Faking in High-Stakes Personality Assessments: A Response-Time-Based Latent Response Mixture Modeling Approach.

Area of Science:

• Psychological measurement
• Psychometrics
• Behavioral science

Background:

• Personality assessments in high-stakes settings risk response bias (faking).
• Existing latent variable models assume uniform response strategies, limiting accuracy.
• Faking can distort results in job applications and other critical evaluations.

Purpose of the Study:

• To present a novel latent response mixture item response theory (IRT) model for faking.
• To account for dynamic test-taker response strategies during assessments.
• To identify and investigate faking-related strategies at the person-by-item level.

Main Methods:

• Developed a latent response mixture IRT model integrating item responses and response times (RT).
• Modeled changes in test-taker response strategies throughout an assessment.
• Utilized person-by-item level analysis for detailed faking strategy investigation.

Main Results:

• Parameter recovery studies demonstrated reliable model estimation under realistic conditions.
• The model was successfully applied to a large empirical dataset (N=1,824) from a job application context.
• The study confirmed the model's utility for analyzing real-world high-stakes assessment data.

Conclusions:

• The proposed IRT model offers a more nuanced approach to understanding and detecting faking.
• It enhances the accuracy of personality assessments in high-stakes situations.
• The model has significant implications for both psychological measurement and substantive research.