How does the radiology community discuss the benefits and limitations of artificial intelligence for their work? A systematic discourse analysis

Abstract

Purpose: We aimed to systematically analyse how the radiology community discusses the concept of artificial intelligence (AI), perceives its benefits, and reflects on its limitations.

Methods: We conducted a qualitative, systematic discourse analysis on 200 social-media posts collected over a period of five months (April-August 2020).

Results: The discourse on AI is active, albeit often referring to AI as an umbrella term and lacking precision on the context (e.g. research, clinical) and the temporal focus (e.g. current AI, future AI). The discourse is also somewhat split between optimism and pessimism. The latter considers a wider range of social, ethical and legal factors than the former, which tends to focus on concrete technologies and their functionalities.

Conclusions: Further precision in the discourse could lead to more constructive conversations around AI. The split between optimism and pessimism calls for a constant exchange and synthesis between the two perspectives. Practical conversations (e.g. business models) remain rare, but may be crucial for an effective implementation of AI in clinical practice.

Keywords: Artificial intelligence; Deep learning; Discourse; Machine learning; Perceived benefits and limitations; Social media.