Mitigating Popularity Bias in Counterfactual Explanations using Large Language Models

2025-08-12

AI Summary

Existing counterfactual explanation (CFE) methods suffer from popularity bias, yielding explanations misaligned with users’ genuine preferences. To address this, we propose a debiasing framework that leverages large language models (LLMs) to semantically interpret user historical interactions and enforce role-consistent filtering—removing highly popular yet preference-irrelevant items. The cleaned interaction sequences are then fed into the neural CFE generator ACCENT to produce actionable, personalized explanations. Evaluated on two public recommendation datasets, our approach significantly reduces popularity bias (average reduction of 32.7%) while improving explanation personalization and user trustworthiness. Our key contribution is the first integration of LLM-driven, preference-aware preprocessing into the CFE generation pipeline—mitigating bias at the *data layer*, rather than solely at the *model layer*. This paradigm shift enables more faithful alignment between generated counterfactuals and individual user intent.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Preferences
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Counterfactual explanations (CFEs) offer a tangible and actionable way to explain recommendations by showing users a "what-if" scenario that demonstrates how small changes in their history would alter the system's output. However, existing CFE methods are susceptible to bias, generating explanations that might misalign with the user's actual preferences. In this paper, we propose a pre-processing step that leverages large language models to filter out-of-character history items before generating an explanation. In experiments on two public datasets, we focus on popularity bias and apply our approach to ACCENT, a neural CFE framework. We find that it creates counterfactuals that are more closely aligned with each user's popularity preferences than ACCENT alone.

Problem

• Mitigating popularity bias in counterfactual explanation methods
• Aligning generated explanations with user preferences
• Filtering out-of-character history items using LLMs

Innovation

• Leveraging large language models for pre-processing
• Filtering out-of-character history items
• Aligning counterfactuals with user popularity preferences