TY - JOUR AU - Ruback, LĂ­via AU - Carvalho, Denise AU - Avila, Sandra PY - 2022/12/30 Y2 - 2024/03/28 TI - Mitigating Bias in Machine Learning: A Socio-technical Analysis JF - iSys - Brazilian Journal of Information Systems JA - iSys VL - 15 IS - 1 SE - Regular articles DO - 10.5753/isys.2022.2396 UR - https://sol.sbc.org.br/journals/index.php/isys/article/view/2396 SP - 23:1-23:31 AB - <p>This work presents a socio-technical analysis of biases included in the machine learning process. We describe in this work four types of biases: historical bias, data bias, model bias, and human interpretation bias, and how they can occur during the learning process, together with their social and cultural implications. We also bring strategies to mitigate those biases, including computation solutions, such as balancing the data used for the training and alternative metrics for the model evaluation, non-computational solutions, regulatory efforts, and initiatives to promote diversity in the tech industry and academy.</p> ER -