Common Sense vs. Morality: The Curious Case of Narrative Focus Bias in LLMs

2026-03-10

AI Summary

This work addresses the tendency of large language models to overlook embedded commonsense contradictions in moral reasoning. To this end, the authors introduce CoMoral, the first benchmark that integrates moral dilemmas with explicit commonsense conflicts, enabling a systematic evaluation of commonsense awareness across ten prominent models through carefully crafted scenarios. The study identifies and formally defines a novel phenomenon termed “narrative focus bias”: models are more likely to detect commonsense violations committed by secondary characters than those made by the narrator themselves. Experimental results demonstrate that current models generally fail to recognize commonsense inconsistencies within moral contexts without explicit prompting, and their performance is significantly influenced by the narrative perspective. These findings underscore the urgent need to enhance the robustness of commonsense reasoning in language models operating in ethically sensitive domains.

Technology Category

• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics 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

Large Language Models (LLMs) are increasingly deployed across diverse real-world applications and user communities. As such, it is crucial that these models remain both morally grounded and knowledge-aware. In this work, we uncover a critical limitation of current LLMs -- their tendency to prioritize moral reasoning over commonsense understanding. To investigate this phenomenon, we introduce CoMoral, a novel benchmark dataset containing commonsense contradictions embedded within moral dilemmas. Through extensive evaluation of ten LLMs across different model sizes, we find that existing models consistently struggle to identify such contradictions without prior signal. Furthermore, we observe a pervasive narrative focus bias, wherein LLMs more readily detect commonsense contradictions when they are attributed to a secondary character rather than the primary (narrator) character. Our comprehensive analysis underscores the need for enhanced reasoning-aware training to improve the commonsense robustness of large language models.

Problem

• commonsense reasoning
• moral reasoning
• narrative focus bias
• large language models
• commonsense contradictions

Innovation

• narrative focus bias
• commonsense reasoning
• moral dilemmas
• CoMoral benchmark
• reasoning-aware training