07/06/2026
Why do studies comparing AI vs. human communicators produce such mixed results? Our recent article led by HMC lab member Bumju Jung offers one possible answer:
Participants in our experiment correctly recognized whether a social media post was attributed to an AI or a human source. Yet the label itself did not consistently shape evaluations of credibility, objectivity, bias, or accuracy. Instead, individual differences in endorsement of the machine heuristic accounted for more variation in how participants evaluated the content.
We introduce the concept of the evanishing effect to describe situations where source cues remain visible but have limited influence on evaluation. The idea is intended as a tool for scholars grappling with findings in which AI-versus-human distinctions matter less than other conceptual factors, including pre-existing beliefs, attitudes, or individual differences.
The study also produced an unexpected methodological result: participants did not reliably distinguish between accurate and inaccurate versions of the factual claims. Yet beliefs associated with the machine heuristic still helped explain how the posts were evaluated.
As AI becomes increasingly normalized in everyday communication, understanding when source cues matter—and when they fade into the background—may be just as important as documenting source effects themselves.
📖 Open access: https://stars.library.ucf.edu/hmc/vol12/iss1/7/
