AI confidence shapes perceived gains in critical thinking and creativity
Updated Aug 24, 2026
A positive answer about GenAI may tell you more about student confidence than about what the technology actually improves. That distinction matters when universities use AI pulse surveys to shape guidance, teaching, and support. Atiyah A. Alghamdi's Innovations in Education and Teaching International paper, "University students’ perceptions of generative AI for critical thinking and creativity: The influence of self-efficacy and disciplinary differences", shows how strongly students' views can vary with confidence and academic discipline. For UK teams building on student experiences of GenAI across universities, the message is practical: ask what students believe AI adds, but do not mistake perceived value for demonstrated learning gain.
Context and research question
Universities need evidence on whether GenAI supports critical thinking and creativity, weakens them, or does both in different contexts. A single question such as "Has AI helped your learning?" cannot separate confidence, discipline, task, prior experience, or concern about dependency. Recent work on AI attitude surveys that separate usefulness, self-expression, and concern points to the same measurement problem.
Alghamdi investigated four questions: how students perceive GenAI's effect on critical thinking, how they perceive its effect on creativity, whether those views differ between STEM and non-STEM disciplines, and how AI self-efficacy relates to both. The quantitative, cross-sectional study surveyed 405 first-year undergraduates at King Saud University in Saudi Arabia. Of the sample, 298 students, or 73.6%, studied STEM subjects, while 107 studied non-STEM subjects. Participants completed five-point scales covering critical thinking, creativity, and confidence in using AI.
That design gives institutions useful evidence about student perceptions, but it does not test whether students became more critical or creative. For UK Student Experience, Digital Education, and Market Insights teams, this distinction is essential because perceptions can shape uptake and trust without proving an effect on learning.
Key findings
Students were moderately positive about GenAI, not overwhelmingly convinced. The mean score for perceived support of critical thinking was 3.63 out of 5, while the creativity score was 3.32. Students saw value in analysing information, synthesising sources, generating ideas, and approaching problems in different ways, but the mid-range scores point to cautious approval rather than certainty.
The paper's abstract captures that measured tone:
"students hold moderately positive views of Generative AI’s impact on critical thinking and creativity"
Discipline mattered for perceived critical thinking. STEM students reported a higher mean score than non-STEM students, 3.70 compared with 3.43, and the difference was statistically significant. The paper suggests that GenAI may fit more visibly with the structured problem-solving and computational tasks common in STEM. Creativity perceptions, by contrast, did not differ significantly across the two disciplinary groups. This is a warning against rolling every subject into one institutional AI score.
AI self-efficacy was closely associated with perceived educational value. Confidence in using AI correlated with perceived support for critical thinking at r = .604 and accounted for 36.4% of the variance in that perception. The relationship was stronger for creativity, where self-efficacy correlated at r = .807 and accounted for 65.1% of the variance. Students who felt more capable with AI were much more likely to believe it supported their thinking and creative work.
Confidence may therefore shape the answer before an institution has measured the outcome. A student who knows how to prompt, challenge, and refine an AI response may report more value than a student who has had little guidance. That does not mean confidence caused stronger critical thinking or creativity. It means survey teams should interpret positive AI ratings alongside experience, training, and discipline, much as evidence on student agency and critical reflection in AI-supported learning separates tool use from who retains judgement.
The study's limits matter. It draws on self-reported perceptions from first-year students at one Saudi university, with STEM students forming nearly three quarters of the sample. Its cross-sectional design cannot establish causality, and reported confidence or benefit may not match actual behaviour or performance. The findings are best used to improve local questions and hypotheses, not as proof that GenAI develops higher-order skills.
Practical implications
For UK universities, the first implication is to measure AI confidence separately from perceived benefit. Ask students how confident they feel checking outputs, identifying errors, refining prompts, and deciding when not to use AI. Then ask what effect they think AI has on analysis, originality, and learning. Separating those constructs gives teams a clearer basis for deciding whether the next step is skills support, policy clarification, or course redesign.
Second, analyse AI-related student voice by discipline and learning context. STEM and non-STEM averages differed in this study, but a university may find more useful variation at school, programme, or task level. Pair a small set of stable survey items with open prompts such as "When does AI help you think more carefully?" and "When does it make your work feel less original?" This follows the broader case for measuring adaptation, critique, and reflection rather than tool use alone and gives faculties evidence they can apply to their own teaching.
Third, use comments to explain the score. Student Voice Analytics can group open responses about confidence, critical thinking, creativity, dependency, and unclear expectations, then compare patterns across disciplines and cohorts. This helps institutions distinguish a capability gap from a trust problem or a mismatch between AI guidance and disciplinary practice, giving Digital Education and Student Experience teams a more precise starting point for action.
Finally, do not use positive perception data as a proxy for learning gain. If an institution wants to know whether GenAI improves critical thinking or creativity, it also needs behavioural or performance evidence, ideally collected over time. Student surveys remain valuable because confidence and belief influence adoption, but triangulation protects teams from making stronger claims than the evidence supports.
FAQ
Q: How should a university apply these findings to an AI student survey?
A: Use separate items for confidence, perceived usefulness, critical evaluation, creativity, dependency, and policy clarity. Add one or two open questions that ask when AI helps and when it gets in the way. Break results down by discipline, prior experience, and year of study before deciding whether the main need is training, clearer expectations, or changes to assessment and teaching.
Q: Does the study show that greater AI confidence improves critical thinking and creativity?
A: No. It shows a strong association between self-efficacy and students' perceptions of GenAI's value. The study did not measure actual changes in critical thinking or creativity, and its cross-sectional design cannot show cause and effect. A university would need performance measures, behavioural evidence, and longitudinal data to test whether confidence leads to stronger learning outcomes.
Q: What does this mean for student voice on AI more broadly?
A: Universities should stop treating AI opinion as a single spectrum from support to resistance. Students may value GenAI because they feel skilled with it, distrust it because expectations are unclear, or judge it differently depending on the intellectual work their discipline values. Open comments make those reasons visible, which helps institutions respond to the experience behind the rating rather than the rating alone.
References
[Paper Source]: Atiyah A. Alghamdi "University students’ perceptions of generative AI for critical thinking and creativity: The influence of self-efficacy and disciplinary differences" DOI: 10.1080/14703297.2025.2600476
Request a walkthrough
Book a free Student Voice Analytics demo
See all-comment coverage, sector benchmarks, and reporting designed for OfS quality and NSS requirements.
- All-comment coverage with HE-tuned taxonomy and sentiment.
- Versioned outputs with TEF-ready reporting.
- Benchmarks and BI-ready exports for boards and Senate.
UK-hosted · No public LLM APIs · Same-day turnaround