HOME    |    NEWS    |    PEOPLE    |    PROJECTS    |    PAPERS    |    RESOURCES    |    CONTACT-US    

Projects

Human Ethical Bias Detection with Machine Learning Software.

Aug 2021 | repo

Instead of trying to reduce bias in machine learning software, this project aims to reduce bias from its source--- the human decisions.

• That is, this work proposes to utilize the machine learning bias inherited from the training data (human decisions) as an indicator for detecting human bias.
• While it is difficult to directly test whether a human has bias, with current research on machine learning bias, it is now easy to test, in large scale at low cost, whether a machine learning software has bias.