Sharing while caring: Five principles for data ethics in children’s services
In 2016, the Department for Education published a report titled Information Sharing to Protect Vulnerable Children and Families (PDF). It identified “poor information sharing between multi-agency partnerships […] as a compounding factor that can lead to the serious harm, abuse or death of a child”. The publication called for “improved information sharing for bodies with local and national responsibility”.
Despite an increasing recognition that data sharing has the potential to transform frontline services, local government is yet to undergo a data revolution. Whilst the desire may exist, the reality is that local authorities are faced with numerous barriers, including:
- a lack of process or guidance
- concerns over ethics
- limited resources to facilitate data sharing
- nervousness about potential risks
- difficulty convening and finding agreement between multiple stakeholders
Prominent amongst these are the ethical issues related to data sharing.
At Social Finance, our approach to data science and digital transformation in local government puts ethics at its core, and it is no exception for our Family Context project.
Family Context is a digital tool to support children’s social care workers. It has been developed in partnership with Leeds City Council and Stockport Council, with funding from the Ministry of Housing, Communities & Local Government’s Local Digital Fund.
We first identified Family Context as a project from a workshop with 12 local authorities to discuss improving the outcomes of vulnerable children. Since then, we have worked with Leeds and Stockport through three stages of agile delivery: discovery, alpha, and (private) beta.
We have developed a digital tool to promote multi-agency working by facilitating social workers’ conversations with lead practitioners from other services. The result is that social workers are better able to support families and safeguard children. The tool functions by allowing social workers to easily access service involvement information on relevant individuals to a child.
With data from four different services being included within the first release of the tool, data ethics has been central to our work. We have approached this by not only complying to legislation such as GDPR but also through applying five key principles.
1. Be clear about the purpose of data sharing
Data about citizens is sensitive, so understanding where and why there is a need (and how data sharing might fulfil that need) is important from the outset. It ensures we strike a balance between meeting individual rights and public interest.
We framed our initial research around questions such as ‘how can data sharing benefit children’s social workers, or the children and families they work with?’ and ‘what is the consequence of not sharing this data?’.
What we learned was that sharing data enables conversations. This means social workers can do their job more effectively, ensuring that more children and families receive appropriate support. Ultimately, this can result in fewer children ending up in care. As one chief social worker commented, “one piece of information can change the decision for a family”.
2. Use the minimum data to achieve the purpose
As the saying goes: take what you need and leave the rest. Data minimisation is based on a similar logic, where the minimum amount of data is identified in order to achieve one’s goal. This approach helps to ensure that the level of data being shared is proportionate to the level of need.
For Family Context, this involved finding the level of information that would indicate to social workers where conversations with other services were required, rather than providing comprehensive enough information that would prevent such conversations happening.
We conducted rigorous testing with social workers to identify the minimum data fields from each service needed in order to provide value. By doing so, we defined a common data model. The advantage of this was that data source owners could share less information — they only provide fields to populate the model instead of complete data sets.
3. Understand the limitations of the data
See the data for what it is, and recognise that data alone is not able to present a complete picture of reality. Instead, work within the bounds of the data and ensure the data you do have is easily understood.
Key to developing the Family Context tool was ensuring that there was sufficient data quality that passed a certain threshold. We conducted thorough data quality checks and took measures to improve the quality where we could. This means social workers can have confidence in the information presented.
It was also important to us that the users interpret the data in the intended way and are aware of its limitations. To this end, we have included a clear breakdown of what information is and is not included within the tool. For the information that is included, we repeatedly tested users’ understanding of labelled data fields to refine them.
4. Work transparently with people to build trust
People should be brought into the process if they are expected to be accepting of it. As humans, we tend to trust something if we understand it, and this also applies to how data is shared.
To achieve this in practice, we conducted user research with people who had previously experienced children’s social care to capture their voice with regard to sharing data. What we heard helped to validate our work: they agreed with the principle of how their information would be shared because they recognised that that social workers “need to protect children.”
5. Work in the open and be accountable
Working in the open has many advantages, namely generating support, sharing learnings and gathering feedback. All the while, it encourages one to be accountable about ethics.
As a project, we wanted to promote all these things. Our comms strategy for informing others about our work included publishing regular progress updates on the Pipeline and speaking at conferences. We have open-sourced our work as much as possible. We hope that this will encourage new local authorities to reuse our work and help us to build a Family Context community.
This will also allow us to gather feedback on the materials that we have made available, building on the already comprehensive quality assurance process we have undergone to improve our work.
Overall, we have found that following these five principles has kept us focused on ethical considerations throughout the development of the tool.
Since completing the recent phase of the project, the Covid-19 situation has reinvigorated the debate around data sharing. Local authorities who had already invested in their data sharing capabilities are on the front foot. With others likely to follow suit, the ethical use of data is as important as ever.