Archive / Language models
Language, Bias & Human Preferences
Two earlier projects examine different sides of language model behavior: what models pick up from linguistic form, and how people respond to what models say.
The shared question
Language is never just a string of tokens. Negation can change what a sentence means; empathy and demographics can change which response a person prefers. These projects made me curious about how model behavior and human judgment meet, and about the care needed when evaluating both.
Papers
The Negation Bias in Large Language Models: Investigating Bias Reflected in Linguistic Markers ↗COLM 2025
Investigating how negation and linguistic markers relate to bias in language models.
Your Mileage May Vary: How Empathy and Demographics Shape Human Preferences in LLM Responses ↗EMNLP Findings 2025
Studying variation in human preferences for LLM responses through empathy and demographics.