Summary

• Led qualitative research on age bias in LLMs by conducting secondary analysis and literature reviews to shape human-centered design and AI research for inclusive technology development

• Presented research at the Dana Knox Student Showcase (2024) and Web4All (2025)

• Analyzed 1,648 GPT-3.5 responses using the BBQ dataset to examine age-bias and stereotypes of aging in AI language models through qualitative coding.

• Presented the research to the ACM ASSETS ‘25 program (10/2025)
• View Published Paper

Abstract

Large Language Models (LLMs) are increasingly used in decision-making, from content generation to employment screening, raising concerns about embedded biases. While biases related to gender, race, and culture have been studied, understanding age-bias and stereotypes of aging in LLMs remain underexplored. This study analyzes LLM-generated responses to prompts on aging, revealing stereotypical biases in technology proficiency, cognitive and physical decline, and job roles. Even responses that did not have an explicit age bias had mentions of stereotypes. We discuss considerations for involving older adults’ perspectives through human-in-the-loop methodologies yet exercising caution about aspects like internalized ageism. We discuss opportunities in exploring the role of educational explanations in minimizing harmful assumptions about marginalized subgroups.

Groups

Faculty Advisor

Alisha Pradhan

My Undergrad Thesis Committee ‘25

Adam Spryszynski, Keita Ohshiro

Co-authors

Sherwin Dewan, Ismail Shaikh, Connie Shaw, Abhilash Sahoo, Akshita Jha, Alisha Pradhan

View the Files Below