Modeling Machine Learning
What do machines learn, and why? To answer these questions we import models of human cognition into machine learning. We propose two ways of modeling machine learners based on this join: feasibility-based and cost-based machine learning. We evaluate and estimate our models using a deep learning convolutional neural network that predicts pneumonia from chest X-rays. We find these predictions are consistent with our model of cost-based machine learning, and we recover the algorithm's implied costs of learning.
We thank Emir Kamenica, Stephen Morris, Nick Netzer, Marek Pycia, Doron Ravid, Jakub Steiner, and audiences at the University of Zurich, the European Summer Symposium in Economic Theory in Gerzensee, and the Sloan-NOMIS Summer School on Cognitive Foundation of Economic Behavior. Caplin thanks the NOMIS and Sloan Foundations for their support for the broader research program on the cognitive foundations of economic behavior. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.