Responsible scaling of artificial intelligence in healthcare: standardization meets customization

Abstract

Organizations across the globe are progressively investing in artificial intelligence (AI) innovations to meet today’s healthcare challenges. Meanwhile, public policy increasingly emphasizes the need for these innovations to be ‘scaled’. As scholars emphasize, scaling innovations is never just ‘more of the same’, but requires adapting innovations to local contexts. In this perspective paper, we aim to explore and draw attention to the tensions and possible alignments between standardization and customization that should lead to a responsible scaling of AI in healthcare. We approach responsible scaling building on the notion of socio-technical configurations. Configurations are unique assemblies of technological and non-technological components, including human factors, integrated in different ways to meet particular local requirements. We explore how conceptualizing AI tools and the broader socio-technical ecosystems in which they are integrated as configurations can offer a framework for envisioning possible pathways for responsibly scaling AI. We contend that standardization and customization can be employed synergistically within AI configurations. Standardization can be an important driver of innovation at the level of configurational components of healthcare AI, as well as the interoperability between these components. Thereby, standardization can expand the configurational options that local AI implementations can draw from and lay a foundation for local customization of healthcare AI ecosystems at the architectural level. Accordingly, we propose key considerations for innovators and policymakers to boost the configurability of healthcare AI, and discuss the need for, and challenges of shaping of healthcare AI configurations at the local scale.

Ähnliche Arbeiten