Disinformation detection is becoming increasingly important and relevant because it is easier than ever to create and disseminate disinformation. How does detection ability affect the incentive to generate disinformation? Given the practical constraints of classification technology, how should a detector be designed? To answer these questions, this paper studies the problem where a sender strategically communicates his type to a receiver, and a lie detector generates a noisy signal on the truthfulness of the sender’s message. The receiver then infers the sender’s type both through the message from the sender and through the signal from the detector. We find a nonmonotonic relationship between the probability that the low-type sender is lying and the accuracy of detection. More accurate detection increases the probability of lying when the true-positive rate is low, because of a persuasive effect. By contrast, more accurate detection decreases the probability of lying when the true-positive rate is high, because of a dissuasive effect. We also characterize the optimal detector design. The designer always chooses the lowest feasible false-positive rate for any true-positive rate. The possibility of false-positive alarms implies that the designer chooses an intermediate true-positive rate rather than the highest true-positive rate. This paper was accepted by Dmitri Kuksov, marketing. Supplemental Material: The online appendix is available at https://doi.org/10.1287/mnsc.2024.06956 .
Yang et al. (Thu,) studied this question.