# The Illusion of Debiasing: Persona Steering Redistributes Rather Than Reduces Bias in LLMs

arXiv: https://arxiv.org/abs/2609.07117 · HTML: https://mlsanity.com/paper/2609.07117/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 13 of its 72 references that are in the map.
- First posted on arXiv 7 Sep 2026.

## Builds on (most cited first, 13 of 13)

- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [Equality of Opportunity in Supervised Learning](https://mlsanity.com/paper/1610.02413/) (2016)
- [Out of One, Many: Using Language Models to Simulate Human Samples](https://mlsanity.com/paper/2209.06899/) (2022)
- [BBQ: A Hand-Built Bias Benchmark for Question Answering](https://mlsanity.com/paper/2110.08193/) (2021)
- [Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies](https://mlsanity.com/paper/2208.10264/) (2022)
- [BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation](https://mlsanity.com/paper/2101.11718/) (2021)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://mlsanity.com/paper/2304.05335/) (2023)
- [Character-LLM: A Trainable Agent for Role-Playing](https://mlsanity.com/paper/2310.10158/) (2023)
- [Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs](https://mlsanity.com/paper/2311.04892/) (2023)
- [Investigating Implicit Bias in Large Language Models: A Large-Scale Study of Over 50 LLMs](https://mlsanity.com/paper/2410.12864/) (2024)
- [OpenCharacter: Training Customizable Role-Playing LLMs with Large-Scale Synthetic Personas](https://mlsanity.com/paper/2501.15427/) (2025)
- [Silenced Biases: The Dark Side LLMs Learned to Refuse](https://mlsanity.com/paper/2511.03369/) (2025)
- [Bias after Prompting: Persistent Discrimination in Large Language Models](https://mlsanity.com/paper/2509.08146/) (2025)

## Abstract (arXiv)

Prompt-based interventions: system prompts, personas, role instructions, reliably reshape what a language model says, but it is unclear which layer they reach. Do they reconfigure internal structure, or only modulate the output channel? We use persona conditioning as a controlled probe, measuring its effects along a depth axis from self-report, through open-ended generation, to word-level parametric association, across three instruction-tuned models. We find a graded dissociation. Personas are legible but not structural: models follow single-trait instructions yet fail to reproduce human inter-trait covariance. The dissociation deepens with depth: personas hold or amplify closed-form QA bias, shift absolute tone while leaving between-group disparity unchanged, and barely perturb an already saturated associative baseline. Prompt-based steering thus operates in the output channel and has a structural reach limit that surface manipulability can mask.