Machine Learning Meets Economics: Using Theory, Data, and Experiments to Design Markets


Economists often build “structural models,” where they specify a specific model of individual behavior and then use data to estimate the parameters of the model. Although such models require strong assumptions, they have the advantage that they can make principled predictions about what would happen if the environment changed in a way that has not been observed in the past. Stanford University’s Susa Athey will describe the application of these techniques to advertiser behavior in sponsored search advertising auctions, focusing on how the models can be used for marketplace design and management. She discusses economists’ focus on causal inference in statistical models as well as the ways in which experiments can be used to estimate and test structural models. Also presented are suggestions about research directions at the intersection of economics and machine learning.


(Source: UWTV)

Leave a Reply