This study examines how people evaluate factual claims attributed to human journalists versus artificial intelligence (AI) news bots, and how these evaluations are shaped by individual endorsement of the machine heuristic. An online experiment (N = 233) used a 2 (Agent: human vs. AI) × 2 (Information Veracity: accurate vs. inaccurate) design in which participants viewed a short tweet embedded in either a human or AI profile and then rated perceptions of source credibility, objectivity, bias, and accuracy. Endorsement of an AI-focused machine heuristic was significantly associated with higher source credibility, perceived objectivity, and perceived accuracy, and lower perceived bias, whereas agent type showed no significant effects in the full sample. In post hoc analyses, the difference by agent emerged only among participants high in the machine heuristic. These findings suggest boundary conditions of the machine heuristic: while it shapes perceptions, its operation appears increasingly agnostic to human versus AI source cues. We call this the evanishing effect, wherein source cues (human vs. AI) are recognized but do not systematically alter evaluations, so neither human nor machine agents are consistently preferred.
Jung et al. (Sat,) studied this question.