Intellectus: an AI model to enhance student cognitive performance, increase knowledge acquisition, boost academic integrity and mitigate AI over-reliance in education

This study presents Intellectus, a fine-tuned large language model (LLM) designed to enhance student cognitive performance, boost knowledge acquisition and mitigate AI over-reliance in education. In contrast to traditional AI tools that provide direct answers, Intellectus engages learners through Socratic questioning, adaptive scaffolding, and structured prompting. Developed using a no-code Custom GPT approach and grounded in a systematic literature review, Intellectus was evaluated through a triangulated methodology combining student self-assessments, expert ratings, and plagiarism analysis. A/B testing with twelve university students showed significant gains across ten higher-order cognitive skills when using Intellectus compared to ChatGPT, alongside a dramatic reduction in plagiarism rates. Expert evaluations further validated these findings, confirming Intellectus' ability to promote deeper thinking and academic integrity, and T-testing proved statistical significance across domains.