Demographic inequality emerges from text alone in language model recommendations
Research report. Peer-review status is not recorded in the source records below; consult the linked publication.
Researchers found that demographic inequality emerges from text alone when large language models rank and recommend online content.
Platform metadata generally amplified demographic inequalities in recommendations, while explicitly disclosing authors' demographic characteristics produced little additional change.
Framing described as controversial intensified both content and demographic skew in recommendations, whereas popularity-oriented framing dampened it, although for some attributes like age the objective could also reverse which group was favored.
Sources behind this report
Analysis
Why it matters
Demographic inequality in whose voices are surfaced emerges from text alone.
The context
A controlled experiment across three LLM providers, six objectives, and five levels of contextual information used 10,000 posts written by 1,000 survey respondents linked to verified self-reported characteristics across nine demographic attributes.
Report history
No earlier revision dates are retained here. This does not establish that the report has never changed; read any dated updates and corrections.
Corrections
No corrections have been appended.