Cognitive dissonance in programming education: A qualitative exploration of the impact of generative AI on application-directed learning

Abstract

Generative AI tools, powered by Large Language Models (LLMs), are already being extensively used by students to support their learning and it is important that educators understand what this might mean for higher education practice. In this study, two researchers external to the faculty teaching team conducted in-depth interviews with 12 students in a small European university of applied sciences who have recently undertaken programming learning as part of their undergraduate studies. The aim was to explore how these students were using LLMs to support their learning and their perceptions of its value as a learning tool. A thematic analysis of the resulting qualitative data revealed trends in the perceived advantages and disadvantages of using LLMs, as well as different levels of LLM usage, with more cautious use associated with a 'meaning-directed' approach to learning (learning pattern) and more enthusiastic and unrestrained use with 'application-directed' patterns of study. A tension was observed between some application-directed learners’ high use of LLMs and their recognition that this is not optimal for effective learning. The authors argue that Cognitive Dissonance Theory (CDT) can explain how this dissonance may motivate learners toward a dissonance-reducing attitude or behavior change. The conclusion reflects on the implications for teaching practice and offers some recommendations for how educators can increase metacognition, instrumentalize CDT to increase self-regulation, and facilitate meaning-directed learning patterns in the age of generative AI.