Our research aimed to fill a gap in knowledge about parental social licence for operational data linkage and predictive analytics. We designed the research to consist of cumulative mixed methods data generation and analysis work packages, moving from an analysis of documents, to a representative survey, to group discussions, and then to individual interviews. This mix of methods took advantage of the combined strengths of systematic breadth and contextual depth approaches.
Our starting point was a discourse analysis of policy and public discussion. We looked at the range and content of discourses about data linkage, predictive data analytics, and family intervention, and drawn out the supportive and critical rationales and their contingencies. We analysed national and local government reports and statements that are key markers in the field; promotional material from data analytics companies serving local authorities; reports and statements from relevant charities and advocacy groups. We adopted a framework analytic method to cover discursive rationales for and against operational data linkage and predictive analytics across the sets of materials and comparing them.
Findings from the documentary analysis provided a grounding for the next data generation packages, including informing the development of the survey questions.
Freedom of Information requests:
We also submitted Freedom of Information requests to local authorities. We asked them about their use of data analytics, predictive analytics, or algorithmic automated systems for processing, risk assessment, scoring systems or automated decision-making in relation to child and family services.
You can read the findings from the Freedom of Information requests in: Freedom of Information Requests on the Use of Data Analytics in Children’s Services: Generating Transparency, Gillies, V. and Gardner, B. with Edwards, R. and Gorin S., (Project Working Paper)
The point of the survey was to see if parents of dependent children agree or disagree about what is acceptable or unacceptable in relation to linking data collected by government services and using predictive analytics for risk modelling as a basis for intervention.
To build a systematic understanding, we commissioned an online and telephone probability-based survey at NatCen, an independent social research organization, designed to be representative of the population and produce reliable estimates of opinions. The sample consisted of 843 parents with a range of demographic characteristics.
We asked questions about awareness of the collection and joining together of data sources, assessments of a range of early intervention rationales for data linkage and predictive analytics, and acceptance of and trust in various bodies and services to undertake linking of different types of information. Most of the questions were multiple choice responses to statements, but around a third were attitudinal vignettes describing hypothetical scenarios in which data linking and analytics might be used to target families for intervention. We adopted the consensus baseline approach for analysing our survey data.
Findings from the survey about the ways in which views on operational data linkage and analytics are segmented provided a basis for the sampling of sub-population groups for the group method in the next data generation work package.
You can read the findings from the survey in: Data linkage for early intervention in the UK: Parental social license and social divisions, Data & Policy, 3. Edwards, R. Gillies, V. and Gorin, S. (2021).
You can also read about an earlier pilot survey.
Group discussions offer insights through group interactions, allowing participants to express the issues that are relevant to them and develop their views in the discussion. Guided by the findings from the survey, we held nine group discussions with specific groups of parents, including parents in professional occupations, lone mothers, fathers, and so on. We had an average of four parents in each group. We mainly used local community groups, neighbourhood networks, voluntary support groups and childcare services for access.
The topic-based discussions were centred on parents’ views on the acceptability of data linkage and predictive analytics, rather than personal experiences. We also used hypothetical case studies. We did a discursive analysis of the negotiated production or contestation of social licence in the interactive exchanges.
Our group discussions were designed to focus on opinions about social acceptability rather than personal experience. But we designed the project to delve deeper into personal experiences through the next method.
Interviews enable participants to discuss issues on their own terms, and so we used them to provide in-depth consideration of how perceptions of operational data linkage may be shaped where parents have experience of engaging with family support and intervention services. We carried out interviews with 23 parents of dependent children who used or were subject to family support or intervention services, mainly accessed through support services and groups for service users.
The interviews were topic-centred and fluid, covering the interviewees’ experiences and views of their position in relation to data held about them by services they access, and their particular bases for trust or distrust in operational data linkage and predictive analytics. We analysed this material using inductive coding and theme development to identify patterns of meaning and identification of contingencies.
N.B. A month before the project was to start, we faced Covid-19 lockdowns and social distancing measures. We were able to carry out the documentary analysis and the representative survey as planned, because neither of these involved face-to-face interaction. However, we had envisaged in-person group discussions and individual interviews. But this was ruled out by Covid, so we had to switch to online recruitment and data generation for both. The quality of group discussions or the individual interviews did not suffer.
+ + +
You can access the findings from our various methods that we have published so far on our Resources and Outputs page.