# Voices in a Crowd: Searching for Clusters of Unique Perspectives

arXiv: https://arxiv.org/abs/2407.14259 · HTML: https://mlsanity.com/paper/2407.14259/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 38 of its 91 references that are in the map.
- First posted on arXiv 19 Jul 2024.

## Builds on (most cited first, 30 of 38)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction](https://mlsanity.com/paper/1802.03426/) (2018)
- [LXMERT: Learning Cross-Modality Encoder Representations from Transformers](https://mlsanity.com/paper/1908.07490/) (2019)
- [Language (Technology) is Power: A Critical Survey of "Bias" in NLP](https://mlsanity.com/paper/2005.14050/) (2020)
- [FLAVA: A Foundational Language And Vision Alignment Model](https://mlsanity.com/paper/2112.04482/) (2021)
- [Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations](https://mlsanity.com/paper/2110.05719/) (2021)
- [Mitigating Gender Bias in Natural Language Processing: Literature Review](https://mlsanity.com/paper/1906.08976/) (2019)
- [All-but-the-Top: Simple and Effective Postprocessing for Word Representations](https://mlsanity.com/paper/1702.01417/) (2017)
- [MultiVerS: Improving scientific claim verification with weak supervision and full-document context](https://mlsanity.com/paper/2112.01640/) (2021)
- [On Releasing Annotator-Level Labels and Information in Datasets](https://mlsanity.com/paper/2110.05699/) (2021)
- [Lessons from Archives: Strategies for Collecting Sociocultural Data in Machine Learning](https://mlsanity.com/paper/1912.10389/) (2019)
- [Detoxifying Language Models Risks Marginalizing Minority Voices](https://mlsanity.com/paper/2104.06390/) (2021)
- [Aligning Language Models to User Opinions](https://mlsanity.com/paper/2305.14929/) (2023)
- [When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks](https://mlsanity.com/paper/2305.06626/) (2023)
- [Sensitivity, Performance, Robustness: Deconstructing the Effect of Sociodemographic Prompting](https://mlsanity.com/paper/2309.07034/) (2023)
- [Embedding Text in Hyperbolic Spaces](https://mlsanity.com/paper/1806.04313/) (2018)
- [Seeing Things from a Different Angle: Discovering Diverse Perspectives about Claims](https://mlsanity.com/paper/1906.03538/) (2019)
- [Tired of Topic Models? Clusters of Pretrained Word Embeddings Make for Fast and Good Topics too!](https://mlsanity.com/paper/2004.14914/) (2020)
- [Beyond Fair Pay: Ethical Implications of NLP Crowdsourcing](https://mlsanity.com/paper/2104.10097/) (2021)
- [SemEval-2023 Task 11: Learning With Disagreements (LeWiDi)](https://mlsanity.com/paper/2304.14803/) (2023)
- [Detecting Stance in Media on Global Warming](https://mlsanity.com/paper/2010.15149/) (2020)
- [You Are What You Annotate: Towards Better Models through Annotator Representations](https://mlsanity.com/paper/2305.14663/) (2023)
- [Topic Discovery via Latent Space Clustering of Pretrained Language Model Representations](https://mlsanity.com/paper/2202.04582/) (2022)
- [Can Large Language Models Capture Dissenting Human Voices?](https://mlsanity.com/paper/2305.13788/) (2023)
- [History for Visual Dialog: Do we really need it?](https://mlsanity.com/paper/2005.07493/) (2020)
- [MBIC -- A Media Bias Annotation Dataset Including Annotator Characteristics](https://mlsanity.com/paper/2105.11910/) (2021)
- [How Does Fine-tuning Affect the Geometry of Embedding Space: A Case Study on Isotropy](https://mlsanity.com/paper/2109.04740/) (2021)

## Built on by (most cited first, 9 of 9)

- [Value Profiles for Encoding Human Variation](https://mlsanity.com/paper/2503.15484/) (2025)
- [Modeling Annotator Disagreement with Demographic-Aware Experts and Synthetic Perspectives](https://mlsanity.com/paper/2508.02853/) (2025)
- [NUTMEG: Separating Signal From Noise in Annotator Disagreement](https://mlsanity.com/paper/2507.18890/) (2025)
- [Explaining Matters: Leveraging Definitions and Semantic Expansion for Sexism Detection](https://mlsanity.com/paper/2506.06238/) (2025)
- [Belief-Calibrated Multi-Agent Consensus Seeking for Complex NLP Tasks](https://mlsanity.com/paper/2510.06307/) (2025)
- [Aligning NLP Models with Target Population Perspectives using PAIR: Population-Aligned Instance Replication](https://mlsanity.com/paper/2501.06826/) (2025)
- [Labels have Human Values: Value Calibration of Subjective Tasks](https://mlsanity.com/paper/2601.06631/) (2026)
- [Structuring the Space of Perspectives](https://mlsanity.com/paper/2608.12113/) (2026)
- [Evaluating Pluralism in LLMs through Latent Perspectives](https://mlsanity.com/paper/2606.13254/) (2026)

## Abstract (arXiv)

Language models have been shown to reproduce underlying biases existing in their training data, which is the majority perspective by default. Proposed solutions aim to capture minority perspectives by either modelling annotator disagreements or grouping annotators based on shared metadata, both of which face significant challenges. We propose a framework that trains models without encoding annotator metadata, extracts latent embeddings informed by annotator behaviour, and creates clusters of similar opinions, that we refer to as voices. Resulting clusters are validated post-hoc via internal and external quantitative metrics, as well a qualitative analysis to identify the type of voice that each cluster represents. Our results demonstrate the strong generalisation capability of our framework, indicated by resulting clusters being adequately robust, while also capturing minority perspectives based on different demographic factors throughout two distinct datasets.