As a small, innovative software company, plugging away for the past 12 years to help local authorities support vulnerable people, we are always a little taken aback at the hyperbolic headlines that our work generates. There are few topics able to provoke quite the same mix of concern and conspiracy as data sharing and predictive analytics.

It’s understandable – the use of digital data is relatively new and increasingly vast. Ninety per cent of the world’s data has been created in the past two years. Every day, individuals upload over 300 million photos and Google processes more than 3.5 billion searches.

And, while we all benefit from the power of data sharing and predictive analytics – from doctors using digital sepsis alerts to pick up early indications of blood infections to television viewers getting helpful suggestions of new box sets to watch – we tend not to hail the algorithm or data processing at their core.

Rightly, we are outraged by the misuse of our data, such as when the secretive UK company, Cambridge Analytica, obtained personal details from millions of Facebook users without their consent and sold them for the purposes of political advertising. And, while they have acknowledged the need to act, the big search engines continue to be found wanting, with continuing evidence of racial bias within some of their search results.

Navigating a complex mix of risks and opportunities

Having worked at the human end of predictive analytics for a long while, it’s clear to us that taking a binary – it’s either good or evil – approach to its application is too simplistic and obscures a complex mix of risks and opportunities.  In most, real-world situations, it’s about considering whether the means can be managed in a way that ensures the end is justified.

In our work, that means helping local authorities achieve the right balance between helping vulnerable people get the support they need and impinging on their right to privacy and potentially the privacy of other citizens.  As such, it’s a common dilemma for social care, public health and many other public services – who have many statutory safeguarding responsibilities – that goes far beyond the use of data.  We argue that building in ethical considerations at every stage allows public services to use their vast wealth of data in an increasingly sophisticated way to improve their decision making for the good of everyone. That doesn’t mean sharing everything about everyone, rather extracting just the insights that are needed while also meeting data privacy regulations, avoiding automating decisions and monitoring and mitigating the risks of bias.

The public sector has employed algorithms without controversy for many years

Predictive analytics is not new to the public sector. There are many examples of risk algorithms that have been in use, without controversy, for some years, such as the Youth Justice Board’s approach to determining the risk of re-offending and the ‘DASH’ risk model used by police and social workers to understand the risk of domestic violence escalating.

The key differences between these ‘legacy’ algorithms and more recent approaches to predictive analytics mean that we now have the capability to build in more protection for individuals, not fewer. Legacy algorithms use static or fixed models that provide a score or a percentage likelihood of an event while new models, like ours, are continuously evolving as they draw on new data flows.  Our models provide a factual assessment of risks that allow the professionals using them to arrive at their own conclusions.  Our approach also embeds the monitoring of bias, measuring whether the algorithms are identifying risk disproportionately amongst individuals with protected characteristics so that action can be taken immediately if required.  We are also working hard to address transparency issues so that we can simply explain how complex algorithms work.

Ensuring a natural evolution, safely

We are already living in a world that is informed and supported by algorithms.  The implementation of more sophisticated approaches to data analytics is a natural evolution that will not only improve outcomes but that will also be more auditable and accountable. The challenge is how to transition to these new systems safely, building in safeguards and transparency so that they are understood and trusted by users and beneficiaries alike.

The power of current data analytics means that so much more of the data routinely gathered by local authorities and other public services can be used to create insights, rather than just existing in a database or, in many cases still, on a spreadsheet. That means that even more care needs to be taken about whether, why and how this information will be used. Recent news articles highlighted the prospect of using personal information about sexual behaviour. We haven’t ever been asked to include that within our software deployments but local authorities will hold information about sexual behaviour for some individuals and there are circumstances where it may well be justified to take it into account – for example, if there are concerns about child sexual exploitation. Access to this data – like all data – still has to be strictly controlled and restricted to those who need to know for a specific and sufficiently important purpose.

We are in agreement with privacy campaigners that personal data must be closely controlled and that there should be clear visibility of how it is used. The ICO has reviewed our approach to data protection positively and much of our work with local authorities is concerned with building processes that ensure data is used appropriately and proportionately. However, there does need to be more willingness on both sides of the debate to build a consensus about how we get the right balance between using data for clear public benefit and individual privacy.

Learning from the pandemic to achieve more for vulnerable people

Recent media interest in our work has been prompted by our work with local authorities to develop systems that used their existing data to identify individuals most at risk of the fallout from the Covid-19 pandemic – unmanageable debt or homelessness, for example – so that support could be targeted at them before it was too late. It was one example in a whole wave of innovations prompted by the pandemic. The opportunity now is to build and improve on those developments – and using digital technology to facilitate collaboration and protect the most vulnerable in society has to be at the top of the list given all we have learned over the past few months.