Messi H.J. Lee

I’m a fourth year PhD Candidate in the Division of Computational and Data Sciences at Washington University in St. Louis. I am part of the Psychological & Brain Sciences Track, doing research at the Diversity Science Lab. My advisors are Calvin K. Lai in the Psychological & Brain Sciences Department and Jacob M. Montgomery in the Political Sciences Track.

My research is at the intersection of Social Psychology and Artifical Intelligence where I study bias in Large Language Models (LLMs). Recently, I’ve been working on projects assessing homogeneity bias in LLMs where we find that these models portray socially subordinate groups as more homogeneous compared to their dominant counterparts. Moving forward, I aim to develop strategies to mitigate this bias, enhancing the fairness and inclusivity of AI systems.

news

Apr 29, 2024	I am thrilled to announce that I have been awarded a $1,000 travel grant to attend the ACM Conference on Fairness, Accountability, and Transparency (FAccT) in 2024.
Apr 29, 2024	You can now access the preprint of my paper “Large Language Models Portray Socially Subordinate Groups as More Homogeneous, Consistent with a Bias Observed in Humans” on ArXiv.
Apr 26, 2024	I was accepted to the OpenAI API Researcher Access Program! As part of the program, I will be granted OpenAI API credits to continue my work on homogeneity bias in Large Language Models.
Mar 30, 2024	My submission to ACM FAccT 2024 titled “Large Language Models Portray Socially Subordinate Groups as More Homogeneous, Consistent with a Bias Observed in Humans” was accepted for publication. See you in Rio!
Feb 08, 2024	I received the Outstanding Research Award at the Society for Personality and Social Psychology (SPSP) 2024 Annual Convention for my submission titled, “Outgroup Homogeneity-like Effect in Large Language Models.”

selected publications

Large Language Models Portray Socially Subordinate Groups as More Homogeneous, Consistent with a Bias Observed in Humans

Messi H. J. Lee , Jacob M. Montgomery , and Calvin K. Lai

Apr 2024

Abs PDF

Large language models (LLMs) are becoming pervasive in everyday life, yet their propensity to reproduce biases inherited from training data remains a pressing concern. Prior investigations into bias in LLMs have focused on the association of social groups with stereotypical attributes. However, this is only one form of human bias such systems may reproduce. We investigate a new form of bias in LLMs that resembles a social psychological phenomenon where socially subordinate groups are perceived as more homogeneous than socially dominant groups. We had ChatGPT, a state-of-the-art LLM, generate texts about intersectional group identities and compared those texts on measures of homogeneity. We consistently found that ChatGPT portrayed African, Asian, and Hispanic Americans as more homogeneous than White Americans, indicating that the model described racial minority groups with a narrower range of human experience. ChatGPT also portrayed women as more homogeneous than men, but these differences were small. Finally, we found that the effect of gender differed across racial/ethnic groups such that the effect of gender was consistent within African and Hispanic Americans but not within Asian and White Americans. We argue that the tendency of LLMs to describe groups as less diverse risks perpetuating stereotypes and discriminatory behavior.
America’s Racial Framework of Superiority and Americanness Embedded in Natural Language

Messi H. J. Lee , Jacob M. Montgomery , and Calvin K. Lai

PNAS Nexus, Jan 2024

Abs PDF

America’s racial framework can be summarized using two distinct dimensions: superiority/inferiority and Americanness/foreignness. We investigated America’s racial framework in a corpus of spoken and written language using word embeddings. Word embeddings place words on a low-dimensional space where words with similar meanings are proximate, allowing researchers to test whether the positions of group and attribute words in a semantic space reflect stereotypes. We trained a word embedding model on the Corpus of Contemporary American English—a corpus of 1 billion words that span 30 years and 8 text categories—and compared the positions of racial/ethnic groups with respect to superiority and Americanness. We found that America’s racial framework is embedded in American English. We also captured an additional nuance: Asian people were stereotyped as more American than Hispanic people. These results are empirical evidence that America’s racial framework is embedded in American English.