Artificial AI Bias

• In Event: Excellence in Education Research: Early Career Scholars and Their Work

Abstract

As Artificial Intelligence (AI) bias increasingly draws attention in every sector of society, concerns of the attribution of bias to AI arise. While many AI biases were due to the design, training, confounding, and algorithm computational capacity, some were not. What if AI accurately predicts something which is the imposition of injustice by nature? Things can be even more complicated if attributions of biases to AI are artificially crafted, which have been seen in the literature broadly. Such artificial AI bias (AAIB) can result in unnecessary AI fear in society and the waste of social capital invested in research. This may be particularly true if understudied "biases" are arbitrarily attributed to AI only because users have a misunderstanding of bias, operate machine algorithms or interpret AI predictions inappropriately, or have an over-expectation of AI predictions. This review presents three types of AAIBs identified in the literature and discusses the potential impacts on society and the solutions.

Author

• Xiaoming Zhai, University of Georgia