When:
Wednesday, April 5, 2023
4:00 PM - 5:00 PM CT
Where:
Online
Webcast Link
Audience: Faculty/Staff - Student - Post Docs/Docs - Graduate Students
Contact:
Sarah Spurlock
Group: AI@NU
Category: Lectures & Meetings, Academic
Join us for a presentation and discussion on new research led by Daniel Martin, Associate Professor of Managerial Economics & Decision Sciences at Kellogg.
Analyzing Machine Learning using Cognitive Economic Methods
In this line of research, we analyze machine learning predictions using cognitive economic methods. We first propose three counter-factual optimality conditions on algorithmic performance and then evaluate these conditions using an experiment that involves training a convolutional neural network to predict pneumonia from chest X-rays. We then show that these three optimality conditions imply two possible models of machine learning: feasibility-based and cost-based. We find that the pneumonia detection algorithm's behavior aligns with our cost-based model of machine learning, enabling us to estimate the algorithm's associated learning costs using tools of cognitive economics.