- Reviews the ethical criteria that would make the use of machine learning (ML) in children’s social care (CSC) justifiable and examines the problematic contexts in which such criteria may not be met;
- Identifies requirements and best practice for the responsible use of ML in CSC;
- Presents recommendations for a way forward.
The aim of the report is to answer the question: “Is it ethical to use machine learning approaches in children’s social care systems and if so, how and under what circumstances?”. The findings are aimed at data scientists, policy makers, local authority (LA) children’s services departments, civil servants, and citizens.
How we went about it
Following a request for proposals, we commissioned The Alan Turing Institute and the Rees Centre, University of Oxford to undertake the review. The research is informed by a review of the literature, the integration of multiple existing ethical frameworks in social care and ML, a stakeholder roundtable with 31 participants, and a workshop with 10 family members who have lived experience of children’s social care.
The Turing and the Rees Centre mapped out common motivations and moral foundations to propose a list of ethical values, practical principles and professional virtues which can be used as guardrails for the responsible use of ML in CSC. The aim of presenting these practical ethics is that they can be actively adopted by all affected stakeholders as a vehicle of common commitment to the shared purpose of using these technologies exclusively in ways that advance public wellbeing and benefit society.
To answer the question whether ‘Can we do this right?’, The Turing and the Rees Centre, present standards for best practice across ML’s design and deployment lifecycle, paying special attention at each step of the way to the CSC context. They cover the data quality and use, model design and implementation.
The report should be utilised both as a means to reflect on questions about the appropriateness and justifiability of using ML applications in CSC (both for specific use cases and in general) and as a preliminary guide for developing projects involving ML in CSC.
The review concludes with some preliminary recommendations for steering the present direction of ML in CSC, namely:
- Mandate the responsible design and use of ML models in CSC at the national level.
- Connect practitioners and data scientists across local authorities to improve ML innovation and to advance shared insights in applied data science through openness and communication.
- Institutionalise inclusive and consent-based practices for designing, procuring, and implementing ML models.
- Fund, initiate, and undertake active research programmes in system, organisation, and participant readiness.
- Understand the use of data in CSC better so that recognition of its potential benefits and limitations can more effectively guide ML innovation practices.
- Use data insights to describe, diagnose and analyse the root causes of the need for CSC, experiment to address them.
- Focus on individual- and family-advancing outcomes, strengths-based approaches, and community-guided prospect modelling.
- Improve data quality and understanding through professional development and training.