Data Trusts Initiative
Data trusts are an approach to rebalance the power asymmetry between data controllers and data subjects.
The first draft of The Atomic Human was written in 2015.1 The last chapter was tentatively titled “A Data Democracy”. It outlined a vision for a new type of institution, one that would rebalance the power asymmetry between data controllers and data subjects.
Prompted by challenges generated by the DeepMind-Royal Free Hospital data sharing agreement, I wrote about the proposal in an article for the Guardian in 2016.2 The purpose of a data trust is to rebalance the power asymmetry between data controllers and data subjects.
This purpose directly addresses a challenge highlighted in The Atomic Human - that data protection regulations don’t “directly protect us regarding the ‘inconsequential’ decisions” made by platforms that can accumulate to have major effects, a key concern discussed in yesterday’s post on power concentration and automation. I agree with the Information Commissioner John Edwards who suggested that existing data protection legislation could potentially address this, but I think the key word is directly, without institutional frameworks like data trusts the rights that data protection confers on us are not practically accessible and effective at scale.
The idea of a data trust is to use the legal infrastructure associated with trust law, that would be enforced by a court of equity,3 to enable a new type of institution that could provide governance based on the collectivised rights of data subjects.
Back in 2016 I wasn’t aware of a phenomenon known as “trust-washing”. Sometimes the term “trust” is applied to an institution that borrows the terminology but isn’t governed by a court of equity. For example, NHS Trusts are not governed by a court of equity so are not trusts in the legal sense. When the notion of data trusts was picked up in the wider public domain, it felt like it was often without the original purpose of forming institutions that better reflect the aspirations of their members, and more as a route to trust-washing data processes.
This is an example of the type of power concentration mentioned in yesterday’s post. For many existing interests, the easy route is not to engage with the data subjects, but to look for quick solutions that enable them to push forward with their core agenda.
To better clarify the purpose and role of a data trust I worked with Sylvie Delacroix on a paper that emphasised the purpose of the trust as a means for enfranchising data subjects.
The Data Trusts Initiative followed on from this work and was funded by the Patrick J. McGovern Foundation. Together with Jess Montgomery and Sylvie Delacroix we proposed an initiative that would fund pilot data trusts as well as build a community of practice around data trusts with supporting research and policy work. This enabled us to provide easily accessible information, e.g. on the EU’s Data Governance Act or on the mandatability of data rights we receive under GDPR. We worked with the Ada Lovelace Institute, the AI Council, the Open Data Institute, and the City of London Law Society to clarify the legal mechanisms behind data stewardship.
But the main objective of the initiative was to fund data trust pilots. The aim being to shift the conversation from perceptions of the challenges around data trusts to experience of the actual challenges.
The three pilot projects were the Brixham Data Trust, a place-based data trust led by Pamela Charlick and Natasha Nicholson of charlick + nicholson architects. The Born in Scotland Data Trust, focussed on supporting a birth cohort study, led by Jessica Bell from the University of Warwick and the General Practitioners Data Trust, led by Sarah Devaney from the University of Manchester.
This video from Dan Andrews of Scribeysense.com is our summary the work of the initiative.
Progress on establishing data trusts has been slower than one might have hoped. There have certainly been challenges in the political landscape, but I think this also reflects the difficulty of creating new institutions that fulfil the expectations of society and can assimilate the complexity of the data systems that we have created.
As we saw in the discussion of bridging domain experts to AI capability, we need to maintain feedback loops between technology, its users and the broader societal context. Data trusts are proposed as a framework for developing institutional solutions to these problems. Misunderstandings and clarifications arise from different perspectives worrying about different parts of the problem. A given pilot project won’t represent all the challenges we face, but they allow discussions to focus on problems that arise in practice rather than hypothetical challenges that may arise in theory.
Progress continues to be made with a recent report from the Ada Lovelace Institute on participatory and inclusive data stewardship. Through the ai@cam policy lab we have also been convening the community around the general area of data intermediaries to support the new government in understanding the possibilities in this space.
Our work on data intermediaries exemplifies the dual challenge we’ve been exploring: creating institutions that can bridge between domain expertise and AI capability while also addressing the power concentration concerns highlighted in our x-risk decomposition.
Data trusts aim to serve both purposes - they should facilitate use of data which is aligned with the aspirations of those who generate the data and in doing so they should be addressing harmful concentrations of power.
-
This draft was a very different book, see The Atomic Human for more details. ↩
-
See this archive piece for more details. ↩
-
The Court of Equity is a court of law in England and Wales that deals with matters of trust and trust law. It is the successor to the Court of Chancery, which was established in the 14th century. Until 1400 the Court of Chancery mainly dealt with abuses of power in the feudal system. For a detailed history see Wikipedia. ↩
Click to see what the machine says about the initiative and the book
Machine Commentary
NL: Could you provide a machine commentary for this post that relates the work of the Data Trusts Initiative to the recent Ada Lovelace report on participatory and inclusive data stewardship?
Claude: Based on the Ada Lovelace Institute’s 2024 report on participatory and inclusive data stewardship, we can see how the Data Trusts Initiative’s work has influenced and been extended by subsequent research:
Evolution of Data Trust Concepts
Power Dynamics & Participation
The report builds on DTI’s work by:
Implementation Insights
Validates several DTI principles while adding new perspectives:
Future Directions
The report suggests areas where DTI’s work could be extended:
Source: Ada Lovelace Institute’s “Participatory and inclusive data stewardship” report
NL: Can you provide further machine commentary on how DTI is fitting with the approach described in this post?
Claude: Alignment with the Five Ps Framework
Purpose
People
Projects
Principles
Process
The initiative demonstrates strong alignment with the five Ps framework, particularly in maintaining clear purpose while developing practical projects and processes. The slower-than-hoped progress noted in the post reflects the challenge of balancing these elements when creating new institutions.
NL: Can you speculate why the progress on data trusts has been slower than we might have hoped?
Claude: Factors Contributing to Slower Progress
These challenges reflect the ambitious nature of creating new institutions that must simultaneously address legal, technical, social, and economic aspects of data governance. The slower progress might be seen as a natural consequence of trying to build robust, lasting solutions rather than quick fixes.
Analysis Through the Five Ps Lens
How the Framework Helps
Purpose
People
Projects
Principles
Process
Remaining Gaps
The 5 Ps framework provides valuable structure for addressing many implementation challenges, but some fundamental difficulties around resources, scale, and cultural change may require additional approaches or tools.
NL: Can you describe how this initiative has inspired or been inspired by ideas in The Atomic Human?
Claude: Connection to The Atomic Human
The initiative serves as a concrete example of how the book’s theoretical framework for maintaining human agency in technological systems can be implemented in practice, while highlighting both the possibilities and challenges of institutional innovation.