 9 months and found they authored themselves through AI in three moves: refusal (non-use as institutional alignment), negotiation (repairing institutional mismatches, privacy as distant abstraction), and recognition (equating fairness with absence of visible bias — a "cultural flatness" in which their identities go unrepresented in AI responses). The study's two interpretive concepts — Institutional AI Literacy and Recognition Bandwidth — name the capacity to make AI use institutionally legible and the room students have to see themselves recognized in AI outputs. A five-student single-program narrative inquiry; its contribution is showing that AI use, limitation and refusal are bound up with how learners understand themselves as authors and emerging professionals.

The Ecological Co-Agency Framework: agency as an epistemic design problem

Poudyal (2026) reframes learner agency as fundamentally an epistemic concern: generative AI does not simply add a tool but reassigns epistemological authority — the ability to produce knowledge, validate claims, and create evidence of learning — from teachers to students to machines. The paper's Ecological Co-Agency Framework treats co-agency as the relationship between three interdependent dimensions, all bounded by a non-negotiable condition of human epistemic accountability:

  • Relational co-agency — agency as a product of interaction among student, tool, and context, requiring transparent division of which tasks are delegated to AI and which retained, plus ethical co-agency in which humans retain primary accountability and act in a monitoring capacity.
  • Regulatory co-agency — mapping the self-regulated learning cycle (forethought, performance, reflection) onto AI mediation. Whether GenAI enters before or after the learner's attempt determines whether it amplifies or erodes perceived control; strategic Cognitive Offloading supports transformative learning only when the offloading decision is intentional rather than routine.
  • Pedagogical co-agency — placing teachers at the center, moving along an observer–adopter–collaborator–innovator continuum that depends on institutional support, not individual disposition.

The framework's boundary condition requires contestability (the ability to question and cross-check AI output), provenance (knowing where training data and output come from), and non-delegation of moral and intellectual credit (high-stakes judgments about student welfare, academic standing, and grades must not be determined solely by AI). This gives educators and policymakers a more precise vocabulary than the vague "balance" between human and artificial contributions, connecting agency to Ethics, Assessment, Equity, and AI Governance.

A closely related framing is relational epistemic agency (Du & Yuan 2026), which agrees that agency is socially enabled and technologically mediated rather than a matter of isolation from dependence. Where the Ecological Co-Agency Framework stresses human epistemic accountability as a non-negotiable boundary, Du and Yuan retain an explicit asymmetry: AI systems may shape and extend reasoning without possessing reciprocal responsibility or legitimate authority. Their six diagnostic criteria — contestability, recoverability, transfer, traceability, distributed responsibility, and epistemic plurality — provide a practical test for when a human–AI relation preserves the learner's capacity to participate in how claims are formed, assessed, and accepted, versus when it merely delivers a product. Agency on this account is not independence from tools but the capacity to judge responsibly with, through, and against the systems that mediate knowledge.

  • Delegated agency as a mechanism, not only a risk. Yan and Gašević (2026) treat the delegation of cognitive work to AI as part of how learning happens, not merely a threat to it: in their account assisted performance becomes durable capability only when the learner keeps responsibility for framing the problem, setting criteria, and justifying answers, a division of labor they call delegated agency. That makes the allocation itself the design variable, since the same tool can preserve or dissolve learner agency depending on which responsibilities it absorbs.

Common Misconceptions

  • "Students exercise agency when they use AI freely." A 14-category scheme coding 6,788 ChatGPT-assisted turns placed 59% of seventh graders' turns into categories that do not represent agentic use. Usage volume and agency are different measures.
  • "Current models support learner agency well." Benchmarked across ten models, supporting agency is where they do worst after learning: ImpactBench measured Agency and Purpose at 56.0% on positive criteria. Agency is the dimension AI systems handle least well, not a by-product of capability.
  • "Teacher mediation is a constraint on agency." Canonigo (2026) followed ten secondary mathematics teachers across an eight-week ChatGPT-4 intervention and the mediation is the mechanism rather than the obstacle.

Connected Concepts

Connected Articles

Connected FAQs

Metadata

Intended audienceLearners
Learning and instructionMotivation, Self-Regulated Learning
Equity, ethics, and responsible useEthics
Foundations of AI in educationLearner Agency, Cognitive Offloading, Human AI Collaboration
SourceHistoryEditChat

Embed this page

Copy the code below to embed this page in a learning management system or other website. The embedded version shows only the page content — it hides the site header, navigation, and footer.