An Algorithmic Approach to Reducing Gender Discrimination: Evidence from a Large-Scale Experiment on Tracking Decisions

This project will study how to use algorithms to equalize students’ educational opportunities in a tracked high school system. The researchers will use a randomized evaluation in around 400 Italian middle schools to test two conceptually different ways to leverage algorithms to close gender gaps in the most rigorous scientific high school tracks. To help teachers predict future performance, the first treatment provides teachers with algorithmic recommendations that indicate whether students are likely to excel in the most rigorous scientific high school tracks. To address behavioral biases and awareness issues, the second treatment provides teachers with real-time feedback on the diversity of their track recommendations compared to an algorithmic benchmark. The researchers hypothesize that the first intervention will lead to fairer track recommendations if the algorithmic recommendations supplant teachers’ inaccurate mental models of how individual students will perform in high school. Meanwhile, the second intervention will lead to fairer track recommendations if the algorithmic benchmark prompts teachers to correct behavioral biases by comparing the diversity of their track recommendations to a more accurate measure.