ALGORITHMIC BIAS:
A bias relevant to studies of automated data analyses (e.g. machine learning, deep learning, artificial intelligence); when the application/use of a computer algorithm results in unfair outcomes, for or against, a specific group, category, or individual1,2.
Algorithmic Bias may be caused by many factors including but not limited to; human bias in the selection of model parameters, unfairness or lack of representativeness of the data used to train the computer model, inappropriate training or testing procedures for the model, human programmer biases in assumptions used to create the model beyond the choice of parameters, using the model within the wrong population or context, or using the model to derive a conclusion for which is was not designed to do.
Algorithmic Bias is related to Automation Bias; such that Automation Bias may allow Algorithmic Bias to propagate throughout a medical system. Algorithms are often assumed (falsely) by society to be unbiased, and thus are sometimes given more authority over decisions compared to human-generated opinions (Automation Bias). If users of automated computer systems do not mitigate Automation Bias, then Algorithmic Bias cannot be understood or corrected. Also see: Estimation Bias, Automation Bias, Significance Bias, Statistical Bias, Forecast Bias, Wrong Sample Size Bias, and No Evidence Bias.
References:
1. Panch T, Mattie H, Atun R. Artificial intelligence and algorithmic bias: implications for health systems. J Glob Health. 2019;9(2):010318. (Link to Reference)
2. Kerasidou A. Ethics of artificial intelligence in global health: Explainability, algorithmic bias and trust. Journal of oral biology and craniofacial research. 2021;11(4):612-4. (Link to Reference)