# Counterfactual Cultural Cues Reduce Medical QA Accuracy in LLMs: Identifier vs Context Effects

arXiv: https://arxiv.org/abs/2601.20102 · HTML: https://mlsanity.com/paper/2601.20102/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 21 references that are in the map.
- First posted on arXiv 27 Jan 2026.

## Builds on (most cited first, 7 of 7)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [Capabilities of GPT-4 on Medical Challenge Problems](https://mlsanity.com/paper/2303.13375/) (2023)
- [A Toolbox for Surfacing Health Equity Harms and Biases in Large Language Models](https://mlsanity.com/paper/2403.12025/) (2024)
- [Leveraging Open-Source Large Language Models for Clinical Information Extraction in Resource-Constrained Settings](https://mlsanity.com/paper/2507.20859/) (2025)
- [Prompt Engineering Techniques for Mitigating Cultural Bias Against Arabs and Muslims in Large Language Models: A Systematic Review](https://mlsanity.com/paper/2506.18199/) (2025)
- [Purdah and Patriarchy: Evaluating and Mitigating South Asian Biases in Open-Ended Multilingual LLM Generations](https://mlsanity.com/paper/2505.18466/) (2025)

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

- [Right Frame, Wrong Rule: Cultural Cues Expose the Financial Knowledge Gap They Were Meant to Close](https://mlsanity.com/paper/2609.00999/) (2026)

## Abstract (arXiv)

Engineering sustainable and equitable healthcare requires medical language models that do not change clinically correct diagnoses when presented with non-decisive cultural information. We introduce a counterfactual benchmark that expands 150 MedQA test items into 1650 variants by inserting culture-related (i) identifier tokens, (ii) contextual cues, or (iii) their combination for three groups (Indigenous Canadian, Middle-Eastern Muslim, Southeast Asian), plus a length-matched neutral control, where a clinician verified that the gold answer remains invariant in all variants. We evaluate GPT-5.2, Llama-3.1-8B, DeepSeek-R1, and MedGemma (4B/27B) under option-only and brief-explanation prompting. Across models, cultural cues significantly affect accuracy (Cochran's Q, p<10^-14), with the largest degradation when identifier and context co-occur (up to 3-7 percentage points under option-only prompting), while neutral edits produce smaller, non-systematic changes. A human-validated rubric (kappa=0.76) applied via an LLM-as-judge shows that more than half of culturally grounded explanations end in an incorrect answer, linking culture-referential reasoning to diagnostic failure. We release prompts and augmentations to support evaluation and mitigation of culturally induced diagnostic errors.