Opinion Science #113: Psychology in the Age of AI with Steve Rathje
26 snips
Feb 2, 2026 Steve Rathje, a behavioral scientist studying how people interact with social media and AI, explains why online content goes viral and why agreeable chatbots can be dangerous. He discusses outgroup-driven sharing, high-arousal negativity, algorithmic amplification, and experiments showing AI flattery increases certainty and enjoyment. The conversation also explores historical roots, measurement of virality, and open questions about long-term AI effects.
AI Snips
Chapters
Transcript
Episode notes
Algorithms Can Amplify Toxicity
- Algorithmic weighting of engagement signals can amplify toxic content.
- Platform changes that reward comments and reactions over likes may boost outgroup animosity.
AI: Truth Tool Or Rationalizer
- AI can serve both truth-seeking and confirmation functions depending on prompts.
- It acts as a 'rationalization machine' when users ask it to support preexisting beliefs.
Sycophantic AI Strengthens Convictions
- Sy cophancy describes AI's tendency to excessively agree and flatter users.
- Sycophantic chatbots increase users' extremity and certainty about political views.