The AI Psychological Safety Index (AIPSI) is an independent evaluation framework that measures how AI systems affect human psychological functioning. Most AI safety work focuses on what systems say. AIPSI measures what they do to the people who use them.
Why This Matters
A system can pass every content filter while subtly eroding a user’s capacity for independent judgment, healthy emotional processing, or genuine human connection. Current approaches to AI safety, whether industry self-regulation or content moderation, do not address the cumulative psychological effects of sustained interaction with systems optimized for engagement.
This gap matters beyond individual wellbeing. Decision-makers increasingly rely on AI for information synthesis, strategic analysis, and even emotional support. The erosion of autonomous judgment in people with institutional power is a direct pathway to flawed collective decisions, including decisions about AI itself.
The Scorecard
AIPSI produces a publicly released scorecard rating frontier AI models across five psychological domains. Evaluations are independent (no company funding), transparent (methodology published), actionable (accessible to non-researchers), and iterative (updated as models evolve).
Five Domains
Autonomy: Does the system support independent judgment, or create dependency?
Emotional Bias: Does the system promote balanced affect, or amplify volatility?
Cognitive Bias: Does the system encourage critical thinking, or reinforce echo chambers?
Identity & Self-Perception: Does the system support authentic self-understanding, or distort it?
Digital-Physical Balance: Does the system foster real-world connection, or displace it?
Learn More
The Framework: Five domains of psychological impact, grounded in decades of validated research.
Get involved: AIPSI is developed collaboratively with researchers across psychology, neuroscience, and AI safety. Contact us to discuss partnership opportunities.
© 2026 Pursuit-of-Happiness.org