Improving worker safety in the Era of Machine Learning (A)

Citation:

Michael W. Toffel, Dan Levy, Jose Morales-Arilla, and Matthew S. Johnson. 10/2017. “Improving worker safety in the Era of Machine Learning (A).” 618-019. Harvard Business School Case Study. Publisher's Version

Abstract:

Managers make predictions all the time: How fast will my markets grow? How much inventory do I need? How intensively should I monitor my suppliers? Which potential customers will be most responsive to a particular marketing campaign? Which job candidates should I employ? Machine learning, data science, big data, and predictive analytics all use statistical techniques to predict an outcome. This case enables students to begin using data to make predictions and teaches the core metrics to evaluate how accurate predictions are. It helps students understand how to choose among alternative model specifications and introduces the concepts of overfitting and in-sample versus out-of-sample prediction. The case discussion also promotes an understanding of factors beyond prediction accuracy—such as transparency and perceived fairness—that managers need to consider when deciding which predictive algorithm to deploy. The class discussion also helps students appreciate the differences between prediction, correlation, and causation. The case protagonist recently joined a new data science team at the U.S. Occupational Safety and Health Administration (OSHA), a government agency, and needs to evaluate and recommend one of several alternative approaches that OSHA should use to improve how it targets its government inspections of workplaces to better assure safe working conditions. The case includes a dataset and exercise.
Last updated on 11/17/2021