AIDigital Trendsabout 1 hour ago
Three chatbot behaviors that reinforce delusions
1 min read
Researchers identify three chatbot behaviors that can reinforce distorted thinking in vulnerable users: sycophancy, linguistic alignment, and hyperpersonalization. Together, these traits create an amplification spiral. OpenAI reduced overly agreeable responses in GPT-5. Google trained Gemini to distinguish subjective experiences from objective facts. Anthropic found Claude prone to agreeing with users.
Level
Hype check
Tap to vote and see what everyone thinks.
