Types, methods, and evaluations of artificial intelligence (AI) in public communication research in the early phases of adoption: a systematic review

Abstract

Listen

Translate article

Abstract Artificial intelligence (AI) has become pervasive in everyday life, and with the publication of large language models such as ChatGPT especially relevant in the context of media and public communication. This paper synthesizes the conceptualizations, types, methods, and evaluations of AI in public communication in the early phases of innovation adoption, before the broad public discussion around generative AI set in. We conducted a systematic review of empirical research focusing on AI in six social and computing-science databases up until and including 2022 (k = 198). Results show a steep increase in the number of studies published on AI in public communication in just four years. To facilitate a common understanding of what AI is and how it is studied, AI applications are grouped into four AI types: (a) AI as method, (b) Generative AI, (c) AI as communicator, and (d) AI generally. People’s interaction with and attitudes towards AI seem to be central in this research. In addition, AI has mostly been investigated with quantitative methods, often with human participants, as evidenced by the dominance of surveys and experiments. The reviewed research primarily comes from English-speaking countries and often fails to define what AI is or to formulate normative implications. Five blind spots of AI research in public communication and their implications for future empirical studies are discussed.