Diverse classrooms are linked to enhanced intellectual engagement and understanding of different perspectives. College admissions decisions have traditionally relied heavily on academic characteristics like GPA and standardized testing. Universities started to adopt holistic strategies while attempting to increase diversity. Yet, increasing subjective assessment may increase risk of human bias. Machine Learning (ML) could assist in admitting a more diverse student body, but algorithmic bias could be introduced. Our goal is to develop software tools to minimize human bias in admissions while actively eliminating algorithmic bias. We will examine past admissions data and identify risks of use of possible privileged values, which have historically put certain groups at a disadvantage. This tool may help reduce bias risk and select a more diverse and academically prepared group for admission.
[1]
Jason D. Wittenbach,et al.
Support Vector Machine for Spatial Variation
,
2013,
Trans. GIS.
[2]
Francisco Herrera,et al.
Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI
,
2020,
Inf. Fusion.
[3]
Eric L. Dey,et al.
How does racial/ethnic diversity promote education?
,
2003
.
[4]
David Danks,et al.
Algorithmic Bias in Autonomous Systems
,
2017,
IJCAI.
[5]
Tom LaGatta,et al.
Conscientious Classification: A Data Scientist's Guide to Discrimination-Aware Classification
,
2017,
Big Data.
[6]
Risto Miikkulainen,et al.
GRADE: Machine Learning Support for Graduate Admissions
,
2013,
AI Mag..