Quantifying Human-AI Synergy

A good team is more than the sum of its parts. This insight seems almost too elementary to lead anywhere new. Yet we keep bumping into the same principle: human groups whose intelligence exceeds their smartest member, human-AI teams that achieve more than either could alone, and multi-agent AI systems that coordinate to solve challenging tasks. Synergy is easy to claim but hard to measure: by definition it exists only jointly, so it vanishes the moment we test the parts in isolation. In this talk I explore collective intelligence across different scales and substrates---how we can quantify, explain, and steer it. Information theory gives us the language to make synergy precise. Across every substrate, one ingredient converts parts into a whole---not more capability, not more communication, but the capacity to model other minds. Each time the principle resurfaces, my suspicion grows: all intelligence is collective.