AI can guide patient care, but skeptical physicians might put it at a standstill: 4 things to know

Listen

The National Institute of Standards and Technology launched a study to quantify how much physicians trust artificial intelligence. Without trust in AI, physicians may be skeptical of adding automated diagnosis and guidance to patient care plans, according to a June 22 article by Wired.

Four things to know:

  1. In the draft report, researchers analyze the use of AI in medical diagnosis scenarios. The study found that age, cultural background, personality, and other factors played a large role in how much a physician would trust AI. 

  2. For example, a 56-year-old caring woman with lots of experience in AI may rank it highest for accountability, safety and explainability. Physicians trusted AI the least in privacy, security and objectivity. Yet, a 26-year-old adventurous man with little experience in AI may give it the highest ranks in accuracy, resilience and reliability. He would give it poor scores in objectivity, security, explainability, safety and accountability.

  3. Harvard University assistant professor Himabindu Lakkaraju, PhD, studies the role trust in AI plays in decision-making in hospitals. She's working with nearly 200 physicians at hospitals in Massachusetts. For common illnesses like the flu, AI isn't very helpful, since physicians can recognize them pretty quickly and easily, Dr. Lakkaraju said. However, Dr. Lakkaraju found that AI can help physicians diagnose illnesses that are more difficult to identify, such as autoimmune diseases.

  4. Dr. Lakkaraju and her colleagues conducted a study where they gave physicians about 2,000 patient predictions from AI systems and asked them to predict if the patient would have a stroke in the next six months. The information supplied by the algorithms varied. The study found that physicians' predictions were more accurate when they were given more information about the AI system. Accuracy decreased when physicians were given an explanation without data to explain its decision.

© Copyright ASC COMMUNICATIONS 2021. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.

 

Featured Whitepapers

Featured Webinars