What if effective AI tutoring started by understanding how students actually learn? This fundamental question drives Deliverable 5.1 of the HITS Project, which explores the analytical and strategic foundations of the HITS chatbot, demonstrating how understanding academic engagement patterns can lead to more meaningful and personalized support.
From Theory to Profiles: The Foundations of the Project
The HITS chatbot was not designed as a generic AI assistant. Its architecture follows a careful analysis of how students approach learning, manage tasks, and interact with institutional support.
A dual foundation: The framework combines pedagogical models like the COM-B model, digital nudging, and choice architecture with Retrieval-Augmented Generation (RAG) technology, bridging educational reasoning with institutional reliability.
Three dimensions of similarity: Student engagement is analyzed through cognitive (how learners approach tasks), behavioural (interaction rhythms and deadlines), and motivational (confidence and persistence) dimensions.
Four dynamic profiles: Four recurring profiles emerge from this analysis—Structured Achievers, Reactive Performers, At-Risk Students, and Self-Directed Learners—which serve as provisional, revisable interpretations rather than fixed labels.
The key message: Personalization should not mean putting students into rigid categories. It should mean understanding evolving needs and responding proportionately.
From Strategy to Action: Architecture, Ethics, and Validation
The second part of Deliverable 5.1 translates behavioral analysis into concrete tutoring strategies, defining the technical architecture, validation methods, and ethical safeguards of the system.
Targeted tutoring strategies: Depending on the profile, the system triggers specific levels of support, ranging from structured coaching to preventive intervention and human escalation when needed.
Technical design and validation: Similarity analysis directly shapes adaptive logic and hybrid routing, while ongoing validation tests technical reliability, fairness, and academic continuity.
Ethics by design: Data protection, consent, transparency, bias mitigation, and human oversight are integrated directly into development, while preserving institutional autonomy and European transferability.
The guiding principle behind the entire initiative remains clear: AI should support human tutoring, not replace it.