The tendency for humans to conform to a group consensus, known as groupthink, is a major pitfall in forecasting and decision-making. A recent competition, however, has revealed a potential nemesis for this common bias: artificial intelligence. An AI from British startup ManticAI placed eighth in the contest, not by following the crowd, but often by standing apart from it.
According to ManticAI’s co-founder, Toby Shevlane, their system’s predictions were often “more original” because it “strongly disagreed” with the average predictions made by human participants. This independence from the “community average” is a key feature that could make AI an invaluable tool for any organization looking to make more robust decisions.
The AI, which competed in the Metaculus Cup, derives its independent streak from its fundamental design. It is not influenced by reputation, social pressure, or the desire to conform. Its conclusions are the product of a systematic, data-driven analysis performed by a team of different AI agents, each tackling the problem from a unique angle. This process allows it to arrive at conclusions that might be unpopular or overlooked by a human group.
Its high ranking proves this isn’t just contrarianism for its own sake; its independent analyses were frequently correct. This suggests AI could be deployed as a “red team” or a designated dissenter in strategic planning sessions, forcing human teams to confront alternative viewpoints and stress-test their own assumptions.
While AI is not infallible, its inherent objectivity and freedom from social cognitive biases make it a powerful ally in the fight against flawed, consensus-driven decision-making. The performance of ManticAI is a glimpse into a future where AI helps protect us from the pitfalls of our own human nature.