Ethical considerations using Predictive Risk Modelling in child and family welfare

ntroduction
Digitalization - in this case, Predictive Risk Modelling based on Machine-Learning - is coming forward in both social work with children and families at risk and the potential underlying research processes. Predictive Risk Modelling poses both potential benefits and risk when applied in social work. Some of the potential benefits are prevention of maltreatment and targeted interventions to the children and families in the direst need. In addition, Predictive Risk Modelling may help reduce decision-time and errors, increase consistency in the way social workers respond to similar risk situations and thereby reduce social worker bias and between-group inequalities. The many potential risks for service users are also well known; such as disproportional impact on people from marginalized groups in relation to e.g. gender, race and socio economic status.

Methods & results
Based on a framework developed especially towards this field, we present and discuss ethical consideration, reflections and actions, in relations to five main themes: Beneficence, non-maleficence, autonomy, justice and explicability.
The purpose is to show and discuss ethical issues and implications concerning the introduction and testing of a Predictive Risk Model based on Machine-Learning in Danish child and family welfare. The Predictive Risk Model is designed to assist child and family welfare services in detecting children at risk and it is specifically aimed at the decision-making process regarding notifications.
In the presentation, we show how we apply relevant ethical considerations to our existing case.

Conclusions
It is our hope that our ethical considerations and challenges regarding the use of Predictive Risk Modelling in child and family welfare can inspire both research and the field of practice when taking a deep dive into this very difficult field within digitalization in social work.