Knowledge of automated journalism moderates evaluations of algorithmically generated news

Drawing on propositions from the HAII-TIME (Human–artificial intelligence [AI] Interaction and the Theory of Interactive Media Effects) and Persuasion Knowledge Model, this study examines how knowledge of automated journalism (AJ) moderates the evaluation of algorithmically generated news. Experiment 1 demonstrates the utility of process-related knowledge in user evaluations of agency: individuals with little knowledge of AJ prefer attributions of human authorship over news stories attributed to algorithms, whereas individuals with high AJ knowledge have an equal or stronger preference for news that is described as algorithmically generated. Experiment 2 conditions these effects to show how prior characterizations of AJ—whether more machine- or human-like—shape evaluations of algorithmically generated news contingent on user age and knowledge level. Effects are found for differing age groups at lower levels of AJ knowledge, where machine-like characterizations enhance evaluations of algorithmically generated news for younger users but ascribing human-like traits enhances evaluations of automated news for older users.