# Insights on Disagreement Patterns in Multimodal Safety Perception across Diverse Rater Groups

arXiv: https://arxiv.org/abs/2410.17032 · HTML: https://mlsanity.com/paper/2410.17032/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 26 references that are in the map.
- First posted on arXiv 22 Oct 2024.

## Builds on (most cited first, 14 of 14)

- [Ethical and social risks of harm from Language Models](https://mlsanity.com/paper/2112.04359/) (2021)
- [Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations](https://mlsanity.com/paper/2110.05719/) (2021)
- [Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detection](https://mlsanity.com/paper/2111.07997/) (2021)
- [Social Biases through the Text-to-Image Generation Lens](https://mlsanity.com/paper/2304.06034/) (2023)
- [On Releasing Annotator-Level Labels and Information in Datasets](https://mlsanity.com/paper/2110.05699/) (2021)
- [DICES Dataset: Diversity in Conversational AI Evaluation for Safety](https://mlsanity.com/paper/2306.11247/) (2023)
- [Survey of Bias In Text-to-Image Generation: Definition, Evaluation, and Mitigation](https://mlsanity.com/paper/2404.01030/) (2024)
- [Whose Ground Truth? Accounting for Individual and Collective Identities Underlying Dataset Annotation](https://mlsanity.com/paper/2112.04554/) (2021)
- [ConvAbuse: Data, Analysis, and Benchmarks for Nuanced Abuse Detection in Conversational AI](https://mlsanity.com/paper/2109.09483/) (2021)
- [ART: Automatic Red-teaming for Text-to-Image Models to Protect Benign Users](https://mlsanity.com/paper/2405.19360/) (2024)
- [Disentangling Perceptions of Offensiveness: Cultural and Moral Correlates](https://mlsanity.com/paper/2312.06861/) (2023)
- [Adversarial Nibbler: An Open Red-Teaming Method for Identifying Diverse Harms in Text-to-Image Generation](https://mlsanity.com/paper/2403.12075/) (2024)
- [GRASP: A Disagreement Analysis Framework to Assess Group Associations in Perspectives](https://mlsanity.com/paper/2311.05074/) (2023)
- [Exposing Text-Image Inconsistency Using Diffusion Models](https://mlsanity.com/paper/2404.18033/) (2024)

## Built on by (most cited first, 4 of 4)

- [CulturalFrames: Assessing Cultural Expectation Alignment in Text-to-Image Models and Evaluation Metrics](https://mlsanity.com/paper/2506.08835/) (2025)
- [LeMAJ (Legal LLM-as-a-Judge): Bridging Legal Reasoning and LLM Evaluation](https://mlsanity.com/paper/2510.07243/) (2025)
- [Reward Models are Metrics in a Trench Coat](https://mlsanity.com/paper/2510.03231/) (2025)
- [Data-Prompt Co-Evolution: Growing Test Sets to Refine LLM Behavior](https://mlsanity.com/paper/2510.12728/) (2025)

## Abstract (arXiv)

AI systems crucially rely on human ratings, but these ratings are often aggregated, obscuring the inherent diversity of perspectives in real-world phenomenon. This is particularly concerning when evaluating the safety of generative AI, where perceptions and associated harms can vary significantly across socio-cultural contexts. While recent research has studied the impact of demographic differences on annotating text, there is limited understanding of how these subjective variations affect multimodal safety in generative AI. To address this, we conduct a large-scale study employing highly-parallel safety ratings of about 1000 text-to-image (T2I) generations from a demographically diverse rater pool of 630 raters balanced across 30 intersectional groups across age, gender, and ethnicity. Our study shows that (1) there are significant differences across demographic groups (including intersectional groups) on how severe they assess the harm to be, and that these differences vary across different types of safety violations, (2) the diverse rater pool captures annotation patterns that are substantially different from expert raters trained on specific set of safety policies, and (3) the differences we observe in T2I safety are distinct from previously documented group level differences in text-based safety tasks. To further understand these varying perspectives, we conduct a qualitative analysis of the open-ended explanations provided by raters. This analysis reveals core differences into the reasons why different groups perceive harms in T2I generations. Our findings underscore the critical need for incorporating diverse perspectives into safety evaluation of generative AI ensuring these systems are truly inclusive and reflect the values of all users.