RisCanvi protocol: addressing its impact from a gender perspective

Abstract

Recidivism risk-prediction systems have historically been grounded in long-established actuarial logics. In recent years, however, these systems have undergone significant transformations with the integration of algorithmic tools, which purport to enhance predictive accuracy and to standardise decision-making processes. A central issue arising from the adoption of such algorithmic tools is the risk of discrimination. The concept of algorithmic discrimination refers to the manner in which algorithms can reproduce or generate novel forms of inequality within the contexts in which they are applied. This article seeks to critically examine the ways in which gender-based discrimination manifests in the implementation of algorithmic risk prediction systems, as well as to explore the extent to which the development of gender-sensitive tools may mitigate this problem. To this end, the study focuses on RisCanvi, a risk-assessment tool employed in prisons in Catalonia, Spain. In doing so, it aims to illuminate the tensions and challenges inherent in incorporating a gender perspective into risk-assessment instruments, and to consider how such a perspective may shape the evolution of protocols that, like RisCanvi, remain in the early stages of development.