Introduction

Session I: Bo Cowgill

Abstract: Why does “algorithmic bias” occur? The two most frequently cited reasons are "biased programmers" and "biased training data." We quantify the effects of these using a field experiment on a diverse group of AI practitioners. In our experiment, machine learning programmers are asked to predict math literacy scores for a representative sample of Americans. One group is given perfectly representative training data, and the other is given a "dataset of convenience" -- a biased training sample containing who confirm to common expectations about who is good at math. We can then measure the bias and fairness properties of the predictions from the engineers under these experimental conditions. Using this experiment, we can quantify the benefits of employing programmers who are diversity-aware, versus obtaining more representative training data. We also measure the effectiveness of training interventions to train AI engineers to reduce bias using recently-developed technical methods.

Break

Session II: Sonny Tambe

Abstract: Using a database of the skills demanded by employers, I show that the human capital required for data science and machine learning remains concentrated by industry and geography, but has a much broader occupational footprint than other technical skills of similar complexity. I then use within-job skills data to show that these patterns are consistent with a model of job design in which employers bundle data science skills with functional knowledge to minimize coordination costs. These findings stand in contrast to data management, network administration, or software development tasks which are primarily bundled into specialized jobs (i.e. the IT workforce). These patterns are also reflected in education requirements. Jobs requiring machine learning often appear in listings seeking candidates educated in domains such as economics or marine biology rather than technical fields such as computer science or engineering. These findings suggest that an understanding of machine learning and prediction techniques may be required by workers in a wider class of occupations than earlier information technologies.

Break

Session III: Xiang Hui

Abstract:Artificial intelligence (AI) is surpassing human performance in a growing number of domains. However, there is limited evidence of its economic effects. Using data from a digital platform, we study a key application of AI: machine translation. We find that the introduction of an upgraded machine translation system has significantly increased international trade on this platform, increasing exports by 10.9%. Furthermore, heterogeneous treatment effects are consistent with a substantial reduction in translation costs. Our results provide causal evidence that language barriers significantly hinder trade and that AI has already begun to improve economic efficiency in at least one domain.

Reception

Dinner

Session IV: Rob Seamans

Break

Session V: Zachary Lipton

Break

Session VI: Florenta Teodoris

Lunch and adjourn