Conversational Forecasting Across Large Human Groups Using A Network of Surrogate Agents
Human teams using AI-facilitated conversations predicted NBA games with 62% accuracy, beating Vegas odds.
A new study from researchers Louis Rosenberg, Hans Schumann, Ganesh Mani, and Gregg Willcox demonstrates how AI can significantly enhance group forecasting accuracy. Their paper, 'Conversational Forecasting Across Large Human Groups Using A Network of Surrogate Agents,' tested the Hyperchat AI architecture—a system that uses intervening AI agents to facilitate real-time conversations among distributed teams. In the experiment, groups of 25-30 basketball fans used the Thinkscape platform (powered by Hyperchat AI) to discuss and forecast 56 NBA games against the spread over 12 weeks. Each game received just five minutes of conversational deliberation.
The results were striking: these AI-facilitated groups achieved 62% accuracy in their predictions, statistically significant against the Vegas benchmark of 50% (p=0.059). Had participants wagered on these forecasts, they would have realized an 18% return on investment. The study found a strong correlation between conversation quality and accuracy—when excluding the bottom 25% of forecasts by conversation rate, accuracy jumped to 68% (p=0.017). This suggests that AI agents don't just enable large-scale conversations (previously tested with groups up to 250 people) but actively improve the quality of collective reasoning. The architecture supports text, voice, or video deliberations for complex tasks like brainstorming, risk assessment, and decision optimization.
- Hyperchat AI's Thinkscape platform enabled groups of 25-30 people to achieve 62% accuracy forecasting 56 NBA games, beating Vegas 50% odds
- The AI-facilitated conversations would have generated an 18% ROI over 12 weeks, with higher conversation rates correlating to 68% accuracy
- The system uses 'surrogate agents' to manage large-scale real-time deliberations, previously scaling to 250-person teams for complex problem-solving
Why It Matters
This demonstrates AI's potential to amplify collective intelligence for critical forecasting in finance, strategy, and risk assessment.