Harvard economist David Deming discusses his research on measuring 'soft skills' like teamwork and management, revealing that tests such as emotion recognition and resource allocation tasks can predict leadership success better than self-nomination or interviews. He also finds that those who most want to be managers often perform worse than randomly assigned individuals, and that these management skills transfer to effectively overseeing AI agents.
Summarized by Podsumo
People who volunteer to be managers often perform worse than those assigned randomly, showing overconfidence bias.
A simple 'Reading the Mind in the Eyes' test (emotion recognition) predicts team performance better than IQ or demographics.
A custom allocative efficiency test—assigning workers to roles—predicts real-world manager success, reducing stockouts and improving cashier throughput.
Management skills for humans and AI agents are highly correlated (r=0.81), suggesting the same core abilities apply to both.
Good managers ask the right questions and facilitate conversation, rather than dominating the discussion.
"The managers who wanted to be managers were worse than randomly assigned. — David Deming"
"If you cannot correctly identify people's emotions, you probably won't be good in a team. — David Deming"
"The same approach that works for measuring hard skills can actually work in the softer, more people-oriented domain. It just requires a little bit of thoughtfulness. — David Deming"