Do Reputation Concerns Make Behavioral Biases Disappear? The Conjunction Fallacy on Facebook and Mechanical Turk

This paper reports the results of experiments designed to test whether individuals interacting on Facebook are more likely to succumb to the conjunction fallacy when they post their answers publicly and are exposed to the answers of others. Using the experimental design in Kahneman and Tversky (1983), we find that the proportion of individuals violating the conjunction rule on Facebook is substantially lower than that reported by previous experiments conducted in the lab, regardless of whether responses are public or private. When responses are posted in a public form, however, the partic- ipation rate is substantially higher. The violation rate on Facebook is also significantly lower than the rate of violation from the same experiment run on Mechanical Turk, a popular online labor market, with monetary incentives. Adding a bonus for the correct answer reduces the violation rate on Mechanical Turk when answers are private, but not when they are public, suggesting that peer effects may indeed counteract the effect of monetary incentives. Our experiment casts doubts about the robustness of behavioral biases for the understanding of real life decisions in environments in which interaction is not anonymous and people are reputation conscious, and suggests the power of social networks to mitigate their effects