Dataverzoek
- Titel:
- Bias in recidivism algorithms
- Aangevraagde data:
- Dataset van verdachten van misdrijven; per leeftijd, geslacht en recidive
- Aangevraagde periode:
- 28-11-2000 tot 28-11-2018
- Gebruik:
- Machine Learning has the potential of improving decision making by eliminating pontential human biases. However, due to the already existing biases in data, the resulting algorithm might inherit the biases of human and the machine learning model might behave in an unfair manner, for example discriminating towards a particular protected group. In this research we propose a machine learning method to predict recidivism and we discuss the potential problems with it and ways to mitigate them .