Between perceived competence and artificial intelligence: academic self-efficacy profiles in university students with motor disabilities

Synthesis: Suriá-Martínez, García-Castillo, López-Sánchez, and García del Castillo (2026) asked whether perceived academic competence tracks with how university students with motor disabilities use generative AI. In a cross-sectional survey of 102 students at two universities in Alicante, Spain, a latent profile analysis of the Attention, Communication, and Excellence dimensions of academic self-efficacy returned three profiles: low (29.4%), moderate (41.2%), and high (29.4%). Reported AI use rose with profile level across educational, informational, and emotional support, and a structural model linked all three dimensions to AI use, with Excellence strongest (β = .47). The authors read the pattern through Bandura's social cognitive theory and the technology acceptance model: students who already believe they can manage academic demands are the ones who reach for AI support. Because the design is cross-sectional, the association is not causal; the authors position AI as a complement to, not a substitute for, structural change toward inclusive, accessible universities.

Key Findings

• Three academic self-efficacy profiles emerged: low (30 students, 29.4%), moderate (42, 41.2%), and high (30, 29.4%), defined by Attention, Communication, and Excellence.
• Reported AI use followed the profiles: means of 2.41, 3.56, and 4.68, F(2, 99) = 27.84, p < .001, η² = .36, a large effect.
• The same gradient held for informational support (η² = .28) and emotional support (η² = .18): every AI support type rose with profile level.
• In the structural model, Excellence was most strongly associated with AI use (β = .47), then Communication (β = .31) and Attention (β = .22); R² = .43.
• The sample was 102 higher education students with motor disabilities aged 18 to 33 (M = 22.56) at two Alicante universities.
• The authors warn that cross-sectional data cannot show whether self-efficacy shapes AI use, whether AI use shapes self-efficacy, or both.

Study Design and Measures

This is a quantitative, cross-sectional survey run in early 2026 at the University of Alicante and the Miguel Hernández University of Elche. The sample comprised 102 higher education students with motor disabilities aged 18 to 33 (M = 22.56, SD = 4.89); 52.9% were female. All held an official disability certificate, with 46% rated below 33%, 39% between 33% and 65%, and 15% above 65%. Academic self-efficacy was measured with the 13-item Academic Behaviors Self-Efficacy Scale (0 to 10) covering Attention, Communication, and Excellence. Perceived AI utility was measured with a purpose-built 12-item questionnaire on a 5-point scale covering educational, informational, and emotional support; the three-factor structure was checked against a separate sample of 85 students, and the authors treat that evidence as preliminary.

Profiles and Patterns of AI Use

Three profiles fit best in the latent profile analysis. Profile 1, low academic self-efficacy, held 30 students (29.4%); Profile 2, moderate, held 42 (41.2%); Profile 3, high, held 30 (29.4%). A four-profile solution had slightly better fit indices but no significant improvement and a small class size (12.7%), so the authors retained three, reporting entropy of .89. Self-reported AI use tracked the profiles closely: M = 2.41 (SD = 0.78) for the low profile, 3.56 (0.81) for moderate, and 4.68 (0.74) for high, F(2, 99) = 27.84, p < .001, η² = .36. The same gradient appeared for informational support (4.51 vs 3.74 vs 2.83; F = 18.92, η² = .28) and emotional support (3.96 vs 3.31 vs 2.76; F = 10.47, η² = .18).

Why Excellence, and Why Inclusion

In the structural model, all three dimensions were positively associated with AI use: Excellence was strongest (β = .47, p < .001), then Communication (β = .31, p < .01) and Attention (β = .22, p < .05), together accounting for 43% of the variance in reported AI use (R² = .43). Excellence covers planning, goal setting, and achievement orientation, and the authors place it at the center: students who already steer their own academic goals may deploy AI deliberately as part of self-regulated study rather than as a crutch. The pattern fits the technology acceptance model.

The authors argue this matters for inclusion and social sustainability. If perceived competence goes with reaching for AI support, students with lower self-efficacy may be least likely to use tools that could reduce access barriers, turning technology uptake into an equity question. They call for interventions that build confidence through mastery experiences, vicarious learning, and feedback, with AI inside university inclusion frameworks rather than in place of structural remedies. Limits: the cross-sectional design rules out causal claims, the sample of 102 constrains power, only motor disabilities are covered, and campus and device factors were not modeled.

What this means for practice

• The authors draw several implications for inclusive higher education.
• Because academic confidence and AI use move together, they argue that support for academic Self-Efficacy should be a programmatic target: vicarious learning, progressive mastery experiences, and positive feedback are the mechanisms they name, aimed particularly at students in the low and moderate profiles.
• Their worry is access: if students with lower confidence are the least likely to reach for AI tools, simply making tools available will not close gaps that tailored scaffolding could.
• Their second point is structural.
• AI should sit inside university inclusion frameworks as a pedagogical component rather than a peripheral add-on, and faculty need professional development in digital competence, AI literacy, and Universal Design for Learning so that guidance accompanies the tools. They also recommend differentiated strategies by profile, with scaffolded support for lower-confidence students and peer-mentorship roles for higher-confidence ones. Finally, they ask universities to judge their AI strategies by equity indicators such as inclusion, accessibility, and autonomy, not only by performance or usage metrics, and to watch for dependency that can accompany excessive use.

Limitations

• The authors are explicit that the cross-sectional design rules out causal claims. The positive link between academic self-efficacy and reported AI use could run in either direction, or both: higher confidence may lead students toward AI, AI use may shape how students judge their own competence, or the two may reinforce each other over time.
• The structural model reports associations only. Sample size is the second caveat.
• With N = 102, the authors write that statistical power, stability, and generalizability of the structural estimates are limited, and the separate confirmatory factor analysis behind the AI utility questionnaire rested on only 85 students, so the authors treat that factor structure as preliminary and call for replication in larger samples. AI use was self-reported, which may overestimate actual use or invite social desirability bias, and usage practices change quickly.
• Generalizability is bounded: the sample covers only students with motor disabilities, and device access, prior digital training, platform accessibility, and campus support infrastructure were left out of the model. The authors ask for ecological frameworks and objective usage data in future work.

Connected Concepts

• Self-Efficacy — the profile variable, measured through Attention, Communication, and Excellence
• Latent Profile Analysis — the person-centered method used to derive the three profiles
• Inclusive Learning and Accessibility — the policy frame the authors apply to AI in higher education
• Technology Adoption Models — the lens used to explain why competence tracks with AI use
• Equity and Differential Effects Across Learner Groups — the risk that support tools reach those who need them least

Connected Articles

• The dark side of AI in education: AI dependency as a mediator linking academic self-efficacy and teacher support to learning burnout among university students — academic self-efficacy and AI dependency in higher education
• Assistive Generative AI for Visually Impaired Learners: Personalization and Inclusion in Higher Education — assistive generative AI for visually impaired learners
• Generative AI, virtual reality, and beyond: A scoping review of digital assistive technologies for neurodivergent students in higher education — digital assistive technologies for students with disabilities

Citation

Suriá-Martínez, R., García-Castillo, F., López-Sánchez, C., & García del Castillo, J. A. (2026). Between perceived competence and artificial intelligence: academic self-efficacy profiles in university students with motor disabilities.