All the participants in Nesta’s event At The Roots of Collective Intelligence agreed that groups are better at reasoning than individuals. This may come as a surprise, but a wealth of studies show that argumentation works well.

Give people a tricky reasoning problem to complete, and most will fail. Let them talk with each other, and soon all will agree on the right answer.

Ask experts to make predictions, and most will be widely off the mark. Let them exchange arguments and their accuracy will improve dramatically. Probe citizens’ views on policy, and you will find that most are uninformed or misinformed. Get the same citizens to discuss the same issues with each other, and they will become more enlightened.

Provided some simple conditions are met—the presence of disagreement, the freedom to voice one’s opinion—argumentation consistently improves on individual reasoning. It does so by turning the flaws of individual reasoning into strengths. Individual reasoners mostly find arguments that support their preconceived beliefs. They do not examine critically their own reasons. When people reason on their own, they often become more confident and opinionated. By contrast, in a discussion, people are forced to contend with arguments for other points of view. Their poorly examined reasons are shot down. They are forced to think of better arguments.

Particularly critical to the success of argumentation is the back and forth of discussion. People start the discussion with relatively weak, generic arguments. Counter-arguments are raised. Counter-arguments are addressed. Argument quality improves. Even when the arguments are as clear cut, demonstrative as they come—logical or mathematical arguments—people are much more likely to be convinced in the course of a discussion than by exposition to a single argument.

This raises the issue of scaling up. Argumentation works best in groups of four individuals or fewer, when the back and forth of the discussion flows seamlessly. In such a context, the best ideas are most likely to spread. How do we scale this up to reach hundreds, thousands, millions of people?

Take vaccine hesitancy as an example. Giving vaccine hesitant people a message that support vaccination typically has little effect. By contrast, discussions with experts can address many of their concerns. Unfortunately, there aren’t enough experts to speak to each vaccine hesitant individual. How can we make sure more people get to argue their ideas with different minded interlocutors so as to ensure that better ideas are spread and adopted? This is a particularly critical question in the light of the debates on the relative homogeneity of friend composition in social media networks.

Collective Intelligence platforms such as Liquid Feedback, Assembl or DemocracyOS which include tools for tracking discussions, are implicitly trying to address this issue by making all threads more visible and traceable. The arguments and counterarguments produced by the participants become accessible and contribute to the deliberative process. While these discussion threads do not always contain all of the possible arguments, they do tend to provide multiple points of view that as we saw above contribute to better decision making.