Multimodal Data Fusion to Detect Preknowledge Test-Taking Behavior Using Machine Learning.

Area of Science:

• Educational Measurement
• Psychometrics
• Computer Science

Background:

• High-stakes decisions in education, medicine, and military recruitment rely on standardized test scores.
• Aberrant test-taking behaviors, like item practice, can compromise score validity and test fairness.
• Detecting such behaviors is crucial for maintaining assessment integrity in remote testing environments.

Purpose of the Study:

• To investigate the application of machine learning (ML) for detecting aberrant test-taking behaviors.
• To explore multimodal data fusion strategies integrating bio-information technology and psychometric data.
• To enhance the reliability and validity of inferences drawn from technology-assisted remote assessments.

Main Methods:

• Utilizing machine learning algorithms for pattern recognition in test-taking data.
• Integrating multimodal data sources: eye-tracking, response times, and item responses.
• Developing data fusion techniques to combine diverse bio-information and psychometric measures.

Main Results:

• The study demonstrates the potential of ML and multimodal data fusion in identifying aberrant test-taking.
• Specific bio-information and psychometric features were found to be indicative of unusual test behaviors.
• The proposed methods show promise for real-time detection of test irregularities.

Conclusions:

• Machine learning combined with multimodal data fusion offers a robust approach to detecting aberrant test-taking.
• This strategy can significantly improve the security and fairness of remote, technology-assisted assessments.
• Further research can refine these methods for broader application in educational and professional testing.