Seite - 575 - in VULNERABLE - The Law, Policy and Ethics of COVID-19

575COVID-19 and Accountable Artificial Intelligence in a Global Context natural language processing and machine learning to gather insights on the emergence and spread of infectious diseases around the globe. Research organizations have used modelling techniques to pre- dict the timing and severity of the crisis and to compare the impacts of different possible intervention scenarios. These epidemiological models typically have a compartmental structure, where the model is divided into four broad categories—susceptibility, exposure, infec- tivity, and recovery (or death). The researchers thus calculate the percentage of the population at each stage, over time. However, the infectivity levels partly depend on the vulnerability of a person, par- ticularly in terms of respiratory weakness. The elderly, but also those with chronic respiratory illness such as asthma, HIV/AIDS, and tuber- culosis, are also more likely to have severe symptoms requiring hospi- talization. As there is no cure (at the time of writing this), the recovery of those who require and receive intensive care is anywhere between 30% and 70%.15 This depends on the quality of care received, as well as on the person’s own immune system. There have been many objectives of impact modelling. Firstly, having a sense of the number of people requiring critical care, as well as the timing of the peak number of cases, has allowed governments to allocate hospital resources appropriately. Secondly, the modelling has allowed the theoretical testing of different intervention strategies, in order to “flatten the curve,” meaning to slow the infection rate and reduce the overall number of cases at the peak. These intervention strategies, in turn, have been implemented, often based solely on the model results. One of the most impactful modelling initiatives in the COVID-19 crisis was spearheaded by Imperial College in London.16 It showed a significant improvement of the crisis by implementing strong mea- sures—physical distancing for the entire population, school closures, and contact tracing. However, there are many countries where the social and economic costs of such interventions in the models threat- ened to exceed the epidemic itself. In these cases, cultural and ethical impact modelling would take these other costs into account by includ- ing appropriate scenarios and population segmentation. An initiative by the London School of Hygiene and Tropical Medicine, for example, 15. Pavan K Bhatraju et al, “Covid-19 in Critically-Ill Patients in the Seattle Area: A Case Series” [2020] New England J Medicine. 16. Robert Verity et al, “Estimates of the Severity of Coronavirus 2019: A Model- Based Analysis” (2020) 20 Lancet 669.