mlsanityData to 14 Oct 2026

One Persona, Many Cues, Different Results: How Sociodemographic Cues Impact LLM Personalization

arXiv 2601.18572 ↗First posted 26 Jan 2026cs.CLcited by 9

Not on the map. Why?

Abstract (arXiv)

Personalization of LLMs by sociodemographic subgroup often improves user experience, but can also introduce or amplify biases and unfair outcomes across groups. Prior work has employed so-called personas, sociodemographic user attributes conveyed to a model, to study bias in LLMs by relying on a single cue to prompt a persona, such as user names or explicit attribute mentions. This disregards LLM sensitivity to prompt variation and the rarity of some cues in real interactions (external validity). We compare six commonly used persona cues across seven open and proprietary LLMs on four writing and advice tasks. While cues are overall highly correlated, they produce substantial variance in responses across personas that can change findings on persona-induced differences and bias. We therefore caution against claims based on single persona cues, especially when they are overly explicit and have low external validity.

Key facts. Cited by 9 papers in the map (data through 14 Oct 2026). Builds on 38 of its 75 references that are in the map. First posted on arXiv 26 Jan 2026.

Builds on

38 of 75 references.

References · 38

  1. 2024Qwen2.5 Technical Reportcited by 4,942
  2. 2023Towards Measuring the Representation of Subjective Global Opinions in Language Modelscited by 437
  3. 2021AI and the Everything in the Whole Wide World Benchmarkcited by 349
  4. 2023Marked Personas: Using Natural Language Prompts to Measure Stereotypes in Language Modelscited by 207
  5. 2024Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Modelscited by 162
  6. 2023Evaluating and Mitigating Discrimination in Language Model Decisionscited by 124
  7. 2025Measuring what Matters: Construct Validity in Large Language Model Benchmarkscited by 119
  8. 2024Democratizing Large Language Models via Personalized Parameter-Efficient Fine-tuningcited by 90
  9. 2025PersonaMem-v2: Towards Personalized Intelligence via Learning Implicit User Personas and Agentic Memorycited by 88
  10. 2024Beyond prompt brittleness: Evaluating the reliability and consistency of political worldviews in LLMscited by 55
  11. 2024Designing a Dashboard for Transparency and Control of Conversational AIcited by 46
  12. 2024Stereotype or Personalization? User Identity Biases Chatbot Recommendationscited by 34
26 more
  1. 2024"You Gotta be a Doctor, Lin": An Investigation of Name-Based Bias of Large Language Models in Employment Recommendationscited by 33
  2. 2023Exploring the Sensitivity of LLMs' Decision-Making Capabilities: Insights from Prompt Variation and Hyperparameterscited by 32
  3. 2024Exploring Safety-Utility Trade-Offs in Personalized Language Modelscited by 27
  4. 2025IssueBench: Millions of Realistic Prompts for Measuring Issue Bias in LLM Writing Assistancecited by 26
  5. 2025Reading Between the Prompts: How Stereotypes Shape LLM's Implicit Personalizationcited by 24
  6. 2024Modeling Human Subjectivity in LLMs Using Explicit and Implicit Human Factors in Personascited by 23
  7. 2024Stop! In the Name of Flaws: Disentangling Personal Names and Sociodemographic Attributes in NLPcited by 22
  8. 2025Presumed Cultural Identity: How Names Shape LLM Responsescited by 21
  9. 2023Aligning Large Language Models with Human Opinions through Persona Selection and Value--Belief--Norm Reasoningcited by 19
  10. 2025Persona-Augmented Benchmarking: Evaluating LLMs Across Diverse Writing Stylescited by 16
  11. 2024LLM Targeted Underperformance Disproportionately Impacts Vulnerable Userscited by 15
  12. 2025The AI Gap: How Socioeconomic Status Affects Language Technology Interactionscited by 15
  13. 2026The Need for a Socially-Grounded Persona Framework for User Simulationcited by 15
  14. 2025Whose Personae? Synthetic Persona Experiments in LLM Research and Pathways to Transparencycited by 14
  15. 2022Monant Medical Misinformation Dataset: Mapping Articles to Fact-Checked Claimscited by 10
  16. 2025No for Some, Yes for Others: Persona Prompts and Other Sources of False Refusal in Language Modelscited by 9
  17. 2025Missing the Margins: A Systematic Literature Review on the Demographic Representativeness of LLMscited by 9
  18. 2025A Tale of Two Identities: An Ethical Audit of Human and AI-Crafted Personascited by 8
  19. 2025Language Models Change Facts Based on the Way You Talkcited by 8
  20. 2025Measuring Bias or Measuring the Task: Understanding the Brittle Nature of LLM Gender Biasescited by 8
  21. 2025The Inadequacy of Offline LLM Evaluations: A Need to Account for Personalization in Model Behaviorcited by 8
  22. 2025Rethinking LLM Bias Probing Using Lessons from the Social Sciencescited by 7
  23. 2025Colombian Waitresses y Jueces canadienses: Gender and Country Biases in Occupation Recommendations from LLMscited by 7
  24. 2025Does the Prompt-based Large Language Model Recognize Students' Demographics and Introduce Bias in Essay Scoring?cited by 6
  25. 2025Benchmarking and Improving LLM Robustness for Personalized Generationcited by 4
  26. 2025Where Should I Study? Biased Language Models Decide! Evaluating Fairness in LMs for Academic Recommendationscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 9.

9 papers.

  1. 2026Affective Context Amplifies Sycophancy in LLM Responsescited by 3
  2. 2026Topics as Proxies for Sociodemographics: How Conversational Context Affects LLM Answerscited by 2
  3. 2026FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocationcited by 2
  4. 2026When Persona Attributes Improve Population Alignment in Large Language Modelscited by 2
  5. 2026Whose Name Comes Up? III: Persona Prompting Effects in LLM-Based Scholar Recommendationcited by 1
  6. 2026ImplicitBBQ: Benchmarking Implicit Bias in Large Language Models through Characteristic Based Cuescited by 0
  7. 2026Creating Multilingual Mental Health Dialogue Datasets: Limits of Persona-Based Localization via Nationality and Languagecited by 0
  8. 2026WildSEEK: Evaluating Language Models for Information-Seekingcited by 0
  9. 2026User Model Extraction via Belief Self-Distillationcited by 0