Etic Lab is delighted to announce the publication of its new report, Data Federations: Digital Collaboration Without Data Sharing. It documents the work we’ve been doing over the past six months with the support of the Open Data Institute’s Stimulus Fund.
The purpose of this project has been to develop and test the concept of Data Federations – a new model for the organisation of digital collaborations, based on the use of distributed and/or privacy-preserving technologies such as federated learning or distributed data mining. With the use of such technologies, it is possible for partners to commit their data to a common project without requiring that it is shared or collected in a central location, allowing for the development of collaborations which might otherwise be impossible.
From our previous work in the charitable sector, we’ve found that whilst the general benefits of strategic data use are increasingly well understood, many organisations are unable to access them, whether because of a lack of resources or concerns over data privacy and security. Our hope with Data Federations is to provide a sociotechnical framework which allows people to overcome these constraints in the context of a purpose-driven collaboration, extending he availability of data analytics and machine learning to coalitions of smaller, less well-resourced organisations, who can then use them to advance their goals and values.
With the support of the ODI Stimulus Fund, we were able to take time to formalise our ideas around what it might take to to develop a working Data Federation, and then to test them with some real-world partners in the charitable sector. This experience has taught us a series of valuable lessons, which we hope will be of use to anyone trying to take on this kind of work in the future:
- The time and labour costs imposed by data standardisation were one of the main barriers to collaborative digital projects that we encountered. A major part of the appeal of the DF concept is that it offers a way for partners to generate shared value without requiring them to first convert their data into a common standard.
- The mitigation of monolithic obstacles such as privacy or non-standard data generally results in the surfacing of a range of context-specific barriers, based on the particular histories and characteristics of the organisations in question. Addressing these barriers is one of the main tasks of an effective Data Federation.
- If they are to be effectively and responsibly deployed, tools such as federated learning should not be seen as a “solution” to a set of narrowly-defined problems, but a set of possibilities which should shape every dimension of a project’s design from the very beginning.
To find out more, read our report!
Reflecting on the ODI Stimulus Fund project, Etic Lab Research Officer Richard Woodall said: “We’d like to thank the ODI for their support over the last six months, which has allowed us to engage in an experimental research project in a field which doesn’t generally have access to cutting-edge digital technologies. It’s been a real pleasure to be a part of a community of passionate and imaginative practitioners, and we’re excited to see how everyone’s work develops in the future.”
For Etic Lab, the experience of the last few months has convinced us that Data Federations have something unique to offer organisations who are considering digital collaboration. Data privacy is now an issue for both legislators and the public as never before, whilst Google’s recent announcements regarding its “Privacy Sandbox” indicate that technologies such as federated learning have progressed from conception to development. It’s high time, therefore, that we start thinking about how to use these tools to create public value in a responsible, ethical and equitable manner. What governance models, organisational frameworks and codes of practice are needed to ensure we can use privacy-preserving data technologies to create public benefit? This will be the focus of our work in the coming months.