I made a bold claim recently that you may want to challenge: AI is not going to transform healthcare in the near future. My research for the last two decades has focused on one area of AI, natural language processing, so I’m a huge fan and see the potential. My brief justification for the claim is that we need layers of infrastructure to support the translation of AI into healthcare, including analysis-ready data and more comprehensive validation that supports integration of AI techniques into clinical workflows. We see the Validitron as an important enabler for implementation of AI and other digital interventions. I found a few practical guides to making the promise of AI a reality.
Evaluation framework
We hosted a presentation last week by Gretchen Purcell-Jackson about how to evaluate AI applications in health care. I liked the focus on the lifetime of the AI (and other digital) intervention from conception and technical development to clinical impact and the steps in between that often get ignored. You can read more about the framework developed at IBM with Gretchen and colleagues here.
Governance framework
University of Wisconsin-Madison published a recent paper outlining the governance framework they are using to facilitate safe and equitable deployment in a large healthcare organization. You will see that the application of AI in healthcare requires oversight and co-design by many groups of people.
AI Research
The director of Melbourne Connect, Ed Hovy, brought together University of Melbourne researchers interested in AI to think about how we might join forces for good. He asserted that we will never do anything big or impactful in AI at the University when almost every AI researcher is bound by a model that requires a lot of teaching and only leaves room for short-term single-PI-led research projects. Ed spent his entire academic career as a research academic--he was able to spend close to 100% of his time on research. The downside is the need to fund your entire salary from grants. Small price to pay for bigger impact?
Regional Linguistic Quirks (RLQ): Brian discovered Erik Singer, a dialect coach, who has some really fun videos talking about accents and critiquing technique of actors. He doesn't have an Australian accent edition, but see if you can recognize the differences in American accents in the two WIRED videos on his home page and compare my accent (American West) with Kayley's (Midwest) for example!
Comments