Developing Effective GenAI Users: The Missing Motivational System and Why It Matters for Equity

Synthesis: Trzesniewski, Gripshover and Master argue that research on Generative AI in education has narrowed to measuring how much students use it and what that does to outcomes, a framing that hides the Motivation driving students' choices in the first place. Their alternative is a motivational model of GenAI engagement: learning environments send cues about what and who is valued, students appraise those cues, appraisals make some goals salient, and goals make particular resource-use strategies rational. They distinguish four goals, learning, achievement (performance), completion (efficiency) and protection (avoidance), and four strategies, instrumental help, executive help, help avoidance and deliberately choosing another resource instead. Because GenAI fluently produces believable work products, achievement, completion and protection goals are all well served by asking it for answers, so only learning goals reliably predict use that enhances thinking. The equity argument follows directly: permitting practice for some students and not others builds the next Digital Divide, and their 2026 national adolescent data show 42% of students whose parents did not have a college degree attended schools where GenAI use was prohibited, against 17% of students with college-educated parents. The paper is a conceptual argument, and it calls for research that measures environment, appraisals, goals and strategies rather than amount of use.

Key Findings

1. The field is studying the wrong slice of the process. Hundreds of observational and experimental studies, including meta-analyses and meta-analyses of meta-analyses, evaluate the outcomes of different amounts of GenAI use (Emslander et al., 2026). The authors argue GenAI is not an intervention adults apply or withhold: it is already in students' lives, and students make use choices every day whether or not anyone scaffolds that decision-making.
2. Effective GenAI use is defined as capacity, not frequency. An effective user can know how, when and when not to use GenAI against where the technology is currently reliable and where it fails; match use or non-use to the goal of the current task; and use GenAI to enhance thinking, learning and work. The authors present this as a complement to existing AI Literacy frameworks (Feng & Carolus, 2026), not a replacement.
3. Knowledge alone does not determine use. A student can be taught to prompt well and still use GenAI to avoid thinking; can know output should be checked and still defer to it; can understand responsible use and still avoid the tool for fear of being accused of cheating. What differs, the authors argue, is the goals she is striving for and the contexts that make those goals salient.
4. Four goal types explain the choice. Learning goals, driven by safety, value and curiosity, seek understanding and competence. Achievement goals, driven by evaluative and high-stakes cues, seek to demonstrate competence and secure a grade. Completion goals, driven by high cost and low value, are transactional: get the assignment done. Protection goals, driven by low safety and low belonging, seek to hide a lack of understanding or avoid the risk of a mistake. Each is a rational response to the learning environment a student inhabits.
5. Four resource-use strategies follow from those goals. Instrumental Help-Seeking asks for limited assistance so the problem can still be solved independently; executive help asks for the answer, which the authors also call offloading or delegating thinking; some students avoid help entirely; and some purposefully choose not to use GenAI where a textbook, peer or their own notes serve the goal better.
6. Goals come from appraisals, not task features. The same assignment can be read as a challenge worth the effort or as proof of inability. Expectations for success, task value, and concerns about safety and belonging are situation-specific and malleable, and identity-relevant cues, such as computing being seen as a field where men belong and excel, can bring out threat appraisals and protection goals. The model is developmental because each choice accumulates into trajectories of skill, confidence, Learner Agency and belonging.
7. GenAI has three new affordances that matter for motivation. It requires an explicit teaching of resource use and instrumental help-seeking, which schools have rarely taught and which students do not acquire on their own (Aleven et al., 2003; Karabenick & Dembo, 2011); it participates in meaning-making as a new kind of authority, demanding a trust-but-verify stance; and its "jagged edge" of reliability shifts constantly, so no fixed playbook replaces calibrated experience.
8. Help-seeking skill is teachable, and that matters. Students given metacognitive Feedback on their help-seeking improved those skills, and the improvements carried into new material a month after the support was withdrawn (Roll et al., 2011). A proactive resource-planning prompt also improved undergraduates' academic performance before GenAI existed (Chen et al., 2017).
9. Access to practice is already unequal. In the authors' 2026 national adolescent data, 42% of students whose parents did not have a college degree attended schools where GenAI use was prohibited, compared with 17% of students with college-educated parents. The authors read permitting practice for some students and not others as how the next digital divide gets built (Capraro et al., 2024).

What is missing when students are taught to use GenAI

The paper's target is a gap between literacy instruction and motivation. Teaching students what Generative AI is, how to prompt it, how to evaluate its output and what ethical use requires is necessary but insufficient: those skills leave untouched the contexts, beliefs and goals that decide whether the tool gets used to think or to avoid thinking. The authors' sharpest illustration is that high-quality submitted work can now result either from a student working through ideas with the tool or from asking it to produce answers, which is why quality of work cannot serve as evidence of learning. Their proposal is to broaden the research question to what cultures, contexts and learning experiences give students the knowledge, skills, beliefs and motivation to become effective users of GenAI, and to do so equitably.

The framing has a corollary the authors state plainly: because within education the desired goal is learning, but not every environment motivates learning goals, studying effective GenAI use for learning requires studying use inside contexts that motivate learning goals in the first place. In environments that reward fast completion, delegating the work is not a failure of character but a rational strategy.

The motivational model the authors propose

The model is a domain-general decision-making sequence drawn from decades of research on learning, development and motivation (Walton & Wilson, 2018). Students appraise their situation, form goals, choose strategies, and accumulate knowledge and experience that shapes the next appraisal. Set against this sequence, the authors note, most AI-in-education studies occupy only the last slice, use and its consequences, often with access assigned rather than chosen, so students appear as products of GenAI rather than agents directing a resource.

• Situation: a person in context. The learning environment supplies cues about what and who is valued. Stereotypes about who has valuable abilities can undermine belonging; task features shape whether work looks valuable; students attuned to cues that their gender, racial or ethnic background, family immigrant status or socio-economic background may be devalued can feel they must work harder to prove themselves, or disengage if effort seems futile.
• Appraisals. Meaning-making processes that create a situation's motivational consequences for a given student: whether she can succeed, whether the task is worth the effort, whether it is safe to struggle or ask for help, and whether her own judgment matters.
• Goals. The four types above. The authors stress no goal is inherently good or problematic: adults routinely offload mundane tasks to preserve effort for more valuable work, and self-protection in a threatening environment may be critical to a student's Well-Being. What educators read as a student problem is often a mismatch between the educator's goal to assess learning and the goal the environment made rational for the student.
• Strategies. Instrumental versus executive help, help avoidance, and purposeful non-use of GenAI, including choices not to use it to learn independently, to avoid accusations of cheating, or because the costs (data-centre environmental costs, Privacy) are judged not worth it.
• Consequences that feed back. Each GenAI choice builds knowledge and experience that shapes the next appraisal, which is what makes the model developmental rather than a snapshot.

The authors separate processes that are unchanged from the pre-GenAI era from three affordances they consider new. First, using GenAI well means knowing how to elicit help that builds learning rather than answers, and monitoring one's own progress, at a moment when adults have less control over the availability of hints, answers and resources than before. Second, unlike earlier technologies GenAI contributes to the knowledge a student is building, introduces questions and offers pushback, so students need a trust-but-verify stance, and classrooms that value learning over getting the right answer build students' authority over their ideas. Faced with the question "Is my thinking worth contributing?", a student whose voice was never valued is unlikely to feel ownership. Third, effective use requires a working map of the tool's shifting reliability, which only structured tinkering can produce.

On teaching, the authors contrast a control response, high standards without support, or lowered demands and blocked tools, with a "mentor mindset" that holds standards high while making it safe to fail and learn from mistakes (Yeager et al., in press). Knowing GenAI output should be checked is not enough; students need to be shown their judgment is valued. They extend the same argument to teachers, who cannot support a judgment they have not had the chance to build.

Why the missing motivational system matters for equity

Every mechanism in the model is differentially available. Whether a school permits GenAI use at all, whether its assignments and grading practices make achievement or completion goals rational, whether struggle is safe, and whether students have been taught to seek instrumental help all vary by context, and those variations track existing inequalities. The authors report that students whose parents lack a college degree were more than twice as likely to attend schools that prohibited GenAI use, so the students with the least access to out-of-school coaching are also the least likely to get supervised, structured practice at school. Permission gaps of this kind are, in their words, how the next digital divide gets built (Capraro et al., 2024).

The equity risk is also motivational rather than merely material. Threat appraisals and protection goals are more likely in environments that signal some students are at risk of being judged negatively, and those goals in turn produce the avoidance strategies that foreclose learning. Because the model is recursive, small task-level inequalities compound into diverging trajectories of Self-Efficacy and Learner Agency. The authors' conclusion is deliberately symmetrical: for classrooms to make the rational choice and the learning choice the same choice for every student, the signals about what is valued have to change, not just the availability of the tool.

Implications for teaching and tool design

For teaching, the argument implies moving from policing use to building the capacity to choose. That means teaching Metacognition and Self-Regulated Learning explicitly, including instrumental Help-Seeking and progress monitoring; designing assignments and grading so that learning goals stay rational rather than making completion the efficient option; making it safe to struggle and make mistakes; and giving every student structured opportunities to test the tool against tasks they understand so they can build their own map of its reliability. For tool and environment design, the relevant levers are the cues a setting sends about what is valued: what assignments reward, how work is judged, and whether students' judgment is treated as worth using.

For research, the paper's methodological demand is measurement of the whole process, situation, appraisal, goal, strategy, consequence, rather than overall amount of use. The authors call for designs that deliberately vary the signals a learning environment sends, such as assignment and grading structures, and then examine how students' goals and GenAI behaviour shift, plus data collection that asks students why they made the choices they did. They also insist on comparability: measures of the full sequence, collected for different student populations, are what would let the field learn what supports effective GenAI use and scale it. Their closing framing is a rebuke to both popular narratives, GenAI as miracle tutor and GenAI as cognitive decay, and an argument that the science should set the pace rather than referee the tool.

Limitations: a conceptual argument

This is a position paper, not an empirical study. The motivational model of GenAI engagement is proposed to organise a research space; the links it asserts between environment, appraisal, goal, strategy and consequence are theoretical claims and predictions, several stated as such, rather than findings from the authors' own designs. The evidence it cites is largely borrowed from adjacent literatures on motivation, help-seeking, expectancy-value, social identity threat and metacognitive feedback, most of it predating GenAI, which supports the plausibility of the mechanism but not its specific operation with GenAI.

The empirical support the authors do present is thin and partly unpublished: two student quotations and a within-person pattern from an in-preparation 2026 study of how and why students use GenAI, and the 42% versus 17% school-prohibition figure from their own 2026 national adolescent data. No causal claims about GenAI and learning follow from that evidence. The paper also offers no instrument, no validated measures and no intervention, so the model's practical value depends on the measurement work it calls for, and the authors' own framing, that the basic motivational processes are largely unchanged from the pre-GenAI era, means the novelty of the account rests on how GenAI's affordances reshape them, which remains an open empirical question.

Connected Concepts

• Learner Agency — the ownership of judgment and epistemic authority the model treats as a developmental outcome
• Anxiety and Stress — threat appraisals and the protection goals that follow from unsafe learning environments
• AI Literacy — the knowledge-based framing of effective GenAI use that the paper argues is insufficient by itself
• Cognitive Offloading — executive help, delegating thinking, and the authors' refusal to treat it as inherently problematic
• Critical Thinking — what GenAI use enhances under learning goals and offloads under completion or protection goals
• Digital Divide — the access-and-permission gap the authors say unequal GenAI policy is building
• Equity — the paper's explicit equity question about who gets to learn effective use
• Help-Seeking — instrumental versus executive help, help avoidance, and the teachability of the skill
• Metacognition — the feedback and self-monitoring skills that make help-seeking instrumental
• Motivation — the appraisal-goal-strategy system the paper proposes as the missing piece
• Self-Efficacy — expectations for success that vary by situation and compound across choices
• Self-Regulated Learning — the strategic resource selection GenAI use is nested within

Connected Articles

• Why Put in This Much Effort?": How AI Availability Shapes Students’ Motivation in Introductory Programming — How the mere availability of AI reshapes students' motivation and effort in introductory programming
• Examining the Impact of Generative AI on Student Motivation and Engagement: The Mediating Role of Autonomy-Support and Autonomous Motivation in Education — GenAI and engagement, with autonomy support and autonomous motivation as mediators
• From disengaged to self-determined: a latent transition analysis of students' AI learning motivation — Latent transition analysis of students' AI learning motivation profiles
• How motivation and roles influence metacognitive engagement in student-GenAI interaction — Motivation and role framing in metacognitive engagement during student–GenAI interaction
• Understanding Student Dependency on AI: The Role of AI Literacy, Academic Self-Efficacy, and Resource Management Strategies — AI literacy, academic self-efficacy and resource management in student dependency on AI
• Efficiency vs. Effectiveness: Self-Regulated Learning with LLM-Mediated Help-Seeking — Self-regulated learning and LLM-mediated help-seeking, framed as efficiency versus effectiveness
• Metacognitive Training Facilitates Optimal Cognitive Offloading — Training the metacognitive judgment that decides whether offloading is optimal
• AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking — Cognitive offloading and critical thinking in everyday AI tool use
• Differential engagement with generative artificial intelligence in higher education: Gender, motivation, and achievement trajectories — Gender differences in GenAI engagement and achievement trajectories
• Decoding Divides: The Role of Socioeconomic Status and Personality Traits in AI Divides and Educational Inequality — Socio-economic divides in AI use and access at primary level

Citation

Trzesniewski, K., Gripshover, S., & Master, A. (2026). Developing Effective GenAI Users: The Missing Motivational System and Why It Matters for Equity. Preprint (manuscript submitted for review at Educational Researcher).