Accurate Approximations About the Truth from Literally False Messages

Communication can be weaponized to manipulate others’ beliefs, most glaringly via explicit lies. We investigate one defense mechanism: people can infer the truth from false messages when they expect that (1) speakers have adversarial motives that direct their lies and (2) bigger lies are costlier. We show in a lab experiment that people can correct for bias in lies when these conditions are satisfied, but with decreased precision. When people adjust what information they glean from expected dishonesty, how might this perturb dyadic, and moreover collective, communication channels? Through probabilistic simulations, we find that deceptive communication systems converge to equilibrium states, in which listeners extract accurate (but less precise) estimates of the truth. Furthermore, when listeners correct for messages assuming that they are distorted, even cooperative speakers (who want listeners to have the correct interpretations) should lie. Liars do not get their way, but they make communication noisier for listeners and other speakers.