Predictive Risk Modeling in child and family welfare

Andrew Whittaker, Ravit Alfandari*, Mary Baginsky, Campell Killick, Beth Coulthard, Brian Taylor, John Mallet, Liesanth Yde Nirmalarajan, Yuval Barak-Corren

*Corresponding author for this work

Research output: Contribution to conference without publisher/journalConference abstract for conferenceResearch


The tensions and challenges associated with the numerous tasks of decision making and risk assessment social worker have to make throughout their day are enormous. Dealing with uncertain and unpredictable practice situations, social workers often have to make difficult judgments and emotionally charged decisions quickly and with insufficient information. Recent initiatives to enhance social workers’ decision-making practice, particularly their capacity to timely identify and evaluate possible risk, focus on developing automatic decision support systems based on big data and machine learning (ML) algorithms.

In a nutshell, data-driven machine learning models use vast datasets that integrates service records to link individual's data to certain high-risk outcomes like suicide, domestic violence, or child abuse. Typically, decision support systems automatically generate assessments and provide actionable recommendations for consideration at appropriate decision-making points. An example from the healthcare field, are pioneering clinical decision support systems in hospitals, embedded in patient electronic health record that alert about high risk of child psychical abuse and offer recommendations or links to official guidelines. Some evidence indicates that these systems can increase identification of children with possible physical abuse, yet further research is needed.

Against this progress, a debate has been evolving in the literature about the possible professional and ethical implications for social work practice taking this direction. How would big data and algorithm-driven practice impact social workers' capacity to exercise professional discretion? How would it effect their engagement with clients? Can we assure these massive databases won't be misused? Would machine learning models reduce or increase the impact of social group biases (e.g., racial and socioeconomic disparities) on decision making?

We propose to organise the pre-conference event around these fundamental questions. We will start with a number of DARSIG members from different countries introducing pioneering local initiatives of incorporating big data and algorithms in social work decision making practice. We will then conduct an open discussion in a World Café format around the implications of such developments for research, practice and policy. This will be followed in the second half of the day by our ACM and DARSIG symposium.

1) Beth Coulthard, Brian Taylor, John Mallett
Using ‘big data’ to explain decisions for children in the family courts

2) Liesanth Nirmalarajan
Predictive risk modelling in child and family welfare

3) Ravit Alfandari & Yuval Barak-Corren
Integration of welfare and healthcare data for early identification of child maltreatment

World Café questions:

1. What initiatives to use big data and machine learning predictive models in social work research and practice are you aware of?

2. What possibilities using big data and machine learning can generate for social work practice and research?

3. What are the challenges that would need to be addressed when using big data and machine learning in social work practice and research?

4. How can we ensure ethical practice when using big data and machine learning in social work practice and research?
Original languageEnglish
Publication date12 Apr 2023
Publication statusPublished - 12 Apr 2023
Event12th European Conference for Social Work Research: Social work research through and towards human relationships - Università Cattolica, Milano, Italy
Duration: 12 Apr 202314 Apr 2023


Conference12th European Conference for Social Work Research
LocationUniversità Cattolica
Internet address

Cite this