Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Is persona enough for personality? Using ChatGPT to reconstruct an agent's latent personality from simple descriptions

Yongyi Ji, Zhisheng Tang, Mayank Kejriwal

TL;DR

The authors' experiments reveal a significant degree of consistency in personality reconstruction, although some inconsistencies and biases, such as a tendency to default to positive traits in the absence of explicit information, are also observed.

Abstract

Personality, a fundamental aspect of human cognition, contains a range of traits that influence behaviors, thoughts, and emotions. This paper explores the capabilities of large language models (LLMs) in reconstructing these complex cognitive attributes based only on simple descriptions containing socio-demographic and personality type information. Utilizing the HEXACO personality framework, our study examines the consistency of LLMs in recovering and predicting underlying (latent) personality dimensions from simple descriptions. Our experiments reveal a significant degree of consistency in personality reconstruction, although some inconsistencies and biases, such as a tendency to default to positive traits in the absence of explicit information, are also observed. Additionally, socio-demographic factors like age and number of children were found to influence the reconstructed personality dimensions. These findings have implications for building sophisticated agent-based simulacra using LLMs and highlight the need for further research on robust personality generation in LLMs.

Is persona enough for personality? Using ChatGPT to reconstruct an agent's latent personality from simple descriptions

TL;DR

The authors' experiments reveal a significant degree of consistency in personality reconstruction, although some inconsistencies and biases, such as a tendency to default to positive traits in the absence of explicit information, are also observed.

Abstract

Personality, a fundamental aspect of human cognition, contains a range of traits that influence behaviors, thoughts, and emotions. This paper explores the capabilities of large language models (LLMs) in reconstructing these complex cognitive attributes based only on simple descriptions containing socio-demographic and personality type information. Utilizing the HEXACO personality framework, our study examines the consistency of LLMs in recovering and predicting underlying (latent) personality dimensions from simple descriptions. Our experiments reveal a significant degree of consistency in personality reconstruction, although some inconsistencies and biases, such as a tendency to default to positive traits in the absence of explicit information, are also observed. Additionally, socio-demographic factors like age and number of children were found to influence the reconstructed personality dimensions. These findings have implications for building sophisticated agent-based simulacra using LLMs and highlight the need for further research on robust personality generation in LLMs.
Paper Structure (7 sections, 3 figures, 8 tables)

This paper contains 7 sections, 3 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Persona Construction
  3. Experiments
  4. Result
  5. Discussion
  6. Appendix
  7. Transformation Procedure

Figures (3)

  • Figure 1: The results of the HEXACO personality test given by personas reconstructed using GPT-3.5-Turbo provided with 1000 personality descriptions. The two columns on the left show the number of dimensions that GPT-3.5-Turbo reconstructed as high, and the two columns on the right show the number of dimensions that GPT-3.5-Turbo reconstructed as low. The two columns on the top show the number of dimensions that are provided as high in the persona descriptions given to GPT-3.5-Turbo, whereas the two columns on the bottom show the number of dimensions that are provided as low in the persona descriptions given to GPT-3.5-Turbo. Hence, the green columns represent consistency, and the red columns represent inconsistency. The results for the six individual dimensions of personality are shown in Appendix Table \ref{['tab:app_overall']}
  • Figure 2: The reconstructed scores (high or low) of each of the six personality dimensions when GPT-3.5-Turbo is provided with a personality type description that omits one of the six personality dimensions. The title of each sub-graph indicates which dimension is omitted. The result for GPT-4-Turbo is provided in Appendix Figure \ref{['fig:app_omit_gpt4']}.
  • Figure 3: The reconstructed scores (high or low) of each of the six personality dimensions when GPT-4-Turbo is provided with a personality type description that omits one of the six personality dimensions. The title of each sub-graph indicates which dimension is omitted.