Digital Technologies: The Unseen Student

Context

Your company provides software that helps universities predict student success. However, a major study shows that your models consistently underperform for racially minoritized students and that bias mitigation efforts are mostly ineffective. The software risks reinforcing systemic inequality in education.

Dilemma

A) Delay the release, invest heavily in fixing the bias, risking losing valuable market share, but ensuring fairness for all students.

B) Continue offering your current models, emphasizing overall performance and offering technical disclaimers on bias.

Summary

A study using nationally representative data from the Education Longitudinal Study of 2002 and various machine learning modeling approaches revealed significant racial bias in college student success predictions. Algorithms, increasingly used by universities, show less accuracy for racially minoritized students. Attempts to mitigate this bias proved largely ineffective, highlighting how these tools perpetuate existing societal inequalities rather than providing objective assessments.

Resources:

• https://journals.sagepub.com/doi/10.1177/23328584241258741
• https://themarkup.org/machine-learning/2021/03/02/major-universities-are-using-race-as-a-high-impact-predictor-of-student-success