Dermatological diagnostic-assistive technologies: a call for regulatory action

Abstract

Artificial intelligence (AI) has been increasingly implemented within the healthcare sector. In that regard, efforts are being made toward commercialising diagnostic-assistive technologies. The ability of such systems to mine a significant amount of clinical data has enabled diagnostics to become progressively more precise. Nevertheless, the usage of this technology poses critical challenges regarding ethical, legal and sociological aspects that ought to be addressed and call for regulation. Diagnostic-assistive technologies may be implemented in multiple healthcare fields, all of which require the policymakers’ attention. Despite having transversal concerns, studying the impact of this technology for dermatological diagnosis purposes is paramount, considering that it poses a specific concern in this particular field of application: discrimination. Algorithmic biases such as misrepresentation due to insufficient data for data training processes render the accuracy of the diagnosis provided by this technology selective which favours one group of people over another, solely based on skin colour (a physiognomic feature highly relevant to this field of medicine). Moreover, this technology poses relevant questions regarding the inefficiency of the current framework on liability, such as the Product Liability Directive, which establishes inadequate regimes to encompass these instruments. Additionally, by being so onerous to the healthcare structure, the current legislation enables physicians to adopt defensive behaviours, which tend to favour their patients less. This article aims at understanding how regulatory actions can prevent the lack of legal certainty and social concerns regarding using dermatological diagnostic-assistive technologies.