Markov Decision Process Measurement Model.

Area of Science:

• Educational Psychology
• Cognitive Science
• Educational Measurement

Background:

• Modeling within-task actions is crucial for understanding student competencies.
• Traditional models often struggle to incorporate detailed behavioral data from within tasks.
• Latent trait estimation can be enhanced by analyzing fine-grained actions.

Purpose of the Study:

• To explore the use of Markov decision processes (MDPs) for mapping within-task actions to latent student traits.
• To evaluate the psychometric properties and parameter recovery of the proposed cognitive model.
• To compare the efficacy of the MDP model against traditional Item Response Theory (IRT) models.

Main Methods:

• Developing a cognitive model based on Markov decision processes (MDPs).
• Conducting simulation studies to assess parameter recovery for the MDP model.
• Applying the MDP model to empirical data from an educational strategy game.
• Comparing MDP model estimates with a partial-credit Item Response Theory (IRT) model.

Main Results:

• Simulation studies demonstrated adequate parameter recovery for the MDP model.
• The MDP model successfully mapped within-task actions to latent traits in an educational game.
• Estimates derived from the MDP model showed a stronger correlation with posttest results compared to the IRT model.
• The MDP approach provided richer insights into student competencies through action analysis.

Conclusions:

• Markov decision processes offer a viable cognitive modeling approach for analyzing within-task actions in educational settings.
• The proposed MDP model demonstrates superior predictive validity for learning outcomes over traditional IRT models.
• Integrating fine-grained behavioral data via cognitive models can significantly improve the assessment of student competencies.