Newsrooms have adopted AI already nearly a decade ago, but this study by Jun Luo from The Chinese University of Hong Kong looked at how audiences respond to it. At first, AI was only capable of handling routine tasks, but is increasingly qualitatively competent too: producing narrative texts and generating news content across a wide range of topics. Here, the essential question on whether the audiences trust AI news amidst wider concerns of media credibility is examined. 

Typically, previous studies suggest that positive machine heuristics may be at play here: AI is perceived as being less biased than humans. On the other hand, ethically sensitive, human-centric topics are still seen as better handled by humans. Newer studies also suggest that AI might reduce perceived bias (Cloudy, Banks, and Bowman 2023). However, studies remain mixed.

Building on previous research, this study examines how perceptions of news credibility vary by authorship and topic, and whether these differences are moderated by individual differences in AI familiarity and topic knowledge. Lastly, whether perceived credibility affects future news selection. 

The study used a 2 (source: AI vs. human journalist) × 2 (topic: economy vs. immigration) between-subjects experimental design. Thus, there were four groups: AI economy, AI immigration, and human economy & human immigration. The articles were found with the terms “undocumented immigrants” and “economic growth”. Explicit partisan signals were removed.

A total of 1012 US adults completed the study. First, they were asked to identify the author: human, AI, or “not sure”. There were five items:believable, accurate, trustworthy, unbiased, complete, and they were measured on a 7 point scale.

People were more accurate in identifying AI-written articles than those of human beings. The news topic did not moderate credibility, but AI familiarity was associated with greater perceived credibility overall. Personal involvement had a major effect: more involved participants saw the stories as more credible. Credibility influenced AI selection only with involvement.

In summary, participants defied the hypotheses of the study: that credibility ratings vary depending on the interactions between news author, news topic, and individual level moderators such as AI familiarity and personal involvement. The findings are interesting and have relevance in research on news credibility and AI.

The article “Beyond Credibility: How News Topic and Cognitive Processing Shape Responses to AI-Authored Journalism” by Jun Luo is in Journalism Studies. (Free abstract). 

Picture: Untitled by Igor Omilaev.

License Unsplash.