# Measuring Mechanistic Independence: Can Bias Be Removed Without Erasing Demographics?

arXiv: https://arxiv.org/abs/2512.20796 · HTML: https://mlsanity.com/paper/2512.20796/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 32 references that are in the map.
- First posted on arXiv 23 Dec 2025.

## Builds on (most cited first, 25 of 25)

- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Language (Technology) is Power: A Critical Survey of "Bias" in NLP](https://mlsanity.com/paper/2005.14050/) (2020)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [BBQ: A Hand-Built Bias Benchmark for Question Answering](https://mlsanity.com/paper/2110.08193/) (2021)
- [Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods](https://mlsanity.com/paper/1804.06876/) (2018)
- [Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://mlsanity.com/paper/2203.14680/) (2022)
- [Gender Bias in Coreference Resolution](https://mlsanity.com/paper/1804.09301/) (2018)
- [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](https://mlsanity.com/paper/2403.19647/) (2024)
- [BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation](https://mlsanity.com/paper/2101.11718/) (2021)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP](https://mlsanity.com/paper/2103.00453/) (2021)
- [Direct and Indirect Effects](https://mlsanity.com/paper/1301.2300/) (2013)
- [Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Them](https://mlsanity.com/paper/1903.03862/) (2019)
- [Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms](https://mlsanity.com/paper/2403.17806/) (2024)
- [Causal Intervention for Leveraging Popularity Bias in Recommendation](https://mlsanity.com/paper/2105.06067/) (2021)
- [SAEs Are Good for Steering -- If You Select the Right Features](https://mlsanity.com/paper/2505.20063/) (2025)
- [NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals](https://mlsanity.com/paper/2407.14561/) (2024)
- [Social Bias Benchmark for Generation: A Comparison of Generation and QA-Based Evaluations](https://mlsanity.com/paper/2503.06987/) (2025)
- [Aligned but Blind: Alignment Increases Implicit Bias by Reducing Awareness of Race](https://mlsanity.com/paper/2506.00253/) (2025)
- [Dissecting Bias in LLMs: A Mechanistic Interpretability Perspective](https://mlsanity.com/paper/2506.05166/) (2025)

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

- [Topics as Proxies for Sociodemographics: How Conversational Context Affects LLM Answers](https://mlsanity.com/paper/2606.02776/) (2026)

## Abstract (arXiv)

We investigate how independent demographic bias mechanisms are from general demographic recognition in language models. Using a multi-task evaluation setup where demographics are associated with names, professions, and education levels, we measure whether models can be debiased while preserving demographic detection capabilities. We compare attribution-based and correlation-based methods for locating bias features. We find that targeted sparse autoencoder feature ablations in Gemma-2-9B reduce bias without degrading recognition performance: attribution-based ablations mitigate race and gender profession stereotypes while preserving name recognition accuracy, whereas correlation-based ablations are more effective for education bias. Qualitative analysis further reveals that removing attribution features in education tasks induces ``prior collapse'', thus increasing overall bias. This highlights the need for dimension-specific interventions. Overall, our results show that demographic bias arises from task-specific mechanisms rather than absolute demographic markers, and that mechanistic inference-time interventions can enable surgical debiasing without compromising core model capabilities.