Building trust in how you handle data: a hierarchy

The ODI’s CEO Jeni Tennison shares four layers to building trust in how you handle data, from basic compliance to making positive efforts to share its benefits

At the ODI we aim to help build an open, trustworthy, data ecosystem. Part of that is helping build good practices within organisations, so that they behave in more trustworthy ways when they collect, use and share data. This includes things like providing training in data ethics or advice on developing good data governance policies and practices.

But there is a difference between being trustworthy with data and being trusted with it. We'd hope that one would lead to the other – that trustworthy organisations would be trusted and untrustworthy ones untrusted – but this isn't always the case. Many organisations are driven to trustworthy data practices because of their values and a desire to do the right thing. They also care about their reputation: they want to be trusted as well as trustworthy.

Gaining trust requires organisations to go above and beyond good data governance practices. I now think of the requirements as being like a hierarchy of needs, because each layer is harder to define and attain than the one below.

That doesn't mean they can only happen in order. It also doesn’t mean that demonstrably doing things at the top of the pyramid means you don’t have to do any of the things at the lower layers. If you're aiming to win people's trust by engaging through a data ethics board, some of the first things they're likely to ask about are legal compliance and visibility of ethics practices. But different organisations will have different starting points. A cooperative, for example, might already have lots of trustworthy practices that make it easy for them to demonstrate both community accountability and fairness with their members; they will still have to work to develop trustworthy data practices and to integrate data into their engagement and how they think about equity.

Layer one: privacy and security

At the bottom of the hierarchy are hygiene factors: basic privacy, security, compliance with GDPR and other relevant legislation. Don't leave USB keys containing data about people on trains. Don't let hackers access your customer database. Monitor and audit how data is being used within your organisation, or by others if you share data with them. Basically, don't do anything that would get you fined by your local data protection regulator. Naturally if you don't do these things (and they find out) people rightly aren't going to trust you with data.

Layer two: ethics and transparency

The next layer up contains the kind of good practices that are commonly recommended and moderately easy for an ethically minded organisation to achieve. Have some ethical principles; publish them; embed them into how you collect, use and share data; talk about the decisions you've made and why. Go beyond the minimum required for compliance with GDPR and adopt good design practices that actually make it easy to opt out and in, to understand what you do with data. Employ data minimisation techniques. Do privacy by design. These are the kind of things that provide visibility, openness and transparency with customers and third parties.

This second layer of practices can help to foster trust in how you handle data. Equally, however, it enables those who disagree with the choices you make to be a lot more specific about what they're unhappy about. They are difficult to adopt because they impact an organisation's confidentiality, and in some cases its competitive advantage. Openness and transparency can also draw attention to what a trustworthy organisation is doing with data, while perhaps an untrustworthy organisation that isn't as open will escape the same scrutiny.

Organisations that only reach this layer will also often face criticism because they don't give their customers or critics any direct power to change what they do with data. The organisation remains the arbiter of what is acceptable and what is not, and what is visible and what is not.

Layer three: engagement and accountability

The third layer of the hierarchy gives some power to other people. Here you would institute independent external data ethics committees that have the power to halt programmes or products. You have data auditors who can issue assessments that impact your ability to trade. You have proper engagement and co-design with customers where you don't do things they tell you are unacceptable. You have routine publication of transparency data and information that makes it possible for others to monitor you from outside. This is a layer that focuses on accountability and engagement, listening as well as telling, recognising you are not the arbiter of what good looks like.

These things are hard to do for any organisation because they reduce autonomy (which organisations like to have as much as individuals like to have). They add delays to decision making. They affect an organisation's ability to achieve their goals because they rule out some paths to those goals. This, of course, is the point of this layer: having a regulator (in the broadest sense of the word) that prevents runaway – and long term damaging – behaviours.

But even still... I have seen data governance structures proposed and adopted along these lines, and even still these organisations face doubt and questions and distrust about how they collect, use and share data. It starts to feel like organisations can't win.

Layer four: equity and fairness

So the final layer of the hierarchy is about equity, about who benefits from the use of data, and about people and communities. It is about not just doing things that aren't bad, but doing things that are positively good, that benefit others. And this layer isn't about the specifics of how data is collected, used and shared, but about how the organisation – or even the sector – behaves as a whole. It's about not avoiding taxes. It's about treating workers fairly. It's about (data and other) philanthropy. It's about avoiding market manipulation or aggressive takeovers.

The reputation of an organisation can be affected by what it does with data, but equally the way people feel about what an organisation does with data depends on the organisation's reputation. The NHS is in part trusted with data because people trust the NHS with their health. The degree of trust in Sidewalk Labs and DeepMind are influenced by people's trust (or lack of) in Google.

Data projects within or involving particular organisations have limited opportunity or leverage to affect broader aspects of how those organisations work. Achieving equity around data requires organisations to align their revenue models and other incentives around benefiting people affected by their use of that data. That's hard to do without an organisational form that prioritises purpose and mission, and provides appropriate external checks and balances, such as a charity or social enterprise. This is one of the reasons we have become particularly interested in data trusts and other data institutions at the ODI: they provide a mechanism to have organisations that can concentrate on stewarding data in independent, equitable ways.

The other thing to recognise is that there might be natural limits to the level of trust any particular data project or organisation can achieve on its own. Sometimes you might have to simply accept those limits. Other times you might be able to shift them, even ever so slightly, by contributing to activities that increase broader trust, and more trustworthiness, in your wider organisation, sector, or the way everyone collects and uses data.

Those leading data projects need to recognise being trustworthy and trusted with data is not a simple checkbox exercise where you can just publish some ethical principles, set up an ethical advisory board, and be done with it. It can require deep, challenging, changes to how organisations interact with consumers and citizens, to their decision making and to their business models. But equally, those of us who advocate for better data practices also need to recognise the enormity of what we're asking for (and keep asking for it). This is a long journey; it will take many small steps to get us there.

Find out more about how you can build trust in how data is handled