Technology usually progresses at a speed that outpaces existing regulation and legislation, and we are then forced to play catch up when the cat is already out of the bag. This is complicated by a tech sector that is becoming increasing powerful and transnational, making regulation more difficult.
The proliferation of artificial intelligence tools in healthcare broadly, and in dentistry specifically alongside the propensity of many practitioners to be technological minded early adopters, has the potential to be a gamechanger. But as we know, games can be both won and lost.
The Therapeutic Goods Administration recently published a report on regulating medical software devices including artificial intelligence and the Australian Commission on Safety and Quality in Healthcare has created resources on the safe and responsible use of artificial intelligence in healthcare.
We need to understand the capability – and perhaps more importantly the reliability and reproducibility – of the tools that we are using. Ambient AI scribes listen in to a patient consultation to transcribe notes, but they can make mistakes and provide inaccurate or nonsensical outputs. This is an intrinsic characteristic of generative AI tools. They fabricate diagnoses, omit or add steps in treatment plans, and confuse similar words like ‘medication’ and ‘mediation’ or ‘meditation’.
AI still struggles with complexity in a way that humans do not. Clinicians have – or at least should have – an ability to understand the nuance in a conversation, which includes interpreting non-verbal cues. Are digital scribes capable of doing that? How often are we checking the accuracy and veracity of the transcription? And if we don’t, what is the possibility of errors finding their way into the clinical record – and what is the potential impact of that?
In this episode of the Dental As Anything podcast I talk about the issues of regulation of artificial intelligence in healthcare, and the importance for practitioners in understanding not only the potential limitations of AI tools, but also their ethical and legal obligations when using such tools.
