Machine Learning Methods in Computational Thinking Assessments: A Scoping Review

• In Event: Computers and Internet Applications in Education SIG Paper Session

Abstract

Researchers have recognized the importance of assessment for integrating computational thinking (CT) in contemporary education. Automatic assessments, especially machine learning-based (ML-based) assessments, have demonstrated promise by outperforming traditional assessment approaches in many aspects. This scoping review examined 22 empirical peer-reviewed publications that used ML methods in CT assessments. For each study, we extracted information regarding the year of publication, source of publication, study purpose, assessment data type, CT concept assessed, ML methods, sample, and validation method. Findings show that ML-based CT assessments tend to be effective, reliable, and outperform human graders. Implications for future research and educational practice of CT assessments are discussed.

Authors

• Bin Tan, University of Alberta

• Hanan Attitalla

• Katherine Wolber, University of Alberta

• Zexuan Pan, University of Alberta

• Maria Cutumisu, University of Alberta