Research shows that crowds can provide more accurate estimates of uncertain quantities than individuals (Surowiecki, 2004). But little is known about how to organize crowd members to maximize accuracy. When should crowd members work independently, and when should they work collaboratively? We examined the effects of social influence on estimation accuracy, consensus, and confidence. Participants first made independent estimates of uncertain quantities, such as the percentage of U.S. deaths due to heart attacks or the height of the tallest building. Then, in some conditions, they interacted with others online. After the discussion, they made second independent estimates. Social interaction improved accuracy. Despite well‐known problems with groups, such as herding and free riding, discussion resulted in more accurate estimates and greater consensus relative to independent estimates. We offer a simple model that describes the process by which group discussion improves the estimates of uncertain quantities. Copyright © 2014 John Wiley & Sons, Ltd.