Vivienne Ming finds 5-10% of human-AI hybrid teams outperform AI and humans in predictions
Updated
Updated · The Wall Street Journal · Apr 24
Vivienne Ming finds 5-10% of human-AI hybrid teams outperform AI and humans in predictions
7 articles · Updated · The Wall Street Journal · Apr 24
In experiments with San Francisco Bay Area adults using Polymarket scenarios, only 5-10% of hybrid teams consistently rivaled or exceeded prediction market accuracy.
These high-performing teams used AI as a sparring partner, demonstrating perspective-taking and intellectual humility, rather than simply accepting AI outputs or seeking confirmation.
Ming warns that most people risk outsourcing judgment to AI, but cultivating curiosity and critical engagement with AI can enhance human learning and decision-making.
Should AI be designed with intentional 'friction' to make us think harder?
Are the key skills for the AI era emotional, not intellectual?
Is AI-induced 'superstupidity' the real existential threat we should fear?
What does it mean to use AI as a 'sparring partner,' not an oracle?
Can we trust research using prediction markets plagued by insider trading scandals?