Algorithmic tools to help social workers in child protection service make well-informed decisions: A scoping review

Introduction

Decisions on whether to investigate a report of child maltreatment or to remove a child from home are often problematic and surrounded by many uncertainties. Several tools have been developed to improve decision-making in child protection services, but little is known about which tools are used and their advantages and disadvantages.

Aim

To provide an overview of algorithm-based decision support tools and to investigate whether these innovative tools can support social workers make better decisions.

Method

We conducted a scoping literature review (from 2000 to 2022) using seven databases. We used ASReview, an innovative program that uses machine learning and algorithms in screening large quantities of text.

Results

All 24 studies originate from North America (77 %) and Europe. Most studies (83 %) focus on tools predicting the risk of maltreatment. Some studies (17 %) focus on prescriptive tools which render advice on the best intervention strategy. Eight tools are discussed, with the ‘Alleghany Family Screening Tool’ (‘AFST’) and the ‘Child and Adolescent Needs and Strengths’ (‘CANS’) as the most frequently referenced.

Conclusion

The use of algorithmic decision-support tools in child protection services is still in its infancy. In general, considering the positive results, we believe that the CANS, the artificial neural networks tool and the boosted tree model are the most promising tools to be further tested at this stage. However, ethical concerns have been raised in almost every study, particularly concerns in respect of racial disparities. More research is required before these tools can be used on a large scale.