Originally appeared as an Op-Ed in the Guardian Media and Tech Network.

Current data-sharing arrangements are more akin to a form of data feudalism than data democracy

The 1832 reform act was a response to social change in the evolving industrial landscape of pre-Victorian Britain. It introduced new parliamentary constituencies. It eliminated “rotten boroughs”: small constituencies controlled by a handful of voters. It also gave the vote to those men with “freehold ownership of land”.

Votes to those men who own freehold on land is distant from universal suffrage, but the freehold land movement soon developed. Land societies purchased large tracts of land with the express intent of subdividing it and allocating the freehold to individual members. My own house was constructed on land bought in exactly this manner.

Land societies were an early form of mutual: a way of sharing resources for mutual benefit. A large freehold land purchase would have been beyond the means of any individual, but by combining resources the individual benefits could be obtained.

In the recent NHS-Google DeepMind data sharing deal the Royal Free Hospital trust shared up to 1.6 million patients’ data with Google DeepMind, the UK-based artificial intelligence company.

By combining patients’ data, a better understanding of disease can be developed. It’s a combination of resources for greater individual benefit, just like a mutual.

However, the sensitivity of this data means that there are challenges of “duty of care”. In an NHS data-sharing deal, who acts on the patients’ behalf? The clinicians? The employees at DeepMind? Who provides oversight?

As the New Scientist has revealed, it’s a thorny subject. Royal Free and DeepMind contentiously claim that they are using the records for “direct care” of patients, meaning the hospital trust is able to oversee the implementation of the data sharing partnership.

The patients themselves have no direct voice in the regulatory framework. The hospital uses the principle of implicit consent and is under no obligation to even make the individual patients aware of the deal. This seems unacceptable.

The NHS recognises the principle of patient ownership of data, but in practice it balances that with the wider “patient interest”. It aims to derive the mutual benefit, just like in a land society, but members of a land society were voluntary participants. The board of trustees of the land society were compelled to operate within the constraints of the contract they formed with their members. No such protections are afforded to today’s patients.

Legally the patient is referred to as a data subject. This term has unfortunate, but perhaps not inaccurate, connotations of royal prerogative. Data is managed on the subject’s behalf by the data controllers, the NHS or Google. Control of data needs to be returned to the patient.

To restore individual rights, and produce more of a sense of citizenship, we might consider the idea of a “data trust”: a mutual organisation formed to manage data on its members’ behalf. Data subjects would pool their data forming a trust, stipulating conditions under which data could be shared. The trust would retain a duty of care without conflicting goals such as making a profit or furthering a research career.

As well as medical data, we could imagine data trusts set up for more trivial concerns: improving product recommendation or matching consumers to suppliers. An ecosystem of data trusts would have the advantage of providing individual choice over who is best placed to manage their data. It would enable co-evolution of regulation and greater democratisation of the data landscape. By aligning data control with data provenance, it would respect a form of data ownership rights.

Trusts would be large enough to be effective partners in controlling how data is used. The legal mechanism would empower each trust to prioritise the data subjects’ interests in negotiations. By collating data, the trusts would become powerbrokers themselves, data-brokers. The trustees become the guardians of individual interests. Oversight of the trustees would be through the founding constitution of the trust.

A data democracy implies data governance by the people, for the people and with their consent. Current data-sharing arrangements are more akin to a form of data feudalism where data is managed on our behalf with irregular oversight. An ecosystem of data trusts would shift the balance back to better reflect the concerns of individual data subjects. It would ensure that when we share our data we don’t also cede our rights. Like the land societies of old, data would be used for our mutual benefit governed by our mutually determined conditions and with our mutual consent.