Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decoding Emergent Big Five Traits in Large Language Models: Temperature-Dependent Expression and Architectural Clustering

Christos-Nikolaos Zacharopoulos, Revekka Kyriakoglou

TL;DR

This work investigates whether large language models exhibit personality-like patterns by applying the Big Five Inventory-2 (BFI-2) across six LLMs and systematically varying the decoding temperature. It combines quantitative trait evaluation with agglomerative hierarchical clustering to reveal architecture-driven groupings in trait profiles, highlighting temperature as a decoding parameter that modulates certain traits. The key findings show significant model-level variation for four traits (Extraversion, Agreeableness, Conscientiousness, Openness) and a robust temperature-related influence on Neuroticism and Extraversion, suggesting that both architecture and decoding shape emergent personality-like behavior. The study offers implications for model tuning, selection, and governance of AI-driven interactions, and provides data-driven directions for future causal analyses of architectural and training-data contributions.

Abstract

As Large Language Models (LLMs) become integral to human-centered applications, understanding their personality-like behaviors is increasingly important for responsible development and deployment. This paper systematically evaluates six LLMs, applying the Big Five Inventory-2 (BFI-2) framework, to assess trait expressions under varying sampling temperatures. We find significant differences across four of the five personality dimensions, with Neuroticism and Extraversion susceptible to temperature adjustments. Further, hierarchical clustering reveals distinct model clusters, suggesting that architectural features may predispose certain models toward stable trait profiles. Taken together, these results offer new insights into the emergence of personality-like patterns in LLMs and provide a new perspective on model tuning, selection, and the ethical governance of AI systems. We share the data and code for this analysis here: https://osf.io/bsvzc/?view_only=6672219bede24b4e875097426dc3fac1

Decoding Emergent Big Five Traits in Large Language Models: Temperature-Dependent Expression and Architectural Clustering

TL;DR

This work investigates whether large language models exhibit personality-like patterns by applying the Big Five Inventory-2 (BFI-2) across six LLMs and systematically varying the decoding temperature. It combines quantitative trait evaluation with agglomerative hierarchical clustering to reveal architecture-driven groupings in trait profiles, highlighting temperature as a decoding parameter that modulates certain traits. The key findings show significant model-level variation for four traits (Extraversion, Agreeableness, Conscientiousness, Openness) and a robust temperature-related influence on Neuroticism and Extraversion, suggesting that both architecture and decoding shape emergent personality-like behavior. The study offers implications for model tuning, selection, and governance of AI-driven interactions, and provides data-driven directions for future causal analyses of architectural and training-data contributions.

Abstract

As Large Language Models (LLMs) become integral to human-centered applications, understanding their personality-like behaviors is increasingly important for responsible development and deployment. This paper systematically evaluates six LLMs, applying the Big Five Inventory-2 (BFI-2) framework, to assess trait expressions under varying sampling temperatures. We find significant differences across four of the five personality dimensions, with Neuroticism and Extraversion susceptible to temperature adjustments. Further, hierarchical clustering reveals distinct model clusters, suggesting that architectural features may predispose certain models toward stable trait profiles. Taken together, these results offer new insights into the emergence of personality-like patterns in LLMs and provide a new perspective on model tuning, selection, and the ethical governance of AI systems. We share the data and code for this analysis here: https://osf.io/bsvzc/?view_only=6672219bede24b4e875097426dc3fac1

Paper Structure

This paper contains 6 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Results
  4. Discussion
  5. Limitations
  6. Appendix A

Figures (3)

  • Figure 1: Hierarchical clustering of models based on personality profiles, revealing distinct groupings and architectural influences on trait expressions.
  • Figure 2: Effects of sampling temperature on personality traits, demonstrating sensitivity in Neuroticism and Extraversion.
  • Figure 3: Comparison of domain scores across different large language models, highlighting significant variations in personality trait expressions.