Assessing the personality in Large Language Models through Diverse Prompting

Date of Award

Spring 2025

Degree Name

Bachelor of Science

Department

Computer Science & Mathematics; College of Arts & Sciences

First Advisor

Weihong Ni

Abstract

As large language models (LLMs) like ChatGPT continue to evolve, questions arise regarding their ability to exhibit human-like personality traits. This study investigates LLMs’ personality by employing the Big Five Personality Test as an assessment tool. We first collected data using both the OpenAI API and the Gemini API by administering the personality test at baseline and for given pre-designed prompts. We then explored how different prompts influence the test results through statistical analyses. This research contributes to AI behavior modeling and enhances our understanding of how prompting shapes LLM responses.

This document is currently not available here.

Share

COinS
 

Assessing the personality in Large Language Models through Diverse Prompting

As large language models (LLMs) like ChatGPT continue to evolve, questions arise regarding their ability to exhibit human-like personality traits. This study investigates LLMs’ personality by employing the Big Five Personality Test as an assessment tool. We first collected data using both the OpenAI API and the Gemini API by administering the personality test at baseline and for given pre-designed prompts. We then explored how different prompts influence the test results through statistical analyses. This research contributes to AI behavior modeling and enhances our understanding of how prompting shapes LLM responses.