Disentangling substantive traits and faking in high-stakes personality assessments using item response theory modeling

Read full text Summarise with VIRAT
Authors: Timo Seitz
Cited by: 0
References: 249

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Self-report questionnaires are frequently used to measure psychological constructs like personality traits, interests, and attitudes. When such questionnaires are employed in high-stakes contexts, where assessment results have important consequences for test-takers, there is a risk that test-takers deliberately distort their responses according to what is socially desirable. This response bias is known as faking, and has been studied by social scientists and research methodologists for decades. Most of the approaches that have been developed either aim at the detection or prevention of faking. In this dissertation, however, I address the response bias of faking by means of latent variable models. Such models allow disentangling different influences on item responses, thus enabling both a statistical control of faking and a substantive investigation of the response process associated with faking. In a series of three articles, I develop and examine psychometric models to separate the measurement of substantive traits from faking, and meet the issue from an item and test construction perspective. The models of all three articles are based on the multidimensional nominal response model (MNRM), which is a flexible item response theory (IRT) model that allows to address limitations of previous latent variable models of faking. Article I (Seitz et al., 2024) presents an examination of conditions under which the MNRM can effectively adjust substantive trait scores and latent correlations for faking, particularly concerning the question of how item content needs to be related to social desirability for the model to perform well. Article II (Seitz, Alagöz, & Meiser, 2025) extends the MNRM in a mixture modeling framework, assuming that test-takers differ qualitatively in the response strategy they employ. Article III (Seitz & Ulitzsch, 2025) introduces a further model extension, which makes use of item-level response times (RT) and allows for switches between response strategies over the course of the questionnaire. Along with simulations to study statistical properties of the models, all articles also include empirical demonstrations in experimental questionnaire data or real high-stakes datasets from personnel selection. The conducted research shows that faking is not an intractable response bias but that the application of sophisticated psychometric models can help to better manage its adverse effects. Beyond methodological contributions, the models of this dissertation also provide a framework researchers can use to enhance the understanding of the substantive nature of faking. Nevertheless, future studies should provide more validation evidence, especially with regard to the added value of the models in applied diagnostic contexts.

Fields of Study & Subjects

Psychometric Methodologies and Testing Psychological Testing and Assessment Personality Traits and Psychology

Publication details

Journal / source
MADOC (University of Mannheim)
Publication date
1 January 2026
Language
DE
Document type
Thesis