When you ask people on Mechanical Turk how well they understand a political issue on a scale of 1-7, the mean of their ratings is 3.82. But if you ask them to explain the issue first, then the mean of their ratings is 3.45.
If you ask them to rate whether they are in favor of some policy on a scale of 1-7, their average ratings are 1.41 away from 4. After asking people to explain the policy, their average ratings are 1.28 away from 4.
These are tiny effects. The effects in the second experiment appear to be just as small, judging by the figures.
The third experiment finds that, for people who rate how much they are in favor of a policy as 6 or 2, if asked to list reasons for their opinion, they are about 70% likely to donate 20 cents to that cause, and if they are asked to explain their opinion, they are about 30% likely to donate to that cause. At last, a non-minuscule effect! But it's about people's willingness to donate 20 cents, a negligible amount of money.
The overall finding seems to be "people slightly underestimate how well they understand things, and if you ask them for an explanation, they get confused and become slightly less confident."
How strong are the generalizations we can draw from such small effects? Are changes of this magnitude really informative about what determines people's behavior? Personally I suspect not.
No comments yet.
