Focal points and blind spots of human-centered AI: AI risks in written online media

Abstract There is a strong tendency in prevailing discussions about artificial intelligence (AI) to focus predominantly on human-centered concerns, thereby neglecting the broader impacts of this technology. This paper presents a categorization of AI risks highlighted in public discourse, as reflected in written online media accounts, to provide a background for its primary focus: exploring the dimensions of AI threats that receive insufficient attention. Particular emphasis is dedicated to the ignored issues of animal welfare and the psychological impacts on humans, the latter of which surprisingly remains inadequately addressed despite the prevalent anthropocentric perspective of the public conversation. Moreover, this work also considers other underexplored dangers of AI development for the environment and, hypothetically, for sentient AI. The methodology of this study is grounded in a manual selection and meticulous, thematic, and discourse analytical manual examination of online articles published in the aftermath of the AI surge following ChatGPT’s launch in late 2022. This qualitative approach is specifically designed to overcome the limitations of automated, surface-level evaluations typically used in media reviews, aiming to provide insights and nuances often missed by the mechanistic and algorithm-driven methods prevalent in contemporary research. Through this detail-oriented investigation, a categorization of the dominant themes in the discourse on AI hazards was developed to identify its overlooked aspects. Stemming from this evaluation, the paper argues for expanding risk assessment frameworks in public thinking to a morally more inclusive approach. It calls for a more comprehensive acknowledgment of the potential harm of AI technology’s progress to non-human animals, the environment, and, more theoretically, artificial agents possibly attaining sentience. Furthermore, it calls for a more balanced allocation of focus among prospective menaces for humans, prioritizing psychological consequences, thereby offering a more sophisticated and capable strategy for tackling the diverse spectrum of perils presented by AI.