Why Corrections (Partly) Fail: Higher-Order Inferences and the Stickiness of Misinformation

The notion that misinformation has persistent effects on beliefs and attitudes despite being debunked is widely recognized, but poorly understood. Through two preregistered experiments, we propose a novel psychological explanation of this phenomenon, based on the conditional nature of spontaneous higher-order inferences (e.g., categorization). In Study 1 (N = 263; American adults in an online (Prolific) panel), inferred beliefs about the subject of the misinformation (attributed behavioral tendencies and personality traits) were assessed after participants read a news article about fictional fraud, followed by a correction for half of the participants. We predicted that corrections would reduce belief in misinformation more than the inferred beliefs about the subject. In Study 2 (N = 429; Dutch adult convenience sample), inferred evaluative beliefs about the subject were additionally assessed, and the design included multiple correction formats. Results confirm our explanatory model: belief in misinformation is more easily reversed than beliefs inferred from misinformation (Study 1 and 2), across correction formats (Study 2). The present work offers initial evidence that the persistence of misinformation may stem from less effective correction at higher levels of inference: Beliefs inferred from false evidence are often not logically addressed by the correction of that evidence, leading to continued persuasive misinformation effects.