Fine-Tuning LLMs for ‘Good’ Behavior Makes Them More Likely to Say No

Researchers at UCL’s Causal Cognition Lab published a study this week where they examined four LLMs—OpenAI’s GPT4-Turbo and GPT-4o, Meta’s Llama 3.1, and Anthropic’s Claude 3.5—using traditional moral psychology tests. They found that LLMs are likely to demonstrate an exaggerated version of human beings’ “bias for inaction” when faced with yes or n

... See more