Fairness implications for preventive services and child protection decision-making

Large linked datasets are increasingly available to researchers, companies and policy makers to create risk prediction algorithms in child welfare contexts.

Alongside these technical developments are calls for attention to issues of fairness, accountability and transparency, that is, the ethical questions these tools prompt.

Predictive risk tools are used in at least two spheres of child welfare policy and practice: to inform the commissioning of and access to preventive community-based services, and to guide decision-making at the point of notification to child protection services.

The fairness implications relating to these uses have areas of convergence and divergence. This talk explores these implications, and suggests that account must be taken of data sources, the specific use they are being put to, and how that compares with alternative practices, in order to draw conclusions about 'fairness'.