Malice, inequality, instability, or ignorance? Disentangling the mechanisms of LLM unfairness

Yang, Ke

Permalink

https://hdl.handle.net/2142/132559 Copy

Description

Title

Malice, inequality, instability, or ignorance? Disentangling the mechanisms of LLM unfairness

Author(s)

Yang, Ke

Issue Date

2025-12-03

Director of Research (if dissertation) or Advisor (if thesis)

Zhai, ChengXiang

Department of Study

Siebel School Comp & Data Sci

Discipline

Computer Science

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• large language model
• unfairness measurement

Abstract

Ensuring fairness in large language models (LLMs) is critical as these models are increasingly deployed in sensitive domains. Traditional fairness metrics typically report a single scalar score, which conflates distinct sources of model failure and obscures underlying biases. In this work, we propose a Hierarchical Bias-Variance Decomposition framework—termed BDSU—that decomposes total discrimination risk into four interpretable components: Bias (systematic global error), Disparity (group-level variance), Sensitivity (context-level variance), and Uncertainty (stochastic or token-level variance). By applying the law of total variance recursively, BDSU provides a principled method to quantify and separate these failure modes, aligning each with ethical and reliability priorities. We further introduce a conditional micro-diagnosis to evaluate fairness at the group level, enabling fine-grained auditing and targeted interventions. Our theoretical framework lays the foundation for more transparent, actionable, and robust evaluation of LLM fairness, highlighting the distinct mechanisms by which models may perpetuate bias or exhibit instability.

Graduation Semester

2025-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/132559

Copyright and License Information

Copyright © 2025 Ke Yang. All rights reserved.

Owning Collections