The ethical perception of AI-generated assignments among Hungarian informatics students

Kusper, Gábor and Zaletnyik, Péter Tibor (2026) The ethical perception of AI-generated assignments among Hungarian informatics students. In: Agria Média 2025 : „Oktatás 5.0: Az ember és a technológia harmóniája”. Eszterházy Károly Katolikus Egyetem Digitális Technológia Intézet, Eger, pp. 123-128. ISBN 9789634963400

Preview

Text 123Kusper.pdf - Published Version Download (921kB) | Preview

Abstract

Generative AI has rapidly entered student workflows, yet ethical norms around assignment submission remain unsettled. To explore these perceptions in a local context, we conducted a survey among Computer Science and Business Informatics students at a Hungarian university. This study is embedded in the broader research project “We Are Not Afraid of the Wolf!”, which examines Hungarian informatics students’ attitudes toward generative AI in programming education. The project investigates how students balance efficiency gains with concerns about reliability, skill development, and labor market impact, and it has shown that most students see AI as a tool rather than a competitor, while stronger programmers tend to review and debug AI-generated code more critically. We studied the following hypothesis: Students with stronger programming skills are less likely to view submitting AIgenerated assignments as ethical than students with weaker skills. We collected responses from 71 students. Programming proficiency was proxied by the Programming 2 lecture grade (on a 2–5 scale). Ethical acceptance was measured on Likert scales by three items: (1) acceptance of using AI-generated code in assignments; (2) acceptance of submitting exclusively AI-generated code as one’s own; and (3) importance of disclosing AI-generated code (interpreted inversely as lower ethical acceptance when disclosure is deemed more important). Spearman correlations were applied in this exploratory stage without additional data cleaning. Ethical acceptance of AI use in assignments showed a small, non-significant negative association with proficiency (r = –0.15, p = 0.24). Acceptance of submitting exclusively AI-generated code was moderately and significantly lower among stronger programmers (r = –0.27, p = 0.032). The disclosure item trended oppositely and was not significant (r = +0.20, p = 0.12). So, we can say, that stronger programming skills are linked to stricter ethical judgments against “pure AI” submissions, while views on mixed use and disclosure are more diverse. Although our study is limited to one Hungarian institution, the full anonymized dataset is openly available at Zenodo, enabling verification and further exploration of our findings by other researchers. As part of the “We Are Not Afraid of the Wolf!” project, these findings enrich a multi-instrument evidence base on how students negotiate ethics, learning, and employability in the age of generative AI.

Actions (login required)

View Item