Mastery experiences are Bandura's term for direct, personal encounters with genuinely challenging tasks that yield to sustained effort. They are the strongest source of self-efficacy Bandura identified — stronger than observing others succeed (vicarious experiences), stronger than being told one can succeed (social persuasion), and stronger than the physiological state of confidence. The distinguishing feature is that the individual's effort produces an outcome she can trace back to her own actions. The tracing is essential: self-efficacy is deposited in proportion to the individual's belief that the outcome was caused by her own actions, not by luck, external assistance, or tools that did the work for her. Twenge's framework treats mastery experiences as the primary developmental currency that AI's architecture disrupts — not because AI prevents outcomes, but because outcomes produced through AI cannot be traced to personal effort with the specificity that self-efficacy requires.
The developmental power of mastery experiences lies in their cumulative structure. A single experience deposits a specific, domain-limited layer of self-efficacy: the child who assembles a model airplane develops confidence in following complex instructions and persisting through manual frustration. That specific confidence transfers, partially and imperfectly, to adjacent tasks. Over years of accumulated mastery experiences across many domains, the layers form what Bandura called a foundation of competence — the psychological bedrock on which adult capability rests. The foundation is not a single belief but a structure of interconnected domain-specific beliefs that together constitute the individual's working sense of her own capacity.
AI's disruption operates through traceability. The student who produces an AI-generated essay receives an output — often superior to what she could produce alone — but cannot trace the output back to her own effort in the way mastery experience requires. She may have prompted effectively, edited lightly, curated outputs. These contributions are real but thin compared to the experience of constructing an argument from scratch, wrestling with organization, producing drafts she recognizes as inadequate and revising them. The AI-generated essay does not deposit a layer of self-efficacy for essay-writing. It deposits something else — perhaps a layer of self-efficacy for AI-directing — but the two are not equivalent, and the student who substitutes one for the other develops different cognitive capacities than the student who completes the mastery cycle independently.
The implication for institutional design is sharp. AI integration in educational contexts that preserves mastery experience structure — assignments where students produce independent work before refining it with AI assistance, where their contribution is substantial and traceable, where the final output is recognizably improvement on their own effort — can capture AI's genuine benefits while preserving developmental function. AI integration that bypasses the mastery structure — prompt-and-submit workflows, assignments redesigned to assume AI generation from the start — eliminates the developmental function even as it improves measurable outputs. The distinction is not between using AI or not using AI. It is between integration structures that preserve traceability and those that destroy it.
Bandura's identification of mastery experiences as the primary source of self-efficacy emerged from his research on behavior change in clinical contexts — particularly phobia treatment, where he observed that patients who directly confronted feared stimuli and succeeded showed greater subsequent improvement than those who merely observed others or were verbally encouraged. The finding, developed in Self-Efficacy: The Exercise of Control (1997), was extended by subsequent research across educational, organizational, and developmental contexts to establish mastery experiences as one of psychology's most empirically robust findings.
Primary source of self-efficacy. Mastery experiences are stronger than vicarious, persuasive, or physiological sources — they are the bedrock on which the others build.
Traceability is the active ingredient. The deposit of self-efficacy requires that the individual attribute the outcome to her own actions with specificity.
Domain-specific accumulation. Mastery experiences build domain-specific confidence that transfers partially to adjacent domains — not generalized self-esteem.
AI disrupts through traceability erosion. Outputs produced through AI cannot be traced to personal effort with the specificity mastery experience requires, even when the outputs are genuinely superior.
Integration structure determines developmental function. AI use that preserves traceability (students produce independent work before AI refinement) captures benefits while preserving mastery; AI use that eliminates traceability (prompt-and-submit) captures benefits while destroying mastery.