In the early days of the COVID pandemic, the UK’s Centre for Data Ethics and Innovation (CDEI) set up a repository to capture AI-driven innovations deployed in fighting the virus. The CDEI’s recent review of the repository concluded that: “aside from driving forward advances in vaccine research, artificial intelligence did not play the outsized role many thought it would in relief efforts.”

The CDEI report identified the following trends

1. Conventional data analysis has been at the heart of the COVID-19 response, not AI: Data has been front and centre in the COVID response: we all hung on the news of the case numbers every day. While AI is data driven, AI and machine learning take-up was minimal. CDEI speculated that with COVID being such a new phenomenon, the data required to train algorithms did not yet exist.
2. Existing datasets provided the basis for much of the pandemic response: In the urgency of the response, health organisations first looked inwards and many found that augmenting and repurposing existing data for the specific COVID-19 circumstances could provide the basis for an effective response. For example, Google, Apple, and Facebook have been publishing "mobility reports" containing aggregated location data drawn from existing databases.
3. New frontiers of data sharing emerged: the main technology-related story of the pandemic seems to be the tumbling of previous entrenched barriers to data sharing. Public services opened up datasets to the private sector for the first time: for example, by giving supermarkets access to information about vulnerable patients most in need of assistance. Independent providers developed a stronger sense of working within a system, such as children’s day care centres pooling data to identify system wide trends. The CDEI noted that ‘while seemingly straightforward to administer, these data sharing initiatives required new legal agreements, oversight measures, technical standards and data storage tools.’
4. Community data sharing increased: In the face of the public health crisis, we have a renewed sense of community and empathy for those hardest hit by the virus. This generated new community data sharing initiatives where individuals and organisations have shared data for the benefit of the wider population. For example, the COVID Symptom Study is an online collaboration between King’s College London, Guy’s and St Thomas’ Hospitals, and ZOE Global Limited with over 4 million contributors globally.
5. Where AI was used, it was in the primary health care sector: facing the risk of being overwhelmed, general practices and hospitals had powerful incentives to harness all available technologies, including those still nascent. Oxford University Hospitals built an AI-driven test which was able to predict whether a patient had COVID within an hour of their arrival at an emergency department.
6. Many existing tools have been repurposed to solve COVID-19 related problems: many existing technologies and AI applications in non-health sectors were pivoted to mitigate the health effects of the pandemic. UK hospitals used gaming mixed reality headset to minimise face-to-face contact with patients who have symptoms of coronavirus whilst ensuring they receive immediate access to specialist opinion.
7. Data-driven tools are also being used to measure and understand the effects of new rules: a number of companies have created wearables for automating social distance control in the workplace. A British company Tharsus has designed an original system, Bump.

Despite this disappointing outing for AI, the CDEI thinks that, as the focus is beginning to shift towards building future resilience, that AI will play a greater role:

• The British Heart Foundation uses AI to measure increased acute and longer-term cardiovascular risk in COVID patients. The project uses novel AI techniques applied to CT chest scans to accurately measure the level of inflammation in the patient's’ arteries, which is suspected to be a cause of severe responses to COVID-19. These results will be combined with previous CT scans in many patients with existing heart disease and repeat CT scans after infection has subsided, allowing a direct comparison of inflammation before, during and after infection to understand whether COVID has lasting effects on heart health.
• Intellegens has received funding from Innovate UK to model COVID data and improve the management of future coronavirus outbreaks and other infectious diseases. The tool, which will be based on Intellegens’ deep-learning algorithm, will help understand how policy changes might impact outcomes, helping inform decision-making.

It’s all about trust (or lack of)

The CDEI’s research found that almost three-quarters (72%) of the UK population believe that digital technology has the potential to be used in response to the COVID outbreak. Indeed, 42% think that the potential of technology to make things better in the UK is not being fully realised.

The main reason given by the public as to why digital technology might not be effectively used in the COVID response was a concern that the technology would not be used properly, rather than problems with the technology itself. While a reasonable proportion of the public (43%) trust that the right rules and regulations are in place, there is a hard core of 24% who disagree. This is largely consistent across age, region and gender.

The CDEI concludes that “to realise a sustainable increase in the use of new technologies, the analysis suggests the critical importance of building and maintaining trustworthy governance.”

Read more - COVID-19 repository and public attitudes retrospective

Authors: Anna Belgiorno-Nettis and Pater Waters