# Engagement Undermines Safety: How Stereotypes and Toxicity Shape Humor in Language Models

arXiv: https://arxiv.org/abs/2510.18454 · HTML: https://mlsanity.com/paper/2510.18454/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security.
- Builds on 30 of its 54 references that are in the map.
- First posted on arXiv 21 Oct 2025.

## Placed in

- LLM safety and security / [Safety alignment, refusal and guardrails](https://mlsanity.com/safety/safety-alignment-refusal-and-guardrails/)

## Builds on (most cited first, 30 of 30)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://mlsanity.com/paper/1909.11942/) (2019)
- [CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models](https://mlsanity.com/paper/2010.00133/) (2020)
- [Bias and Fairness in Large Language Models: A Survey](https://mlsanity.com/paper/2309.00770/) (2023)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://mlsanity.com/paper/2304.05335/) (2023)
- [HateBERT: Retraining BERT for Abusive Language Detection in English](https://mlsanity.com/paper/2010.12472/) (2020)
- [A New Generation of Perspective API: Efficient Multilingual Character-level Transformers](https://mlsanity.com/paper/2202.11176/) (2022)
- [A Pretrainer's Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicity](https://mlsanity.com/paper/2305.13169/) (2023)
- [Jasper and Stella: distillation of SOTA embedding models](https://mlsanity.com/paper/2412.19048/) (2024)
- [Community Interaction and Conflict on the Web](https://mlsanity.com/paper/1803.03697/) (2018)
- [The Next Chapter: A Study of Large Language Models in Storytelling](https://mlsanity.com/paper/2301.09790/) (2023)
- [Humor Detection: A Transformer Gets the Last Laugh](https://mlsanity.com/paper/1909.00252/) (2019)
- [From Role-Play to Drama-Interaction: An LLM Solution](https://mlsanity.com/paper/2405.14231/) (2024)
- [HoLLMwood: Unleashing the Creativity of Large Language Models in Screenwriting via Role Playing](https://mlsanity.com/paper/2406.11683/) (2024)
- [Collaborative Storytelling with Large-scale Neural Language Models](https://mlsanity.com/paper/2011.10208/) (2020)
- [ColBERT: Using BERT Sentence Embedding in Parallel Neural Networks for Computational Humor](https://mlsanity.com/paper/2004.12765/) (2020)
- [Exploring the Impact of Personality Traits on LLM Bias and Toxicity](https://mlsanity.com/paper/2502.12566/) (2025)
- [Uncertainty and Surprisal Jointly Deliver the Punchline: Exploiting Incongruity-Based Features for Humor Recognition](https://mlsanity.com/paper/2012.12007/) (2020)
- [Innovative Thinking, Infinite Humor: Humor Research of Large Language Models through Structured Thought Leaps](https://mlsanity.com/paper/2410.10370/) (2024)
- [Bridging the Creativity Understanding Gap: Small-Scale Human Alignment Enables Expert-Level Humor Ranking in LLMs](https://mlsanity.com/paper/2502.20356/) (2025)
- [Collaborative Storytelling with Human Actors and AI Narrators](https://mlsanity.com/paper/2109.14728/) (2021)
- [Language Models can Subtly Deceive Without Lying: A Case Study on Strategic Phrasing in Legislation](https://mlsanity.com/paper/2405.04325/) (2024)
- [Can LLMs Simulate Social Media Engagement? A Study on Action-Guided Response Generation](https://mlsanity.com/paper/2502.12073/) (2025)
- [Pantheon 1.0, a manually verified dataset of globally famous biographies](https://mlsanity.com/paper/1502.07310/) (2015)
- [AI Humor Generation: Cognitive, Social and Creative Skills for Effective Humor](https://mlsanity.com/paper/2502.07981/) (2025)
- [Prompt to GPT-3: Step-by-Step Thinking Instructions for Humor Generation](https://mlsanity.com/paper/2306.13195/) (2023)
- [Engagement-Driven Content Generation with Large Language Models](https://mlsanity.com/paper/2411.13187/) (2024)
- [CleanComedy: Creating Friendly Humor through Generative Techniques](https://mlsanity.com/paper/2412.09203/) (2024)
- [Stereotype Detection in LLMs: A Multiclass, Explainable, and Benchmark-Driven Approach](https://mlsanity.com/paper/2404.01768/) (2024)

## Built on by (most cited first, 1 of 1)

- [Towards High-Level Semantic Intelligence](https://mlsanity.com/paper/2607.24082/) (2026)

## Abstract (arXiv)

Large language models are increasingly used for creative writing and engagement content, raising safety concerns about the outputs. Therefore, casting humor generation as a testbed, this work evaluates how funniness optimization in modern LLM pipelines couples with harmful content by jointly measuring humor, stereotypicality, and toxicity. This is further supplemented by analyzing incongruity signals through information-theoretic metrics. Across six models, we observe that harmful outputs receive higher humor scores which further increase under role-based prompting, indicating a bias amplification loop between generators and evaluators. Information-theoretic analyses show harmful cues widen predictive uncertainty and surprisingly, can even make harmful punchlines more expected for some models, suggesting structural embedding in learned humor distributions. External validation on an additional satire-generation task with human perceived funniness judgments shows that LLM satire increases stereotypicality and typically toxicity, including for closed models. Quantitatively, stereotypical/toxic jokes gain 10-21% in mean humor score, stereotypical jokes appear 11% to 28% more often among the jokes marked funny by LLM-based metric and up to 10% more often in generations perceived as funny by humans.