Experiences Build Characters: The Linguistic Origins and Functional Impact of LLM Personality

TL;DR

This paper explores how diverse training experiences influence the personality traits and problem-solving abilities of large language models, revealing that tailored training can enhance specific competencies and establish causal links between data and model behaviour.

Contribution

It introduces a framework for quantifying LLM personality traits and demonstrates how domain-specific training impacts linguistic style and reasoning performance.

Findings

Model competence peaks at 'Expressive Generalists' and 'Suppressed Specialists'

Reduced social traits ('Suppression') can improve complex reasoning

Training data linguistics directly influence lexical diversity

Abstract

Human problem-solving is enriched by a diversity of styles and personality traits, yet the development of Large Language Models (LLMs) has largely prioritized uniform performance benchmarks that favour specific behavioural tendencies such as assertiveness. To investigate how diverse experiences shape machine personality and influence problem-solving, this study employs continued pre-training to expose models to domain-specific texts in an unsupervised manner, simulating the accumulation of experience. By adapting the Big Five framework via the Machine Personality Inventory (MPI), we quantify the personality traits of these model variants and analyse their relationship to linguistic style and reasoning behaviour. The findings reveal that model competence is bimodal, peaking at "Expressive Generalists" and "Suppressed Specialists," while identifying a "Suppression Advantage" where reduced social traits enhance complex reasoning performance. This study further establishes a causal link between training data linguistics, such as imperative frequency, and lexical diversity, providing a roadmap for "Personality Engineering".