Can LLMs predict demographics from social media usernames alone?
This explores whether web-browsing language models can infer personal attributes like gender, age, and political orientation from just a username and public profile. The finding matters because it reveals a privacy vulnerability that traditional API-based assumptions didn't anticipate.
Does chatbot personalization build trust or expose privacy risks? Do reasoning traces actually expose private user data? Why do LLM judges fail at predicting sparse user preferences? How do personalization granularity levels trade precision against scalability? Why do similar user profiles produce worse personalization errors?
Chatbot personalization increases trust and privacy risk simultaneously Reasoning traces leak private data primarily through direct recollection Persona sparsity makes LLM judges unreliable without verbal uncertainty filtering Personalization granularity involves distinct precision-scalability-data tradeoffs across three levels Similar user profiles produce worse personalization errors than dissimilar ones
Do personality inferences from text show the same demographic biases as norm predictions? Can post-hoc reranking improve fairness for demographic minorities in shared accounts? Can LLMs infer psychological profiles without explicit user disclosure? Which personalization techniques expose user data most directly? Why do language models infer political orientation from seemingly innocuous user signals? How does data scarcity in user populations amplify persona similarity errors? What social information is missing from language data? What data types carry the most privacy risk in personalization systems? Why do feature-based approaches struggle when privacy or latent factors are involved? Why do sparse user profiles trigger stereotype-driven demographic predictions? What governance safeguards could constrain misuse of demographic inference? How does direct web access change privacy assumptions built on API limits?See all 16 inquiring lines on this note →