Authors:

Henry J Lee

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Intelligent tutoring systems (ITSs) have been utilized to assist humans in learning new skills in psychomotor learning. Self-confidence, workload, and learning stage are three important factors that contribute to psychomotor learning. Self-confidence and workload have been shown to influence learning outcomes, and psychomotor task performance can be measured in terms of the learning stage. To design ITS algorithms, models will need to capture interactions between relevant cognitive stages and learning outcomes. Previous works utilized reinforcement learning methods by training self-confidence state, workload state, and learning stage state Markov decision process (MDP) models to improve ITS's assistance capability and the models were trained with the assumption that the states were independent from one another. However, there may be cases where the states could be interdependent (e.g., the relationship between self-confidence and learning progression), or a different model can be explored entirely. In this work, the MDP framework is modified to capture and investigate the interdependencies between different cognitive states, including self-confidence, workload, and learning stage. Our findings suggest that coupling self-confidence with workload or learning stage may enhance the model's predictive capabilities; however, the increased number of model parameters in these coupling states could lead to overfitting. Additionally, we investigate how the models could be improved through k-means clustering, resulting in multiple models for each state to capture a range of human cognitive state behavior. Keywords: Intelligent Tutoring Systems (ITS); Cognitive State Modeling; Markov Decision Process; Reinforcement Learning

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Henry J Lee