Neural cognitive diagnosis modeling incorporating response times

Area of Science:

• Educational Technology
• Artificial Intelligence in Education
• Cognitive Science

Background:

• Cognitive diagnosis is crucial for understanding student mastery of knowledge concepts in intelligent education.
• Response time (RT) is valuable process data from computerized testing that can enhance diagnostic accuracy.
• Traditional statistical models often overlook RT, limiting diagnostic precision.

Purpose of the Study:

• To propose a novel Joint Response Times Neural Cognitive Diagnosis (JRT-NCD) model.
• To leverage neural networks for modeling complex, nonlinear student-exercise interactions.
• To incorporate RT as a feature for refining student ability diagnostics.

Main Methods:

• Developed the JRT-NCD model using neural networks to process complex interactions and RT data.
• Applied the model to three diverse datasets: PISA2012, 2MFC, and ASSIST09.
• Compared JRT-NCD performance against traditional statistical models and a non-RT neural network model (NCD).

Main Results:

• Neural networks demonstrate superior fitting capabilities for complex, nonlinear educational data compared to traditional models.
• The JRT-NCD model achieved higher diagnostic accuracy than the NCD model, which ignores RT.
• The proposed model effectively reduces the impact of 'overspeed behavior' on diagnostic outcomes while maintaining interpretability.

Conclusions:

• The JRT-NCD model offers enhanced accuracy and interpretability in cognitive diagnosis by integrating response times.
• Neural network-based approaches provide a more robust framework for cognitive diagnosis than traditional statistical methods.
• Incorporating process data like RT significantly refines the assessment of student abilities.