STUDENTS’ PERCEPTIONS AND RESPONSIBLE USE OF GENERATIVE AI IN PHYSICS PROBLEM SOLVING


Akı F. N.

ULUSLARARASI YAPAY ZEKÂ VE DİSİPLİNLERARASI UYGULAMALAR KONGRESİ, İstanbul, Turkey, 21 - 22 October 2026, pp.105-110, (Summary Text)

  • Publication Type: Conference Paper / Summary Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.105-110
  • İstanbul Ticaret University Affiliated: Yes

Abstract

The increasing use of generative artificial intelligence (AI) in education has raised important questions about how students perceive and use these tools, particularly in physics, where conceptual reasoning and problem-solving skills are essential. This study investigates engineering students’ perceptions of generative AI in physics problem solving, their trust in and critical evaluation of AI-generated solutions, perceived effects on learning and motivation, and their intention to continue using AI. The study involved 86 engineering students from three engineering programs. Data were collected using a 10-item, five-point Likert-type questionnaire addressing the perceived usefulness of generative AI for understanding and solving physics problems, trust in AI-generated explanations, the ability to identify errors in AI-generated solutions, physics self-confidence, the tendency to obtain answers directly from AI, conceptual understanding, verification of AI-generated answers, motivation, and future use intentions. Items were coded from 1 (“strongly agree”) to 5 (“strongly disagree”).

The findings indicate that students generally perceive generative AI as a useful tool for understanding and solving physics problems. However, their trust in AI-generated physics explanations and their perceived ability to identify errors in AI-generated solutions were comparatively moderate. A substantial proportion of students agreed that AI-generated answers should be verified. Students also reported a relatively low tendency to obtain answers directly from AI without first attempting to solve problems themselves. The 10-item instrument demonstrated high internal consistency (Cronbach’s α = .905). However, the item measuring the tendency to obtain answers directly from AI (Q6) showed a relatively weak corrected item-total correlation. When Q6 was treated as a separate behavioral indicator, the remaining nine-item scale demonstrated stronger internal consistency (α = .922). The nine-item structure showed adequate sampling adequacy (KMO = .868), and Bartlett’s test of sphericity was statistically significant, χ²(36) = 603.27, p < .001. Exploratory factor analysis and parallel analysis supported a dominant single-factor structure, with the factor accounting for approximately 58.1% of the total variance. No statistically significant differences were observed across gender, academic year, or department in the present sample.

Overall, the findings suggest that engineering students view generative AI as a useful support tool for learning and solving physics problems while recognizing the importance of verifying AI-generated information. The results indicate that the educational use of generative AI in physics should extend beyond answer generation and emphasize students’ physical reasoning, solution verification, and critical evaluation of AI-generated content.

Keywords: Generative artificial intelligence, physics education, problem solving, engineering students, AI literacy, critical evaluation