Biased Programmers? Or Biased Data? A Field Experiment in Operationalizing AI Ethics

Author: Bo Cowgill, Fabrizio Dell’Acqua, Samuel Deng, Daniel Hsu, Nakul Verma, Augustin Chaintreau

Publisher: Navigating the Broader Impacts of AI Research Workshop at the 34th Conference on Neural Information Processing Systems

Publication Year: 2020

Summary: The following study was done to analyze if it was the programmers or the algorithms that are truly creating biases. Whether or not it is the programmers or the algorithms may play into our discussion of bias of algorithms inherently learning as a child would or if it is the direct influence of the programmer. What the study found was the root cause they found for most algorithmic bias was from a biased training dataset. Additionally, there was no less bias found between programmers of minorities or women when compared to men; however, prediction errors were found to be correlated within demographic groups which could be an interesting finding to take a deeper look at.


Posted

in

, , ,

by