Schedule of Events | Symposia

Implementing Deep Learning Models to Personalize Learning in Cognitive Training

Poster Session A - Saturday, March 29, 2025, 3:00 – 5:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Elnaz Vafaei¹ (e.vafaei@northeastern.edu), Jaap Munneke¹, Susanne Jaeggi¹, Aaron Seitz¹; ¹Northeastern University

Cognitive training programs have demonstrated potential in improving cognitive abilities. However, due to a one-size-fits-all approach, the reported effectiveness of cognitive training has been inconsistent. This highlights the need for adaptive interventions that personalize characteristics of cognitive training tasks based on participants’ performances in a real-time manner. In this study, we propose a framework for a closed-loop adaptive interactive system designed to enhance individual performance on a series of cognitive training tasks. In the first stage, we aimed to model individual differences in performance trajectories (Accuracy and N-level) extracted from a working memory (WM) task, serving as the initial prototype. Predicting the performance trajectory for the next trial can serve as a strategy to define the proportional characteristics of subsequent tasks. We employed several deep learning models, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), to predict participants' performance based on data from 277 undergraduates recruited from the Universities of California, Irvine, and Riverside, who completed 20 sessions of the N-back task. The results demonstrate that deep learning models accurately predict performance trajectories in the N-back task, achieving an R-squared value of 0.9168. In addition, no meaningful difference was observed between the GRU and LSTM models' performance. Deep learning models can accurately predict participant performance and can be used to create personalized cognitive training systems that enhance engagement and task compliance. Future research would benefit from exploring the application of the proposed model in a real-time closed-loop system.

Topic Area: EXECUTIVE PROCESSES: Other