Net Benefit: A Framework for Assessing the Utility of Predictive Models

• In Event: AERA Virtual Poster Session 3
  In Poster Session: Building a Better Dashboard: Input From Teachers, Students, and Service Providers

Abstract

The field of educational data science—a broad label for the fields of educational data mining, learning analytics, and artificial intelligence in education—regularly produces predictive models to support a diverse array of educational processes. This paper demonstrates how a widely used approach from the field of clinical prediction in healthcare can be applied to educational prediction tasks to reduce the number of harmful models used in district central offices, classrooms, and digital learning environments.

Author

• Andrew E. Krumm, University of Michigan - Ann Arbor