Skip to content

Latest commit

 

History

History
37 lines (32 loc) · 2.69 KB

Impacts on Trust of Healthcare AI.md

File metadata and controls

37 lines (32 loc) · 2.69 KB

Roles for Healthcare AI

  • AI robotics proposed for diagnosis, simple surgeries, patient monitoring and more
  • Focus on the impact of these technologies in patient-doctor trust
  • A decision to use a robot to care for someone may affect the bonds created by human caregiving.

The nature of trust

  • Trustor makes himself vulnerable based on expectations about the trustee likely actions
  • "Behavioral trust" - Grounded on reliability - the trustee is predictable by the trustor
  • "Understanding trust" - trustor believes firmly trustee will act in certain ways in a particular case, for example in novel situations.
    • This kind of trust can be established with a machine too, but it's easier with humans.
  • Role-based trust is not a third kind of trust, but as a vehicle to establish one of the other two
  • Doctors are a canonical example of role-based trust. We expect doctors to have certain values, beliefs and good intentions.
  • The doctor's role has shifted from a paternalism standpoint to a collaborative one.
  • Problems when patient's care needs differ from near-term desires: e.g - not enough lithum in bipolar disorder

Healthcare AI and patient-doctor trust

  • Doctors are licensed and grantors of licenses ensure objective criteria is met.
  • If AI is introduced to one task, it threatens to displace some patient-doctor trust. Impact will depend on whether regulatory mechanisms and approval are put in place.
  • It's different to have an AI that performs procedures than an AI that judges whether procedures are appropriate.
  • AI for procedures could be regulated in a similar way we regulate drugs.
  • We would expect the social role of doctors to go from 'know-it-all' to 'mere-user of an AI'. It's important to make the public know that the doctor is an power-user of an AI system and has a broader scope.
  • Patient experiences may also undermine trust, e.g: a doctor which does not care about patient's wishes, or faulty AI
  • People are often more willing to provide info to an AI than a human.
  • One good example for AI use could be patient monitoring - has a patient taken meds when they're supposed to, do they comply?

Regulatory Policy Recommendations

  • Doctors using AI systems and their results must have educational training that's overseen, measured and approved by an independent outside group
  • Regulators may compel certain types of education as a precondition to use AI or robotics technologies
  • AI should not be used for patient care without the educated consent of the patient.
  • Use of a technology does not take priority over a patient's wellbeing - e.g: the patient could opt out