When AI and Humans Produce Partial Truths: Examining Acceptability of Perceived Error and Perceived Associated Harms

Existing research on misinformation often focuses on messages that are completely false. For greater external validity, our experiment examines reactions to messages that contain both false and accurate information. Using a framework of uncertainty attributes of truth claims, we examine perceptions of how acceptable error is when the source is (perceived to be) human-only, generative artificial intelligence-only, or a combination of the two. We examine the acceptability of perceived error and harms associated with the message and its topic, and how they interact as predictors of intentions to engage with the message or intervene.