CoPAMoE: Cognitive Process-Aware Mixture of Experts for Bias Understanding and Mitigation
Authors:
- Xingshen LiuXLXingshen Liu
- Fuxin WangFWFuxin Wang
- Fan WangFWFan Wang
- Zhixuan JiaZJZhixuan Jia
- Haowei WangHWHaowei Wang
- Congjing RanCRCongjing Ran
Preprint
- 24 June 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Bias handling in expert-augmented LLMs is not merely a matter of routing tasks to experts, it requires coordinating how biased cues are detected, how their social meanings are attributed, and how interventions are generated. However, existing parameter-efficient fine-tuning and mixture-of-experts approaches typically organize specialization around task labels, domains, or input-level routing signals, rather than explicitly modeling how evidence of bias should inform its interpretation and subsequent rewriting. This paper proposes CoPAMoE, a cognitive process-aware PEFT-MoE framework for bias understanding and mitigation. CoPAMoE organizes bias-related subtasks into three functional stages, Cue Perception, Attribution Reasoning, and Intervention Control. It combines stage-partitioned LoRA experts, shared experts, a task--stage routing prior, and an Adjacent Stage Gated Coordination (ASGC) module. The resulting architecture explicitly separates stage-specific adaptation from task-general bias signals and supports controlled information transfer between adjacent stages. Experiments on a unified benchmark built from CrowS-Pairs, StereoSet, and WinoBias show that CoPAMoE improves average performance over representative PEFT and MoE baselines on two backbone LLMs. Ablation, routing, representation, oracle, and efficiency analyses further suggest that the gains are associated with stage-aware specialization and cross-stage dependency transfer.Keywords
This publication has 18 references indexed in Scilit:
- A Survey on Stereotype Detection in Natural Language ProcessingACM Computing Surveys, 2025
- Bias and Fairness in Large Language Models: A SurveyComputational Linguistics, 2024
- Five sources of bias in natural language processingLanguage and Linguistics Compass, 2021
- The Social Neuroscience of PrejudiceAnnual Review of Psychology, 2021
- Measuring and Mitigating Unintended Bias in Text Classification2018
- Social Attributions from Faces: Determinants, Consequences, Accuracy, and Functional SignificanceAnnual Review of Psychology, 2015
- Prejudice and Stereotype Maintenance Processes: Attention, Attribution, and Individuation.Journal of Personality and Social Psychology, 2005
- Neural Signals for the Detection of Unintentional Race BiasPsychological Science, 2004
- A Continuum of Impression Formation, from Category-Based to Individuating Processes: Influences of Information and Motivation on Attention and Interpretation1990
- Stereotypes and prejudice: Their automatic and controlled components.Journal of Personality and Social Psychology, 1989