She builds AI for medicine, and insists the machine should inform your decision, never make it for you. A conversation about why a convincing explanation can be a trap, what her fantasy novels taught her about research, and how to stay the person in control.
Finale used to think the big question was which treatment for which patient. She doesn't anymore. The machine's job, she argues, is to surface what a busy human might miss, a risk buried in a record, a side effect worth a second look, and then leave the actual choice to the person, who knows things the model never will.
Studies show people will follow an AI's advice just because it came with an explanation, whether or not the explanation holds up. Finale's test is simpler and harder: do people actually make better decisions because of it? You cannot judge an explanation until you can measure the quality of the choice it led to.
In her fantasy novel, Finale writes characters who will not accept a forced choice between two bad outcomes. She says research works the same way. When you are handed "you can't have your data" or "your data will stress you out," the job is to find the third option, the one nobody put on the table.
When an institution is dazzled by a demo, Finale's first question is what core need it actually solves, and whether something simpler would solve it. Her example: we build AI to do your taxes, when plenty of countries just hand you the taxes already done. Sometimes the fix is to simplify the system, not add a smarter layer on top.