Managing bias and unfairness in data for decision support: a survey of machine learning and data engineering approaches to identify and mitigate bias and unfairness within data management and analytics systems

Abstract The increasing use of data-driven decision support systems in industry and governments is accompanied by the discovery a plethora bias unfairness issues outputs these systems. Multiple computer science communities, especially machine learning, have started to tackle this problem, often developing algorithmic solutions mitigate biases obtain fairer outputs. However, one core underlying causes for training data which not fully covered such approaches. Especially, yet central topic engineering management research. We survey research on several domains, distinguishing between publications other domains. This covers creation fairness metrics, identification, mitigation methods, software approaches crowdsourcing activities. identify relevant gaps show activities could be repurposed handle ones might reinforce biases. In second part, we argue novel data-centered approach overcoming limitations current algorithmic-centered methods. focuses eliciting enforcing requirements constraints that are trained, validated, used on. need extend database discuss associated future directions regarding algorithms, formalization, modelling, users,