The Impacts of Process Behavior on Item Functioning Employing Machine Learning Methods and Generalized Linear Models

• In Event: AERA Graduate Student Research-in-Progress Session 3
  In Roundtable Session: Educational Assessment: Policy, Models, and Practices (Table 9)

Abstract

With technological advancements in cognitive assessments, the influence of response behaviors on item functioning has become a crucial factor in measuring cognitive constructs or abilities. Cognitive response process is not directly observable; however, it can be well reflected in the process data captured during assessments. This study investigates effects of process features (behavior) on item-functioning using Machine Learning methods. Furthermore, this study explores differential impacts of response behaviors on item functioning due to test-takers' background factors employing Generalized Linear Modeling approach. The results suggest that at different ability levels, various aspects of the response process exhibit differential impacts on item-functioning. This study provides insights for designing interactive digital assessments to promote fairness, equity, and inclusivity for diverse groups of test-takers.

Author

• Mehdi Rajeb, University of Alabama