AIEd Knowledge BaseAIEd KB
On this page

Concept

Self-Efficacy

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

Self-efficacy — a learner's belief in their capability to successfully perform a task or achieve a goal. Drawing on social cognitive theory (Bandura), self-efficacy shapes motivation, effort, persistence, and learning engagement. In AI in education, self-efficacy matters in two ways: AI tools can build learners' confidence and autonomy (e.g., by providing Feedback and Scaffolding), and learners' AI self-efficacy — their confidence in using AI technologies — influences how effectively they engage with AI, including how AI-related knowledge translates into career-relevant readiness.

Questions to Consider

  • Self-efficacy is a belief about your capability — distinct from actual competence. Have you ever been highly capable at something yet doubted yourself, or confidently wrong about something you couldn't do? What does that gap between belief and ability tell you about why self-efficacy matters?
  • AI self-efficacy (confidence working with AI) is a separate construct from AI literacy, and research finds literacy translates into readiness only when learners also have confidence. If someone knows about AI but doesn't believe they can use it, what happens to that knowledge — and what does that imply for training?
  • Research found that using AI to support understanding was fully mediated by academic self-efficacy in its link to performance, while shortcut use predicted worse outcomes partly independent of self-efficacy. Why would the same tool build confidence when used one way and fail to when used another?
  • The page distinguishes self-efficacy from the everyday word 'confidence.' Before you read, how are they different, and why would a researcher insist on the distinction rather than treating them as the same thing?
  • Robotics and embodied learning build confidence by grounding tasks in observable outcomes, and feedback can build learner self-efficacy. Think of a task where you gained real confidence only after seeing a concrete result. What does that say about what kinds of AI learning experiences are most likely to build — rather than merely report — self-efficacy?
  • Teacher self-efficacy affects adoption and integration of AI. If a teacher doesn't believe they can use AI effectively, does any amount of AI literacy fix it? What would build a teacher's confidence, and how is that different from giving them more information?

Introduction

Self-efficacy is distinct from actual competence: it is a belief about capability that drives behavior. It is closely related to — and often used interchangeably with — the everyday notion of confidence in one's abilities. Self-efficacy connects closely to Motivation, Self-Regulated Learning, and Student Experience. In the AI context, AI self-efficacy (confidence in working with AI) is a distinct construct from AI literacy, and research shows it plays a crucial role in whether learners actually activate and apply AI-related knowledge.

How self-efficacy appears in the knowledge base's research

  • Declines under GenAI-plus-XR studio work: in a 27-student architectural design studio, teams using a GenAI and multi-user XR pipeline showed larger relative pre–post declines in design self-efficacy confidence (β = −1.675, p = 0.021) and outcome expectancy (β = −2.088, p = 0.002) than teams continuing the normal workflow, with no significant difference in blinded panel ratings of their presentations (Xiao et al., 2026). Tool-rich environments can depress efficacy beliefs even when the work itself is judged equivalent.

  • AI self-efficacy and career readiness: Research on AI readiness shows that AI self-efficacy moderates the relationship between AI literacy and AI readiness: literacy translates into readiness only when learners have confidence in using AI, and self-efficacy directly predicts career adaptability.

  • Confidence is not always the lever. A cross-sectional SEM of 502 university faculty found technological self-efficacy a small predictor of AI Well-Being (β = 0.082) while AI literacy dominated (β = 0.713), bounding the claim that confidence gates AI outcomes: for well-being, knowledge outweighed confidence (Liu et al. (2026)).

  • Skills can enable dependency while efficacy buffers it. Among 478 Israeli undergraduates, the skill dimensions of AI literacy were positively associated with AI dependency (using/understanding AI β = 0.404) while AI self-efficacy was negatively associated (β = −0.132), so literacy training that builds only tool skill may increase reliance (Maizel et al. (2026)).

  • Low confidence routes into heavier AI use, then into lower scores. In a 153-student proof course, one point higher baseline self-efficacy predicted 39% fewer chatbot messages, and one point lower self-efficacy corresponded to 0.62 fewer exam points through chatbot use (Chen et al. (2026)).

  • Trust as the pivot between literacy and confidence: Hu (2026) surveyed 450 university students who already use ChatGPT and found an ordered chain rather than two parallel correlates: AI literacy related to Trust in the tool (beta = 0.50), trust to academic self-efficacy (0.48), and self-efficacy to continued use, with the serial indirect effect significant (0.07, 95% CI [0.04, 0.10]). AI anxiety weakened the literacy to trust link (interaction beta = -0.25, simple slopes falling from 0.76 to 0.25 across the anxiety range), so the same knowledge bought less confidence, and reached less use, among more anxious students. Confidence is thus a downstream link in the chain rather than a starting point.

The anxiety route through which AI-assisted practice relates to willingness can reverse rather than fade with confidence: the conditional indirect association fell from 0.142 at low speaking self-efficacy to 0.044 at the mean and −0.054 at high, an effect concentrated in a small upper tail (Wang & Li (2026)).

  • Profiles of academic self-efficacy and who reaches for AI: Suriá-Martínez et al. (2026) ran a latent profile analysis of academic self-efficacy among 102 university students with motor disabilities in Spain and found three profiles (low 29.4%, moderate 41.2%, high 29.4%) whose reported AI use rose stepwise with profile level (means 2.41, 3.56, 4.68; F(2, 99) = 27.84, p < .001, eta squared = .36). Excellence, the planning and goal-setting dimension, was the one most strongly associated with AI use (beta = .47). The authors read this as an equity problem: if confidence tracks with uptake, students with lower self-efficacy may be the least likely to reach for AI support that could reduce access barriers, so support for academic self-efficacy belongs inside inclusion frameworks.
  • AI use patterns and self-efficacy: Stamatoulis et al. (2026) found that using GenAI to support understanding (evaluative integration) was fully mediated by academic self-efficacy in its association with performance — understanding-oriented AI use builds confidence — whereas shortcut use (low-verification uptake) predicted worse outcomes partly independently of self-efficacy. Self-efficacy is thus both a pathway through which productive AI use helps and a factor that shortcut use may fail to build.
  • Attribution shapes whether efficacy grows. Crediting success to the AI rather than to one's own effort constrains self-efficacy growth and reinforces shortcut patterns, so sequencing complex activities into steps where students attribute progress to their own strategies matters more than the tool (Øby, 2026).
  • Robotics and hands-on learning: REMind's robot-mediated role-play built children's self-efficacy in anti-bullying intervention; robotics and embodied learning generally build confidence by grounding tasks in observable outcomes.
  • Creative self-efficacy in children: Niu et al. (2026) report, in a systematic scoping review of 22 primary studies of generative AI with children aged 6 to 15, that gains in creative self-efficacy cluster with divergent thinking and narrative creativity, mostly through text-to-image tools that lower the barrier between idea and artifact. They stop short of claiming efficacy: the designs are heterogeneous, the review performed no critical appraisal, and the corpus is nearly silent on disability, low-connectivity and underserved learners, so creative confidence is an outcome reported to rise rather than one the field has confirmed.
  • Teacher self-efficacy: Teacher AI competency research examines how professional development builds teachers' confidence in using AI, which affects adoption and integration. Ismael, Luo & Li (2026) add quasi-experimental evidence that an AI-supported e-mentoring model raises EFL pre-service teachers' self-efficacy and emotional intelligence during the practicum: the experimental group gained substantially more than controls on both measures, with a large between-group effect and gains across all self-efficacy subdomains — and the authors attribute them to mentoring AI-mediated within structured reflective cycles rather than to the AI alone.
  • Feedback and confidence: AI feedback and tutoring can build learner self-efficacy by providing actionable, supportive feedback.
  • Empowerment in AI Problem Solving: Zhu and Kong (2026) find that students' empowerment in using AI for problem solving mediates the relationship between perceived project-based learning and satisfaction with an AI literacy course. In their SEM analysis of 1,027 students, PBL fostered conditions that empowered students to use AI for problem solving, which in turn drove course satisfaction — evidence that building students' confidence and capability with AI is a key mechanism of effective AI literacy education.

Self-efficacy connects to Motivation, Self-Regulated Learning, Student Experience, AI Literacy, Learner Agency, and Robots in Education. Building self-efficacy is a key mechanism through which AI supports engagement and learning. Self-efficacy is measured almost entirely by self-report, so its associations with observed behavior deserve the usual caution.

  • AIGC self-efficacy as the pivot between tool and learning. Liang et al. (2026) find that perceived affordances of AI-generated content raise AIGC self-efficacy (beta = 0.583), which then mediates the paths to learning motivation (indirect effect 0.329) and to self-regulated learning (0.145), with the serial path affordance to self-efficacy to motivation to self-AI Regulation in Education also significant (0.173). The contrast that makes the finding useful is that the quality of AI Assessment feedback did not predict self-efficacy (beta = 0.131, n.s.) even though it strongly predicted satisfaction — confidence with the tool is built by directing it, not by receiving good output from it.

  • No treatment effect in a semester-long RCT — confidence rose with course time: in Thoeni and Fryer's (2026) randomized field experiment with 454 undergraduates in a Principles of Marketing course, self-efficacy rose over the semester for all students (time effect β = 0.099, p = 0.0009, Cohen's ƒ2 = 0.025). Neither the group effect (p = 0.1162) nor the group × time interaction (p = 0.5306, d = 0.023) was significant, so the RAG chatbot added nothing. Confidence gains in AI-supported courses can track course progression rather than the tool — an attribution risk for a construct the field measures almost entirely by self-report.

  • Confidence can move opposite to behavior. In competitive programming, students who stopped competing reported higher theoretical confidence than their active peers yet showed significantly weaker upsolving and peer-circle habits (p < .001), so a self-efficacy signal alone will mislead an early-warning system (Ruhan et al. (2026)).

  • Confidence can rise while the capability behind it declines. Among university students, frequent generative-AI use raised perceived confidence and efficiency and increased technological dependence in the same learners, so a self-efficacy score taken during AI-assisted work can read high at the moment the underlying skill is eroding (Yan et al., 2025).

  • AI use rode with confidence, but confidence did not predict attendance. In a PLS-SEM survey of 291 Portuguese students, general AI use was positively associated with self-efficacy for independent learning (β = 0.266, f² = 0.106). The construct was modeled as a second-order formative composite of Bandura's four sources — mastery experience, vicarious experience, verbal persuasion, and emotional state. Self-efficacy in turn did not predict absenteeism (β = −0.068, p = 0.276), even though AI use was separately associated with higher absenteeism (β = 0.148, f² = 0.029). The AI-use → self-efficacy path was significantly stronger among males than females (β = 0.449 vs. 0.159; Δβ = −0.289, p = 0.013), under partial measurement invariance (Franco et al. (2026)).

  • Confidence after a structured protocol tracked structure, not dose. Post-protocol confidence in solving similar problems independently averaged 7.18 on a 1–10 scale (σ = 1.62, n = 217), with institutional means ranging only 6.86 to 7.52 across four campuses. It tracked the perceived benefit of the AI explanation (r_s = 0.37) and of replicated practice (r_s = 0.35), more weakly prior AI confidence (r_s = 0.20), and not usage frequency (r_s = 0.02, p = 0.749) or self-reported GPA (r_s = 0.13, p = 0.064) (Paul et al., 2026). That is a mastery route — authored difficulty, then an independent attempt — rather than more time in the tool, a distinction the page lacked. The study measured no learning outcome, so it bounds what students believed.

  • Open-ended AI design tasks sort students by pre-existing confidence. Zangari et al. (2026) found that performance self-esteem correlated with extroversion (r = 0.262, p < 0.001) in a course where 345 students each built a conversational agent to teach a peer, and that female students reported significantly lower self-esteem than male peers (Kruskal-Wallis H(2) = 9.87, p < .005). The design inference the authors draw: open-ended, conversational creative tasks advantage the already-confident, so "open-ended design tasks [should] include sufficient scaffolding to equitably build performance self-esteem across all personality types" — a caution that student-created-artifact assignments can widen, not close, confidence gaps if scaffolding is not deliberate.

  • Self-efficacy interacts with format preference, not just with outcomes. A counterbalanced within-subject study (97 ninth-graders; content-matched tutor-based vs. paper-based practice of graph interpretation and construction) found tutor-based practice beat paper overall (t(62.30) = 2.38, p = .020), and — counterintuitively — lower baseline self-efficacy predicted greater gains in both formats (t(98.40) = −2.14, p = .035). The interesting structure is a significant three-way self-efficacy × favorability × format interaction (t(99.84) = −2.58, p = .011): among low-self-efficacy students, favorability toward the tutor predicts larger gains specifically in tutor practice. A sobering null for the assumption that tutor practice is motivationally self-reinforcing: ITS practice did not significantly improve post-training self-efficacy over paper — attributed to the short intervention (two ~40-minute sessions). The practical signal: baseline self-efficacy and technology favorability are cheap, actionable inputs for format assignment; tutor practice capitalizes on existing favorability rather than creating it. (Self-efficacy and favorability shape learning from tutoring systems and paper practice)

  • Knowledge reached intention through trust, anxiety and self-efficacy. In a PLS-SEM model of 526 design students, perceived AIGC knowledge was positively associated with trust, self-efficacy and usage intention and negatively with anxiety, and trust, anxiety and self-efficacy each carried a significant indirect path from knowledge to intention (Wei et al., 2026). Anxiety sat on the pathway rather than blocking it.

  • "Low writing confidence shows up as a request for help." It shows up as a request for prose. In an instrumented essay study, the Self-Efficacy for Writing Scale predicted total queries negatively (-0.258, SE 0.075, p < 0.001), and the association ran through Translating (-0.949) and Reviewing (-0.588) rather than Planning (-0.204) or whole-essay delegation (-0.016). Students who doubted their writing asked the model to produce and correct text; they did not ask it what to argue. Self-efficacy predicted reliance, and the reliance took a specific shape.

Confidence does not protect a learner from trusting a tool too much. Among 300 university students, Kim, Kim and Zhang (2026) found academic self-efficacy associated with AI overtrust (r = 0.230) and predicting it at B = 0.3901 (95% CI [0.2190, 0.5612]) in a serial mediation model, reversing the authors' hypothesis. The productive version of confidence is confidence about one's own judgment, which is what a high overtrust score says is missing.

  • Confidence reaches adoption partly through literacy. Belief to behavior: how AI self-efficacy shapes AI acceptance through AI literacy tested a sequence on 622 undergraduates in Türkiye. AI self-efficacy predicted AI acceptance (B = 0.43, R² = 0.29), and AI literacy carried part of that link (indirect B = 0.15, 95% CI [0.12, 0.19]) without absorbing it all. Confidence also has to be accounted for as a predictor of uptake, not only of capability. The design is cross-sectional with no demographic controls, so direction and mediated share are associations rather than effects.

  • Necessary for usefulness, not merely correlated with it. Chu and Liu (2026) placed AI self-efficacy in front of the two acceptance beliefs for 405 design graduate students and found it the strongest antecedent of both: β = 0.458 on perceived usefulness and β = 0.394 on perceived ease of use. Their necessary condition analysis adds a stronger form of claim than the path above. Efficacy was the only condition identified as necessary for perceived usefulness (d = 0.125, p = 0.015). Usefulness at 80%, 90% and 100% of its scale required efficacy at 10.0%, 30.0% and 50.0%. Confidence therefore sits as a floor under the instrumental judgment rather than as one contributor among several. The design is cross-sectional and the model explains R² = 0.150 of intention, so treat the thresholds as within-sample.

  • Behavioral traces can carry the signal, but the labels set the ceiling. A two-university pipeline detected low self-efficacy from interaction traces; survey-occasion labels beat a mid-term split, and the MSLQ subscale alone labeled as well as the full instrument (Somny, Chouati & Lefevre, 2026).

  • AI's career implications split into two routes. Among 415 Chinese college students, Zhang (2026) found positive perceptions reached career decision self-efficacy through career planning while negative perceptions reached it only through career anxiety (indirect β = −0.184).

Common Misconceptions

  • "Students who lack confidence avoid AI." The opposite pattern appears: in a 153-student proof course, one point higher baseline self-efficacy predicted 39% fewer chatbot messages. Low confidence routed into heavier AI use — the students most in need of practice were the ones most likely to offload it.
  • "More AI use with low confidence is a harmless coping strategy." The same finding links that routing to lower scores, so the coping pattern has a measurable cost rather than being neutral choice.
  • "Literacy alone builds confidence." Hu (2026) surveyed 450 university students who already use ChatGPT and finds trust as the pivot between literacy and confidence. The path runs through trust, so raising literacy without addressing trust leaves the mechanism untouched.
  • "Self-efficacy profiles are uniform in a disabled population." A latent profile analysis of 102 university students with motor disabilities in Spain found three profiles, with the low group at 29.4% and the remainder split across moderate and higher profiles. Treating the group as one population averages over distinct motivational states.

Connected Concepts

Connected Articles

Connected FAQs

Metadata

Intended audienceLearners
Learning and instructionMotivation, Self-Efficacy, Self-Regulated Learning
Assessment and measurementAssessment
Foundations of AI in educationAI Literacy
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.