Humans Team Up With AI to Predict the Future — Does It Help?

In the Summer 2026 Metaculus Cup, a forecasting tournament, two human forecasters, Scott Eastman and Robert de Neufville, teamed up with an AI bot called Preseen to form a team named Symbiosis. Their goal was to see if human input could improve the bot's predictions about future events. They had four main ways to intervene, like adjusting the bot's answers or adding their own insights. The tournament was a test of whether combining human and machine intelligence leads to better forecasts than using AI alone.

The overall result was inconclusive. Symbiosis finished slightly ahead of the baseline bot, which is the bot without any human help. But the difference was small and could have been due to chance. So, the tournament doesn't give us solid evidence that human interventions reliably improved performance. This means we can't yet say that human-AI teams are better than AI working solo.

This matters because accurate forecasting is valuable in many areas, from business planning to government policy. If humans and AI can work together to make better predictions, it could lead to smarter decisions and fewer surprises. But if AI alone is just as good, then we might not need to spend time and money on human involvement. The study is a step toward understanding the strengths and weaknesses of both.

One important detail: the author of the retrospective works at Preseen and coauthored the analysis, so there's a potential bias. However, they were transparent about this. The takeaway is that human-AI collaboration in forecasting is still an open question. More research is needed to see if and when humans add value. For now, we shouldn't assume that adding people to AI always makes things better.