Abstract:

The rapid spread of generative artificial intelligence is transforming how university students access information, solve problems and engage with academic tasks. In Computer Science degrees, these tools offer new opportunities for personalised support, but they also raise concerns about reduced cognitive effort, over-reliance and academic integrity. Understanding how students integrate AI into their day-to-day learning strategies is becoming increasingly relevant for higher education. This study presents an ongoing research project analysing the impact of generative AI on learning in an undergraduate Computer Science programme. The project adopts a mixed-methods, longitudinal design over two academic years, combining structured questionnaires, open-response prompts and anonymised academic data. Its goal is to examine how students use tools such as ChatGPT and Copilot to understand concepts, debug code, plan and monitor their study process and manage cognitive load. Quantitative and qualitative analyses, supported by clustering techniques, are used to identify distinct learner profiles and emerging patterns of AI-assisted study behaviour. During the conference, the project will present preliminary findings from the first semester of implementation. These early results draw on data from a compulsory subject positioned at an advanced stage of the degree. In Intelligent Systems (3rd year), students with greater academic experience carried out practical activities involving search algorithms and constraint satisfaction problems. These technically demanding tasks provide a suitable scenario in which to analyse how AI tools are incorporated into complex technical workflows and decision-making processes. Although the study is still ongoing, the preliminary results already reveal meaningful tendencies in students' AI-assisted learning behaviour.

Keywords:

Generative AI, Self-regulation, STEM education, Learning analytics, Student behaviour.