Who Laughs with Whom? Disentangling Influential Factors in Humor Preferences across User Clusters and LLMs

2026-01-06

arXiv.org

AI Summary

This study addresses the challenge that large language models (LLMs) struggle to accurately assess humor due to significant individual and cultural variations in humor preferences. To tackle this, the authors propose a novel approach that first clusters users based on voting logs to identify distinct humor preference groups. They then integrate interpretable humor features with a Bradley–Terry–Luce model to estimate group-specific weights for these features. These learned preferences are subsequently incorporated into the LLM via role-based prompt engineering, enabling the model to simulate the humor judgments of specific user groups. This work is the first to combine user clustering with interpretable humor modeling, achieving targeted control over LLMs’ humor preferences. Experimental results demonstrate that the method significantly improves the alignment between model evaluations and group-level human humor judgments.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Humor preferences vary widely across individuals and cultures, complicating the evaluation of humor using large language models (LLMs). In this study, we model heterogeneity in humor preferences in Oogiri, a Japanese creative response game, by clustering users with voting logs and estimating cluster-specific weights over interpretable preference factors using Bradley-Terry-Luce models. We elicit preference judgments from LLMs by prompting them to select the funnier response and found that user clusters exhibit distinct preference patterns and that the LLM results can resemble those of particular clusters. Finally, we demonstrate that, by persona prompting, LLM preferences can be directed toward a specific cluster. The scripts for data collection and analysis will be released to support reproducibility.

Problem

• humor preferences
• user clustering
• large language models
• preference heterogeneity
• cultural variation

Innovation

• humor preference modeling
• user clustering
• Bradley-Terry-Luce model
• persona prompting
• large language models