Knowledge Representation for Debiasing
ISTCS.org, Paolad Di Maio · MTS Press eBooks · 2021
Bias in Artificial Intelligence and Machine Learning is a top concern for developers and users of intelligent technology. Intelligent systems can, in theory, be designed to identify and resolve algorithmic bias, but for this to be possible it is necessary to have an adequate knowledge representation of bias itself, which has never been fully achieved. Here a KR of bias is presented as a matrix and a relational set constituting the logical and functional foundation of an intelligent system capable of autonomously identifying, measuring and minimise algorithmic bias.