Graph Theory Approach to Detect Examinees Involved in Test Collusion

Area of Science:

• Psychometrics
• Graph Theory
• Educational Measurement

Background:

• Test collusion (TC) involves sharing test materials or answers, posing a significant threat to the validity of educational assessments.
• Item preknowledge is a critical form of TC, potentially granting unfair advantages to examinees.

Purpose of the Study:

• To introduce a novel graph theory-based methodology for detecting test collusion.
• To identify groups of examinees engaged in TC without prior knowledge of affected test sections.

Main Methods:

• Application of graph theory to analyze response similarity among examinees.
• Utilizing various response similarity indices tailored to specific TC types.
• Identifying groups through connected components, cliques, or near-cliques.

Main Results:

• The proposed graph theory approach effectively identifies groups involved in test collusion.
• Demonstrated effectiveness through comparisons with existing methods using real and simulated data.

Conclusions:

• Graph theory offers a powerful framework for detecting test collusion.
• The method provides a flexible and robust tool for enhancing test security and score validity.