Risk of ethnic profiling in recidivism prediction tool OxRec

Ethnic profiling

In a growing number of cases, risk assessment tools have been found to be associated with ethnic profiling. A recent example is System Risk Indication (SyRI), which the Dutch government used to combat fraud in areas such as benefits, tax credits and taxes. A court ruled that the tool could no longer be used. Even more recently, it came to light that the Dutch tax authorities had selected taxpayers for extra checks partly on the basis of whether or not they had dual nationality. With these examples in mind, Gijs van Dijck evaluated the risk assessment tool OxRec, which attempts to predict the risk of recidivism and has been used in Dutch legal practice for some time. He looked at the research on which the algorithm is based and came to some striking conclusions.

Risks and uncertainties

In the first place, Van Dijck considers the quality of the algorithm questionable from a practical point of view. ‘Its application is only possible if you accept wide margins of error,’ he says. Moreover, the use of OxRec creates or reinforces the possibility of unequal treatment based on race, social class or other forms of social inequality. ‘We don’t know whether the prediction model uses ethnic profiling, but equally we don’t know that it doesn’t use ethnic profiling. This means there is a chance that judgments may be made that are partly influenced by a tool that creates or increases a disadvantage in criminal proceedings for people of a particular origin or with a particular ethnic profile.’ Given these risks and uncertainties, Gijs van Dijck says that for now it would be better to end the use of the OxRec tool in Dutch legal practice.