No tracking. No cookies
Subscriptions
Subscriptions
Hubris and Technology TriumphalismThe Government’s £12m COVID-19 Contact Tracing App
Paul Dempsey shows how ‘big data’ cannot replace ‘no data’ and that the UK’s centralised plan to combat the Coronavirus was flawed at the outset
On Monday, in the House of Lords, Lord Bethell revealed at the cost of the Government’s COVID-19 contact tracing app has reached a staggering £11.8m, and it is still unworkable.
The Government has already been criticised for failing to take a system-level approach to the complexities of the Coronavirus outbreak. Though its current approach to track-and-trace is system-based, the problem is that it is the wrong system for the UK’s immediate needs.
NHSX, the healthcare technology arm, has been directed to take an inappropriate and overly complex technology-led approach. As Jason Bay, product lead on the TraceTogether Covid-19 app in Singapore points out in his April Medium post, apps and other digital tools make an incremental contribution towards effective contact tracing. They leverage what you already have. They are not – and must not be – at its core.
“There are critical factors that a purely automated system will not have access to. You cannot ‘big data’ your way out of a ‘no data’ situation. Period. Any attempt to believe otherwise is an exercise in hubris and technology triumphalism. There are lives at stake. False positives and false negatives have real-life (and death) consequences. We use TraceTogether to supplement contact tracing – not replace it.”
Yet that is what appears to have happened in the UK
Tha Apps’ Role in Track and Trace
The UK system is a three-legged stool. There is the app. There is a mostly centralised network of 25,000 contact tracers, under the aegis of Baroness Dido Harding and administered by contactor Serco. These feed data in a common format into an AI-based hub. This is enabled by data cleaning and warehousing, and analytics from two other contractors, respectively Palantir and Faculty.
The problem with this architecture is that it imposes an architectural and engineering rigidity upon a system that has already been described as “not fit for purpose” in a lengthy report by the Independent SAGE group of scientists and public health experts.
The effective failure of the NHSX app project illustrates how technology is becoming more of a hindrance than a benefit because of how its objectives have been defined.
There have been claims that NHSX has treated at least one other UK app project as an “enemy” rather than a potential collaborator and developers elsewhere in the world confirm that they have had much less contact with or from our shores compared with other countries.
NHSX insiders dispute this, but do say that the nature of the UK approach has rendered much of the research elsewhere hard, if not impossible, to reuse. They add that the terms of commercial confidentiality with contractors across the UK system has prevented them from consulting outside the project’s members. At a time when pretty much every other app team has sought to open-source its work — Taiwan even developed many of its tools with the help of public hackathons — this is bizarre, to say the least.
Moreover, while health secretary Matt Hancock now says that the app can be fixed by moving from an entirely bespoke (and, it would appear, proprietary) strategy to one that leverages the best of the open platform offered by Apple and Google, this may be hard to do.
Apple contradicted the minister’s claim that active collaboration had already begun and the main elements the UK wishes to change are thought to be at odds with the company’s stance on privacy — one that has already seen it ready to stand firm against the FBI.
The app, then, is still an issue. Even if that can be overcome, the real challenge is the architecture.
The Architecture
Its degree of centralisation is driven by the needs of an AI (clean inputs from a manageable network from which insights are derived at the macro level) rather than the demands of an outbreak, particularly one where you are trying to build out the architecture as a virus is already moving through the population.
Other countries have used much-hyped digital tools during the outbreak, but in every case they have been subordinated to manual tracing. Let’s say it again: incremental, not core.
For example, South Korea has pulled on just about every tracking technology available – almost certainly too many for a directly equivalent app to overcome resistance from the UK population. As developers on its digital side explain, the results are still largely intended to give manual tracers as fine-grained a map of the outbreak as possible, as well as ensure compliance with self-isolation.
China, for its part, has also employed extensive smartphone-based tools, including digital passports, but while these too raise surveillance concerns, its healthcare professionals still point primarily to its localised contact tracing as key, such as the 8,000 tracers that contributed to the effort in Wuhan.
Closer to home, there is Germany. A significant part of the £18m it has spent on its Corona app, released for download last week (including the open source code), went towards meeting the demands and requirements of local governments and healthcare networks.
In this context, you must conclude that the UK has exhausted a lot of time and effort – when both are in the shortest supply – trying to reinvent the wheel.
This speaks to another general concern that has surrounded Boris Johnson’s government.
Technology Triumphalism versus Humility and Progress
Technophilia is generally seen in engineering as a good thing, but placing a faith in technology that is beyond what it can deliver can be extremely dangerous. For example, when Matt Hancock recently sang the praises of cutting-edge genetic testing, he needed to be, shall we say, politely talked down by a raft of experts.
The signs are that the Government is persisting with its harmonised strategy. As recently as Monday’s Today programme, Home Office minister James Brokenshire pointed out that the app was not so much the challenge as “how it all fits together”.
Perhaps the UK could one day have a “world-beating” contact-tracing system with a best-in-class AI to combat any future pandemic. These things take time. Look at how many years it is taking engineers in a safety-critical industry like automotive to develop viable autonomous driving. Most people still think Tesla is less than 10 years old. In fact, it was founded and has been working on R&D for its cars since 2003.
Yet, having said that, tearing the wires out of the back of what the UK has already built – or even mothballing them – simply won’t do either. What next?
This proposal has three components that can immediately start to be implemented so that the country can benefit.
1. Prioritise contact tracing under the traditional model at local level and redistribute the centralised team to work under the management of regional public health authorities that are already carrying out a significant part of the work. Data can still be fed back centrally.
2. Reconfigure the app towards as much off-the-shelf reuse as possible, with a particular focus on how that might be done based on existing code with regionalised functionality (e.g. Corona).
3. Downgrade the AI-based analysis component to a pathfinder R&D project, intended to evolve and adapt to the existing best practices in contact tracing rather than having it try to enforce new ones. Privacy criticisms that have arisen here also need much more consultation.
This will not be easy. Pandemics are not easy. There is evidence that it would be doable and is a quick and practical way to deliver a track-and-trace infrastructure that is fit for purpose and, critically, better tuned to the more local resurgences of Covid-19 that many – including both the prime minister and health secretary – expect to be the next major challenge.
All that will leave just one more thing: the government finally needs to admit it has made a mistake. Sometimes, humility will take you further than simply moving fast and breaking things.
A version of this article first appeared in Engineering and Technology magazine