AI Code Generation : Trust & Risk Awareness Across Educational Levels

Sammanfattning

Generative Artificial Intelligence has experienced a boom of growth in both its capabilities and alsopopularity. These systems can be used for many different things like summarising text, explainingmathematical theories, and brainstorming to name a few. One use case is to use generative artificialintelligence to generate code. Using generative AI for code generation is very common amongstudents, and multiple factors influence how they use this technology. This study focuses on trust andrisk awareness because they play a big role in using automated systems effectively. Having too muchtrust in generative AI could lead to blindly accepting the generated result while too much distrustcould lead to an inability to identify correct recommendations. Additionally, having low-riskawareness may lead to problems such as accidental plagiarism or the introduction of security flawsinto a codebase. The aim of this paper is to investigate the differences in trust and risk awareness between universitystudents and upper secondary school pupils and to see if their current education level is influencingthese factors while using generative AI to generate source code. This paper shows the result of asurvey with 77 upper secondary school pupils at different schools in Sweden, and 33 students atBlekinge Tekniska Högskola. Additionally, 9 interviews were conducted among the two groupsfocusing on the participants’ risk awareness. The results showed a distinct difference in the amount of risks recognized by the university studentscompared to the upper secondary school pupils. Additionally, it was found that a lot of participants inboth groups did not recognize any risks at all. Therefore, further education about the limitations ofgenerative AI could potentially be integrated into lessons in school to help students use itappropriately. Trust was shown to be slightly higher among university students, however, the resultsfrom this study alone are not enough to prove that the difference in trust occurs due to their educationlevel.