Keywords: Algorithms, Discrimination, Efficiency, Facial Recognition Systems, Manipulated Data, Training Data, Unbalanced Data
Author: Jabril Ashe
Publisher: CrashCourse
Publication Year: 2019
Summary: The following video discusses how bias is a natural outcome as systems such as the human brain or a computer algorithm try to take shortcuts in order to increase efficiency. It becomes a problem when we do not acknowledge it, we fail to recognize exceptions to the pattern, or when it results in incorrect or skewed results (including discrimination). Some sources of bias are: unbalanced data (such as a facial recognition software being trained with a dataset devoid of pictures of people of color), having to work around data that are difficult to quantify, algorithms creating or reenforcing a feedback loop, and introduction of manipulated data by malicious actors.