2026 ASEE Annual Conference & Exposition

Educational AI Agent for Student Team Interactions and Analysis

This paper presents the development of a multi-conversational AI agent designed to support student teams in managing collaborative projects more effectively. Student groups often face challenges such as communication breakdowns, lack of continuous support, and early-stage team conflicts. Current approaches, such as instructor interventions and predictive modeling, provide partial solutions but lack real-time, personalized engagement.

Our proposed system integrates an AI agent within a Discord-based environment to monitor conversations, identify potential conflicts, and generate feedback for both students and instructors. Using Natural Language Processing (NLP) techniques, the AI agent summarizes discussions, recognizes individual contributions, and maintains contextual memory across conversations.

All the group conversations will be logged into our MySQL database. Then we will conduct an NLP sentimental analysis to check for the team member's satisfaction. Every team with a score less than or equal to zero will be noted. Their logged conversations will be checked for team inefficiencies using the CATME five team work dimensions (Contributing to the Team's work, Interacting with Teammates, Having Relevant KSAs, Expecting Quality, Keeping the Team on Track ) using NLP text classification. We’ll note any low scores in any dimension. Their logged information will further be summarized using NLP abstractive summarization. Then, the professor will then be notified via email about the team inefficiency with the logged team conversations along with the highlighted low scores in those five dimensions.

We use a database in a HPC for easier retrieval and uploading of the logs that the students make in the team chat. We plan to use a vector database to better capture the context when a regular database reaches its limit especially in retrieving information from a long time.

The study follows a phased methodology, starting with two-member test groups, extending to larger teams, and evaluating the AI agent’s performance in understanding, recall, and contextual feedback generation.

Expected outcomes include improved communication flow, better conflict resolution, and enhanced team productivity through automated feedback and instructor insights. This work contributes to the field of AI-assisted educational collaboration by bridging social interaction analysis and intelligent facilitation in group learning environments.