SOMtime the World Ain$'$t Fair: Violating Fairness Using Self-Organizing Maps
arXiv:2602.18201v1 Announce Type: new Abstract: Unsupervised representations are widely assumed to be neutral with respect to sensitive attributes when those attributes are withheld from training. We show that this assumption is false. Using SOMtime, a topology-preserving representation method based on high-capacity Self-Organizing Maps, we demonstrate that sensitive attributes such as age and income emerge as dominant latent axes in purely unsupervised embeddings, even when explicitly excluded from the input. On two large-scale real-world datasets (the World Values Survey across five countries and the Census-Income dataset), SOMtime recovers monotonic orderings aligned with withheld sensitive attributes, achieving Spearman correlations of up to 0.85, whereas PCA and UMAP typically remain below 0.23 (with a single exception reaching 0.31), and against t-SNE and autoencoders which achieve at most 0.34. Furthermore, unsupervised segmentation of SOMtime embeddings produces demographicall
arXiv:2602.18201v1 Announce Type: new Abstract: Unsupervised representations are widely assumed to be neutral with respect to sensitive attributes when those attributes are withheld from training. We show that this assumption is false. Using SOMtime, a topology-preserving representation method based on high-capacity Self-Organizing Maps, we demonstrate that sensitive attributes such as age and income emerge as dominant latent axes in purely unsupervised embeddings, even when explicitly excluded from the input. On two large-scale real-world datasets (the World Values Survey across five countries and the Census-Income dataset), SOMtime recovers monotonic orderings aligned with withheld sensitive attributes, achieving Spearman correlations of up to 0.85, whereas PCA and UMAP typically remain below 0.23 (with a single exception reaching 0.31), and against t-SNE and autoencoders which achieve at most 0.34. Furthermore, unsupervised segmentation of SOMtime embeddings produces demographically skewed clusters, demonstrating downstream fairness risks without any supervised task. These findings establish that \textit{fairness through unawareness} fails at the representation level for ordinal sensitive attributes and that fairness auditing must extend to unsupervised components of machine learning pipelines. We have made the code available at~ https://github.com/JosephBingham/SOMtime
Executive Summary
This article challenges the assumption that unsupervised representations are neutral with respect to sensitive attributes. Using Self-Organizing Maps, the authors demonstrate that sensitive attributes such as age and income can emerge as dominant latent axes in unsupervised embeddings, even when excluded from the input. The findings highlight the risk of fairness violations in machine learning pipelines and the need for fairness auditing in unsupervised components.
Key Points
- ▸ Unsupervised representations can recover sensitive attributes such as age and income
- ▸ SOMtime outperforms other dimensionality reduction methods in recovering monotonic orderings
- ▸ Unsupervised segmentation of SOMtime embeddings can produce demographically skewed clusters
Merits
Methodological Contribution
The article introduces a novel method, SOMtime, which can effectively recover sensitive attributes in unsupervised embeddings.
Empirical Evidence
The authors provide comprehensive empirical evidence using two large-scale real-world datasets to support their claims.
Demerits
Limited Generalizability
The study focuses on two specific datasets, and the findings may not generalize to other domains or datasets.
Expert Commentary
The article's findings have significant implications for the development and deployment of machine learning systems. The fact that unsupervised representations can recover sensitive attributes highlights the need for careful consideration of fairness and bias in machine learning pipelines. The study's methodological contribution, introducing SOMtime, provides a valuable tool for researchers and practitioners to audit and mitigate fairness violations in unsupervised components. However, further research is needed to fully understand the generalizability of these findings and to develop effective strategies for ensuring fairness in machine learning.
Recommendations
- ✓ Develop and implement fairness auditing tools for unsupervised components of machine learning pipelines
- ✓ Prioritize transparency and explainability in machine learning practices to mitigate the risk of fairness violations