Abstract

This thesis investigates the conceptual design and prototyping of an AI-driven chatbot intended to support university students in discovering academically relevant reading materials. The proposed system leverages a multi-agent architecture in which four specialized agents, the User Agent, Recommendation Agent, Content Retrieval Agent, and Interface Agent collaborate to deliver personalized, context-aware academic recommendations through natural language interaction. The research employs an exploratory, qualitative methodology consisting of semi-structured interviews with nine university students. Thematic analysis of the interview data reveals that participants experience significant challenges in navigating academic databases, with time consumption (86%), relevance assessment (86%), and credibility verification (71%) identified as the most prevalent difficulties. All participants actively use AI tools most prominently ChatGPT (100%) and Google Scholar (86%) yet a majority (71%) have never received personalized reading recommendations from their institution. The proposed agent-based architecture addresses these needs by combining retrieval-augmented generation with hybrid recommendation algorithms and adaptive user profiling. The findings contribute both a conceptual design framework and empirical evidence for informing the future development of AI-supported academic learning environments. …víceméně