Metaphors in digital radiology: ethical implications for responsibility assignments of human-AI imaginaries

Abstract

Abstract The advent of artificial intelligence (AI) in radiology triggered identity-threatening fears for radiologists of becoming replaced by machines. Beyond this competitive narrative of humans versus AI, a collaborative narrative for human–AI-interaction emerged with a new metaphorical landscape both for the functions of AI and the roles of radiologists. This article aims to raise awareness of the ethical implications of figurative language in human–AI interaction in digital radiology. The paper is divided into two parts. The first part justifies the approach of metaphor analysis in medicine, draws a spectrum of ethical implications for language choices, and introduces taxonomies of human–AI interaction. We use these preliminaries as a hermeneutical tool to conduct such a metaphor analysis in the second part. There, we identify prevalent metaphors in the radiological community and discuss their ethical implications regarding responsibility assignments. We argue that while metaphors can facilitate a collaborative narrative, they may also lead to the undesirable ethical consequence of attributing moral responsibility to AI, which lacks the necessary features for such responsibility. The spectrum of metaphorically constructed functions of AI ranges from “time-saving tool” to “assistant” and “ally”. For the roles of radiologists, we found metaphors and analogies which are derived from contexts of aviation (radiologists as “pilots” and AI as “auto-pilots”), war (radiologists at the “forefront of technological development”), music (radiologists as “conductors” of multi-disciplinary teams), and hierarchical power contexts (radiologists as “technology and thought leaders”). Despite radiologists’ expressed willingness to collaborate actively with AI, the prevailing analogy of AI as a “tool” primarily suggests mere delegation of routine tasks, at the same time allowing radiologists to maintain their professional competencies. However, a new competitive narrative of AI-savvy versus non-AI-savvy radiologists also emerged, transforming the initial competitive narrative from human versus AI to human versus human competition.

Ähnliche Arbeiten