"Good" and "Bad" Machine Agency in the Context of Human-AI Communication: The Case of ChatGPT

Machine agency, defined as the ability of machines to act autonomously and interact with users, is becoming increasingly significant in the field of human-machine interaction research. This is especially evident in relation to ChatGPT and other generative artificial intelligence (AI) tools. In this paper, we present an initial extension of S. Shyam Sundar's theory of machine agency, specifically in the context of human-AI communication. We review existing literature on this topic and use real-life news reports on ChatGPT from November 2022 to April 2023 as a basis for illustrating the factors that influence people's perceptions of machine agency as either positive or negative. These perceptions are influenced by a range of factors, including ethical alignment, privacy, transparency, social inclusiveness, human autonomy, and well-being. We propose a more explicit differentiation between "good" and "bad" machine agency, a conceptualization that can enhance our understanding of the complexities of human-AI communication. We believe this approach can contribute to the development of guidelines and best practices for using generative AI tools and similar AI technologies. Finally, this conceptualizationmay help people and the public to better benefit from generative AI and identify its risks.