Watch “How I’m Fighting Bias in AlgorithmsLinks to an external site..” Then read “Saving Face: Investigating the Ethical Concerns of Facial Recognition AuditingLinks to an external site..”

Document a use case scenario for each of the vendors CelebSET APIs that would result in minimal bias when in use. More specifically, how would you apply each of the models (Microsoft, Amazon, and Clarify) that would cause the least harm to stakeholders. You may use this template below:

“Microsoft has the highest overall accuracy (99.94%) and lowest difference in accuracy (0.25%) between the best and worst performing intersectional subgroups (dark/light male and light female vs dark female) for the gender prediction task. Therefore, in order to minimize the risk of bias, Microsoft would do well in use case scenarios that focuses on…”

You can download the PDF here