Strategies to improve fairness in artificial intelligence: A systematic literature review

Abstract

Decisions based on artificial intelligence can reproduce biases or prejudices present in biased historical data and poorly formulated systems, presenting serious social consequences for underrepresented groups of individuals. This paper presents a systematic literature review of technical, feasible, and practicable solutions to improve fairness in artificial intelligence classified according to different perspectives: fairness metrics, moment of intervention (pre-processing, processing, or post-processing), research area, datasets, and algorithms used in the research. The main contribution of this paper is to establish common ground regarding the techniques to be used to improve fairness in artificial intelligence, defined as the absence of bias or discrimination in the decisions made by artificial intelligence systems.