Key points are not available for this paper at this time.
The effects of automated authorship on news perceptions have been inconsistent in past research in both directionality and statistical significance. Could null or opposite direction effects found across prior studies be attributed to differences in manipulation recall or variability in news story polarization? What about inconsistencies in the measurement of key outcome variables like message credibility? An online experiment (N = 820) tested these possibilities using a 2 (declared source: human vs. AI) × 2 (source priming: present vs absent) × 2 (news topic: data-focused vs polarizing) between-subjects design. The effect of declared source was not statistically significant across most tested variables, even varying across different measurement approaches for the same outcome. The two exceptions out of eleven tested outcomes showed a preference for human authorship, such that news attributed to human authors was perceived as more authentic and trustworthy than news attributed to AI authors. Theoretical and practical implications of these findings are discussed.
Thomas K. Waddell (Wed,) studied this question.