Participants in three experiments wrote gratitude letters and then predicted how surprised, happy, and awkward recipients would feel. Recipients then reported how receiving an expression of gratitude actually made them feel. Expressers significantly underestimated how surprised recipients would be about why expressers were grateful, overestimated... See more
"Violation of Expectation via Metacognitive Prompting Reduces Theory of Mind Prediction Error in Large Language Models"
This paper investigates how Violation of Expectation (VoE) and metacognitive prompting can be used to reduce Theory of Mind prediction errors in Large Language Models (LLMs) in the context of human-AI interaction.
Metacognitive Prompting
The overall goal is to create AI systems that can deeply understand and align with individual human users. To achieve this, the researchers propose using a learning technique inspired by cognitive science theories about how human brains work.… Show more
A fourth paper looks at spontaneous contacts, when we reach out to someone in our social circle who we may not have spoken to in a bit - to say “hi” or to catch up. We don’t do it enough, and (I know this will shock you!) the reason is because we think it will be akward and not appreciated. And, of course, as you might suspect, people really... See more
Greg Brockman’s unpublished chemistry textbook. Designed around problems.
artificial empathy is worthy of research. Millions of people have no one they can talk to about their problems. Millions of people rarely receive compassion from another human being. For these people, having access to an AI chatbot that can perform care for them may be an important contributor to their mental wellbeing. Esther Perel compares these... See more