Artificial Intelligence-Assisted Socially Interactive Agents to Transform Undergraduate STEM Advising

This project aims to serve the national interest by investigating a new approach to college academic advising by developing a virtual, computer-based advising assistant to support undergraduate science, technology, engineering, and mathematics (STEM) students. Academic advising is crucial for keeping students on track toward graduation and preparing them for the workforce. However, universities face significant challenges due to limited advising staff and the rapid pace at which job markets and college course requirements change. Traditional computer text chatbots often fail to understand complicated student questions, while newer artificial intelligence tools frequently hallucinate incorrect information or recommend courses that violate official rules. To address this issue, this Level 1 Engaged Student Learning project plans to create an advanced virtual assistant that looks and talks like a helpful human counselor while offering useful advice and relevant recommendations. This virtual agent will help students visualize clear career pathways by connecting their coursework directly to real-time job market data. By providing personalized, accurate, and emotionally supportive guidance, the project is worthwhile because it reduces student anxiety, makes it easier for students to ask for help, and ultimately improves graduation rates and workforce readiness for students in STEM fields. The project seeks to advance the fields of educational data mining, neuro-symbolic artificial intelligence, and human-computer interaction to build a trustworthy institutional decision-support system. The project goals are to design and evaluate a multi-layered framework that balances conversational fluency with absolute policy compliance. The research methods and approaches are structured into three interconnected components. First, develop predictive student models that leverage historical academic data to forecast course performance, attrition risks, and relevant course recommendations. Second, implement a neuro-symbolic reasoning engine that matches student intent against educational predictive and recommender systems, institutional rules, and real-time job market information to guarantee error-free, policy-compliant path planning. Third, deploy these capabilities through a conversational agent featuring adaptive socio-emotional intelligence, using multimodal cues to scaffold student interactions. The contributions of this project include novel open-source architectures for hybrid AI reasoning, design guidelines for virtual agents, and empirical mixed-methods evidence regarding AI-assisted systems in high-stakes institutional workflows. The NSF IUSE: EDU Program supports research and development projects to improve the effectiveness of STEM education for all students. Through the Engaged Student Learning track, the program supports the creation, exploration, and implementation of promising practices and tools. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.