An experimental investigation of persuasion through selective disclosure of evidence

We experimentally study the interaction between a persuader and a decision-maker. The former would like to persuade the latter to approve a project by providing evidence on the project's value. He may choose a selective disclosure strategy on the basis of his private information. Our experimental design contrasts situations where the persuader observes private information or not and where the decision-maker interacts with a human or robot persuader. The experimental results confirm the theoretical prediction that the human persuader manipulates the production of evidence. Although the decision-maker does not adequately take into account such manipulation, the comparative static analysis across treatments is mostly consistent with theoretical predictions with a rational decision-maker. Our findings on the welfare effect of the persuader's manipulation on the decision-maker are consistent with theory. In particular, the decision-maker may benefit from such manipulation. However, the welfare effect on the persuader is not always consistent with theory, in that there are instances in which the persuader is not hurt by manipulation even though theory predicts that he is.