Others have shown that people benefit from being shown numeric information, both because it increases their ability to understand information and because it increases desirable behaviors. In my own work, I was interested in how best to present numeric information. One large randomized trial showed that percentages (e.g., 10% chance) rather than frequencies (e.g., 10 out of 100 people) are best, especially when they are accompanied with descriptive labels (e.g., moderate risk). When drug side effect likelihoods were formatted this way, people correctly reported the information from a table 95% of the time (compared to 81% for other formats), and their willingness to take the drug increased to 3.3 from 2.9, on a 7 point scale. In another series of studies, I showed that it is the reference class that matters (10% of people vs. 10% chance matters), not the numeric format (10% of people vs. 10 of 100 people does not matter).

Check out my paper and poster on this topic.