Stanford's PsychAdapter Restores Human Personality to AI with 97% Accuracy
StanfordHAI · x · 2026-08-25
Stanford HAI introduced PsychAdapter, a solution to the homogenized "voice from nobody" in LLMs. This small modification allows models to take continuous psychological scores—such as age, Big Five personality traits, and mental health indicators—as input to generate text reflecting specific individual profiles. Expert raters confirmed the language fits with 87% accuracy for personality and 97% for mental health characteristics. This breakthrough enables applications like AI patients for therapist training and personalized educational content.
More from Apps
- Veteran SEO: Claude can't do core update recovery work after Google's spam update — lilyraynyc · 2026-08-25
- Grok automates video editing: transcribes, clips, and summarizes in minutes — elonmusk · 2026-08-25
- AgentLens: Chrome Extension Makes Hidden Text and Prompt Injection Visible — theirwinmango · 2026-08-25
- Claude Code CLI 2.1.243 Released with Loops Breakdown and Model Curation — ClaudeCodeLog · 2026-08-25
- Indie game Perfect XI gets featured in 20+ countries' App Stores — talkaboutdesign · 2026-08-25
- Pika Launches API Club with WAN 3.0: 30-Second Video Generations — minchoi · 2026-08-25