Case Study: The William and Flora Hewlett Foundation achieves affordable, scalable, and more consistent essay grading with Kaggle

The William and Flora Hewlett Foundation: Targeting Education Through Data Science

Kaggle partnered with the William and Flora Hewlett Foundation to run the Automated Student Assessment Prize (ASAP), a data‑science competition to develop automated grading tools that make essay scoring consistent, fast, and affordable without sacrificing quality. The two‑phase challenge asked teams to reproduce expert scores for more than 22,000 long‑form essays (150–550 words) in Phase 1 and to score over 27,000 short answers (~50 words) in Phase 2.

In Phase 1 the top five Kaggle teams outperformed eight commercial vendors and showed greater consistency than expert human graders—the winning team later sold its intellectual property. In Phase 2 teams improved on an automated benchmark by nearly 20% (though did not surpass human graders); winners presented to sponsors and publicly released code and writeups, demonstrating strong promise for scalable, data‑driven assessment in education.

The William and Flora Hewlett Foundation