Abstract: Humans are notoriously devoted to predicting, explaining, and controlling their environment. Our understanding of these related, but computationally distinct, tasks is far from perfect, and has been a subject for psychologists, philosophers, and computer scientists. Looking at the three tasks together reveals complex relationships. For example, correct predictions don't always come with correct explanations, and explaining why something went wrong doesn't mean you know how to fix it. I’ll present new work on how humans and machines learn to predict, explain, and control in a variety of scenarios, on the ways in which human performance might deviate from baselines provided by machine-learning algorithms, and on how to model human thinking under cognitive constraints.
Noyce Conference Room
US Mountain Time
Our campus is closed to the public for this event.
Roman TikhonovPostdoctoral Fellow, Dept. of Social & Decision Sciences