cs.CLFeb 19, 2026

Projective Psychological Assessment of Large Multimodal Models Using Thematic Apperception Tests

Authors: Anton Dzega, Aviad Elyashar, Ortal Slobodin, Odeya Cohen, Rami Puzis

Organizations: Faculty of Computer and Information Science, Ben-Gurion University of the Negev, Israel · Department of Computer Science, Shamoon College of Engineering, Israel · School of Education, Ben-Gurion University of the Negev, Israel · Department of Nursing Sciences, Ben-Gurion University of the Negev, Israel

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Thematic Apperception Test (TAT) is a psychometrically grounded, multidimensional assessment framework that systematically differentiates between cognitive-representational and affective-relational components of personality-like functioning. This test is a projective psychological framework designed to uncover unconscious aspects of personality. This study examines whether the personality traits of Large Multimodal Models (LMMs) can be assessed through non-language-based modalities, using the Social Cognition and Object Relations Scale - Global (SCORS-G). LMMs are employed in two distinct roles: as subject models (SMs), which generate stories in response to TAT images, and as evaluator models (EMs), who assess these narratives using the SCORS-G framework. Evaluators demonstrated an excellent ability to understand and analyze TAT responses. Their interpretations are highly consistent with those of human experts. Assessment results highlight that all models understand interpersonal dynamics very well and have a good grasp of the concept of self. However, they consistently fail to perceive and regulate aggression. Performance varied systematically across model families, with larger and more recent models consistently outperforming smaller and earlier ones across SCORS-G dimensions.

Figures & tables

Figure 1: The seven TAT images used in the TAT procedure, which SMs underwent.
Figure 2: The methodological pipeline.
FamilyModels
GPTgpt-5 (provider),
 gpt-5-mini (provider),
 gpt-4.1 (provider),
 gpt-4.1-mini (provider),
 o4-mini (provider),
 gpt-4o-mini (provider)
Table 1: List of LLMs used in study, grouped by family
Figure 3: Heatmap showing the comparison between EMs and average human ratings from the SCORS-G scoring guidelines using MAE and Spearman r, together with the consistency metrics of EMs’ ratings across iterations using A-WS-Std and APSC.
Figure 4: Repeated Measure ANOVA p-value results per SCORS-G dimension showing the AFF and ICS being significantly different between instructions.
Figure 5: Count of SCORS-G dimensions per SM, which were significantly different (p-value <= 0.05) between instructions.
Figure 6: Subject Model Performance Across Dimensions
Figure 7: Subject Model’s Average SCORES-G Dimension Score Per Image.
Figure 8: Subject Model Performance Across Dimensions
Figure 9: Overall Model Performance with 95% Confidence intervals
More in this topic

Personality Modeling in Language Models

Personality modeling in language models studies how traits are measured, represented, simulated, and controlled in model behavior.

See all 169 papers

Explore similar work

CardsList

  1. GenPT: Beyond Self-Report for Reliable LLM Psychometrics via Generative Projective Testing

    May 30, 2026Ming Wang, Shuang Wu, Bixuan Wang +7Psychometric Validation

    School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China · Mental Health Education Center, Northeastern University, Shenyang 110819, China · School of Psychology, Northeast Normal University, Changchun 130024, China +3

  2. Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring

    May 11, 2026Jiaqing Zhang, Sandeep Elluri, Bhanu Cherukuvada +7LLM AuditingMultimodal Large Language Models

    Department of Electrical & Computer Engineering · Department of Computer and Information Science and Engineering · Department of Clinical and Health Psychology +1

  3. Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality?

    May 21, 2026Caixin Kang, Tianyu Yan, Sitong Gong +8LLM GroundingPersonality Modeling in Language Models

    1The University of Tokyo · 2Shanda AI Research Tokyo · 3Dalian University of Technology