Check out this StatNews piece addressing the question: should patients consent for use of artificial intelligence in the clinic setting? It builds the case for an emerging crisis in healthcare where patients are the victims of a failure to disclose the use of AI in the clinical setting.
The concerns expressed reflect the a false dichotomy of man or machine. We like to see something as done by the doctor or done by the machine — with a clear boundary separating where the computer stops and we begin.
But given our relationship with technology things aren’t shaping up this way.
Technology as partner or independent player
Increasingly stuff we once did with our hands (diagnose appendicitis) are now done through smart technology (advanced imaging). And instead of turning all aspects of appendicitis care over to intelligent machines, we’re working with them in a kind of partnership. Even as doctors we don’t always realize when it’s happening.
This is best reflected in the difference between artificial intelligence (AI) and intelligence augmentation (IA). In the simplest sense AI looks to effectively replace us while IA looks to complement us. Think of IA as machines helping us to be better versions of ourselves. Clinically, at least. A good example is IBM Watson as that friend on our shoulder giving us a nudge one way or another.
The practical nature of consenting for smart machines
Consenting for the use of machines in medicine could get hairy given their scaling role in nearly every aspect of clinical care. If we believe that the preliminary interpretation of MRIs, the screening of biopsies by AI or the use of deeply augmented clinical decision support requires some kind of ethically approved sign-off by patients, it’s gonna be a long 21st century for all of us.
Of course turning sentient pieces of technology onto the wards where patients may have a hard time identifying them as machines will require some serious conversation. But autonomous appliances that can independently master the premed curriculum, apply to medical school and then magically roll into the clinic are a ways off.
Until then, seeing every machine algorithm in clinical care as needing patient consent will be practically impossible.
If you like Should Patients Consent for Use of Artificial Intelligence? then you’ll probably like our 33 charts Artificial Intelligence Archives. It’s everything on the site written about AI.