Bias, ethics and social responsibility in predictive models

Abstract

Predictive models based on Machine Learning have been widely used to support diverse decision-making processes. Their decisions, however, can have an impact on human rights of different societal groups. This is because the models are trained based on data from the society itself, often with built-in or learned biases. In this sense, it is necessary to go beyond the traditional objective of obtaining high-performance predictive models by incorporating ethical and responsible principles in their conception, training, and implementation in order to guarantee the greater social good. This paper discusses aspects associated with biases, ethics, and responsibility in the construction of predictive models.