Where did it all go wrong for the UK’s Covid-19 contact tracing app?
19th Jun 2020
Back in May, the UK’s oh-so-imminent contact tracing app was hailed as the jewel in the crown of its “world-beating” tracing programme. In the three months since, it’s charted an undignified fall from grace. Now, the project head-down in a ditch, legs akimbo, the government has finally admitted defeat and meekly acknowledged a shift to Google and Apple’s model.
A potent concoction of technosolutionism and hubris over the UK’s ability to solve impossible technical problems led the country down this tortuous path. The fatal flaw was the decision to go for a centralised, as opposed to decentralised, app. Google and Apple had said that only decentralised apps would be able to run continuously (i.e. work effectively) in the background on their handsets (i.e. 99 per cent of all phones).
When it became clear the two tech giants were unbending on this decision, countries with more humility – Germany, Italy and Denmark – decided to switch to a decentralised approach. Many more chose to develop decentralised apps from the beginning. But some, including the UK, Australia, Norway and France, decided to plough ahead regardless. It’s not ended well for them.
In early May, while the UK was still doggedly pursuing its centralised app, Australia was already reporting issues with its own Covidsafe app. The government admitted that the app wasn’t working properly on iPhones, and signalled that it was considering switching to Google and Apple’s system instead. You might think that at this point, the UK would’ve sat up and taken notice – if Australia wasn’t able to solve the issues plaguing centralised apps what chances would the UK have? Unsurprisingly, it did not. Norway’s centralised app also ran into trouble. This week it was forced to delete all the data collected by the app because it was too invasive.
The point of choosing a centralised app is to create a centrally stored pool of data that can in theory be used for epidemiological modelling. With a decentralised approach, the data remains on the phone, although advocates for this approach argue that epidemiologists are still able to gain relevant information from decentralised tracing apps.
In the UK’s case, there was another reason that it opted for a centralised app. Most of the countries adopting decentralised methods were asking users to enter positive coronavirus test results into the app as evidence they had been diagnosed with the virus. But due to the UK’s lack of available tests, its app instead asked people to enter symptoms – because this was a more unreliable indicator of whether someone had Covid-19 or not, the UK government decided that a centralised method with more central oversight would be required to make it work.
However, Google and Apple were set on a privacy-preserving decentralised framework that didn’t let governments collect as much data on citizens, meaning less likelihood for mission creep – something that privacy experts repeatedly warned about with regard to a centralised app.
A trial on the Isle of Wight generated mixed reports about the effectiveness of the UK’s app. At the time, some technologists claimed that the UK’s tech savvy had managed to leap Apple and Google’s barriers. They claimed it worked sufficiently well because although Apple phones couldn’t run the technology in the background, Android phones were able to ping the app on iPhones, waking it up and allowing it to register contacts. The over reliance on Android phones for this approach to work prompted the coining of the term “Android herd immunity”.
Despite public commitment to its app, in May the UK government privately asked one of the companies carrying out work on the app, Zuhlke Engineering, to start testing out how easy it would be to transition to Google and Apple’s model. Shortly afterwards, the government began to step back from the app, with the head of the contact tracing programme Dido Harding saying that it would be the “cherry on top” of the wider tracing system rather than an integral part.
Recent research finds that the app only recognised 4 per cent of Apple phones and 75 per cent of Android models in the trial. By comparison, the Google-Apple model can identify 99 per cent of phones. The UK government claims that the Google-Apple model has issues with distinguishing between whether someone is one metre or three metres away and that its own app is far better on this front. However, an Apple official speaking to the Times said they were unaware of what the government was talking about, and hadn’t heard about any proximity issues from Germany, Italy, Holland or Ireland, where the framework has already been used to build apps.
Health secretary Matt Hancock snatched up Apple as a useful scapegoat for the failure of the app, saying it was all because the company “won’t change their system”. It’s not like this wasn’t widely known from day one. Hancock signalled his intention to “join forces” with the two tech giants and “bring the best bits of both systems together”. The Apple source told the Times they had not been made aware of Hancock’s new grand plan.
Months later and £8m down, the UK’s contact tracing app may not be ready until winter, or at all. Will the government learn from this? It’s doubtful.