‘Big data’ and social work: using natural language processing and predictive analytics to explain decisions for children in the family courts

Abstract

Factors driving a steep rise in UK care applications over the past decade are unclear. This study addressed the issue using ‘big data’ techniques including predictive risk modelling and natural language processing (NLP). Decisions made for 67,000 children entering care proceedings in England between 2015 and 2017 were examined. National data held by Cafcass were combined with published indicators of socio-economic deprivation. In addition, five thousand social work statements were subjected to NLP to identify case level factors for eight thousand children. Ministry of Justice data were obtained and linked to provide key details concerning interim court decisions.

Findings confirmed well-recognised associations between care outcomes and area levels of poverty but identified stronger links with an index of health deprivation, based on indicators of premature death, disability, and mood disorders. The study also quantified the substantial effects of the child’s early arrangements: children already out of home at first hearing were seven times more likely to become subject to an interim care order, and to remain in care permanently, compared to those still with parents when proceedings commenced. These effects were more marked in highly deprived authorities.

NLP-identified factors were tested against comparable surveys of prevalence, and a case rating exercise conducted by family court lawyers. Higher than 89% agreement was demonstrated between automated and human ratings in clear-cut cases, whilst within ambiguity, automation offered greater consistency.

Compared to the process factors outlined above, the contribution of identified parental difficulties to care outcomes was less clear in most regression analyses. Neglect-related factors, however, apparently affected children still residing with parents at the outset of proceedings disproportionately, also playing a role in permanent care decisions. The potential contribution of innovative ‘big data’ technology to social work research and practice was confirmed. Future research avenues and methodologies are highlighted.

Date of Award	Oct 2021
Original language	English
Sponsors	Department of Education and Learning
Supervisor	Brian Taylor (Supervisor) & John Mallett (Supervisor)