In this talk, the speaker will introduce Messages From A Future You (MFAFY), an AI-driven system designed to enhance undergraduate STEM performance through personalized, just-in-time psychological interventions. The research models student performance as a dynamic system where cognitive and non-cognitive factors evolve together over time. By forecasting both long-term cognitive outcomes and short-term changes in non-cognitive traits such as self-efficacy and social belonging, MFAFY enables tailored interventions that create meaningful impact.

At the core of MFAFY is a novel use of generative AI, which leverages pre-trained large language models (LLMs) to forecast academic trajectories based on students’ longitudinal experiential (LE) data. This multidimensional dataset, curated from a multi-semester study at the University of Nebraska-Lincoln, captures the evolving cognitive and non-cognitive characteristics of students. The main scientific challenge is adapting LLMs to handle the unique complexities of LE data, including missing values, data imbalance, and long context dependencies. To address these challenges, advanced data processing techniques and novel methods for adapting LLMs were developed, allowing the models to capture intricate temporal patterns and co-evolving traits. Although initial results are promising, ongoing research aims to enhance model robustness and predictive accuracy by incorporating multimodal learning. This innovative work marks a significant step in using generative AI to provide personalized, data-driven support for student success in STEM education, moving closer to realizing personal AI for learning.

Speaker: M. R. Hasan