Bias in Machine Learning Algorithms

ECSU Author/Contributor (non-ECSU co-authors, if there are any, appear on document)
Beny Baker, student (Creator)
Christaljah Bethea, student (Creator)
Able Dodo, student (Creator)
Malcolm Dcosta , Professor (Contributor)
Elizabeth City State University (ECSU )
Web Site:

Abstract: Machine Learning is a branch of artificial intelligence focused on building applications that learn from data and improve their accuracy over time without being programmed to do so. It relies on two steps. First is training a model to learn and recognize items or classifications of items based on training examples. And second is testing the model against new examples previously unseen to the model. An example would be training a model to recognize a variety of dogs, and testing it against a picture of a wolf. Whether the model successfully classifies the wolf as an input not in the class of dogs depends on the number examples and scope of the training set. Similarly, several algorithms designed to recognize human emotion using facial expression analysis fail to correctly recognize expressions of darker skin individuals. These biases are due to several reasons. Perhaps the test population may not have been a good representation of the population or the conditions under which the testing was carried out may have been suboptimal, such as poor lighting, which may not affect classification of lighter skin individuals as drastically as darker skin individuals.

Additional Information

Language: English
Date: 2021
Machine Learning, artificial intelligence, facial expression

Email this document to