Events
Modeling Machine Learning
Speaker: Andrew Caplin, Daniel Martin, and Philip Marx
Location: 60 Fifth Avenue, Room 7th floor
Date: Wednesday, February 15, 2023
Economic models are about counterfactuals: rationalizing what did happen by comparing the outcome with clearly feasible alternatives. We introduce three directly computable counterfactual optimality conditions on algorithmic performance. We implement them in a particular case and show 2 of the 3 to apply. We explain the powerful and surprising links between these conditions and cognitive economic models. In the example we show how to process algorithmic performance data to recover implicit costs of learning.