A CONNECTED INQUIRY
Three directions, one recurring problem: what a decision needs to know, and what happens when part of that gets left out.
What do we expect a machine to understand?
How people picture an AI system affects what they hand over to it, how much context they bother to give, and who they think is responsible when it goes wrong. I study those assumptions, and what collaboration should look like when they turn out to be wrong.
What is lost when experience becomes information?
Clinical reasoning keeps moving between what a patient goes through and the codes, notes, and labels used to describe it. I look at where records, diagnostic categories, and computational tools keep the context a decision needs, and where they quietly drop it.
How can computation help us read genetic variation?
I use computational methods to study genetic variation and the molecular mechanisms behind it, and I try to be clear about what those methods assume and where they stop being reliable. The direction grows out of my own experience as a person with Usher syndrome.
International Journal of Human–Computer Interaction
If we treat AI like a new employee, what do we expect it to notice, and what do we never think to tell it?
Explore the paper