A preprint paper coauthored by researchers at Microsoft, the Indian Institute of Technology, and TCS Research (the R&D division of Tata Consultancy Services) describes an AI framework designed to help cities and regions make policy decisions about lockdowns, closures, and physical distancing in response to pandemics like COVID-19. They claim that because it learns policies automatically as a function of disease parameters like infectiousness, gestation period, duration of symptoms, probability of death, population density, and movement propensity, it’s superior to the modeling tools that have so far been used.
If the peer-review process bears out the researchers’ claims, the framework could be useful to organizations and governments in the nearly 200 countries with cases of the coronavirus. Asian nations including Singapore and Taiwan have demonstrated that containment strategies like contact tracing — the process of identifying people who may have come into contact with an infected person — can effectively mitigate the spread of disease.
The coauthors first generated a graph network — a model containing objects in which some pairs are related, where the objects correspond to vertices and each pair of vertices is called an edge — with 100 nodes and 1,000 individuals. Each node stood in for a city or a region containing a certain number of individuals, and the strength of the connections between pairs of nodes was directly proportional to the product of the population between nodes and inversely proportional to the square root of the distance between them.
Next, the researchers modeled the best disease parameters as available for COVID-19: an incubation period of 5-10 days, an infected period of 7-14 days, an 80% likelihood of showing visible symptoms, a 2% death rate, and a 100% transmission probability for infected persons who come into contact with susceptible persons. Multiple simulations were run to obtain reliable statistics.
Throughout the study, the researchers assumed that an open node allowed people to travel to and from other open nodes in the network. People showing symptoms weren’t allowed to travel to other nodes but asymptomatic and exposed people could do so. (When a node was locked down, all travel to and from the node was blocked.) Additionally, they accounted for the fact that while symptomatic people were quarantined within nodes, a small number of people broke quarantine and circulated within.
The researchers also established several baseline lockdown policies, in which they assumed each node had the option to be locked down or opened once per week. They then defined a set of policies that locked down any given node if the fraction of symptomatic people in that node crossed a predefined threshold of 5%, 10%, 20%, 50%, or over 100%.
Lastly, the team trained a Deep Q Network reinforcement learning algorithm (an algorithm that spurred on software agents via a reward) that made a per-node binary decision each week — “open” or “lockdown” — by running a number of simulations of the spread of the disease. To have the algorithm identify the optimal policy for lockdowns, they quantified the cost of each outcome of the simulation: A weight of 1.0 was placed on each day of lockdown and each person infected; a weight of 2.5 was placed on each death; and the reward was defined as the negative of those costs so that higher rewards corresponded to lower costs.
In experiments, over the course of 75 simulations with simulations lasting 52 weeks (364 days), the researchers determined that policies with 5% to 10% lockdowns experienced a lower peak of infections. Predictably, the policy was wary of decisions contributing to an increase in the fraction of symptomatic people within the same node and the population overall, and so it locked down larger nodes earlier once the infection started spreading and nodes where the potential for outside infection was higher as soon as infection began spreading within the node.
The coauthors caution that none of the authors are experts on communicable diseases and that the AI model in the study doesn’t account for population size and geography, and that they didn’t use real data for the network model. But they say that a deeper analysis is in progress and that they’ll continue to add more detailed descriptions and literature review in stages.
Beyond this study, various teams are developing AI systems to track the spread of COVID-19. Carnegie Mellon researchers are in the process of retraining an algorithm to predict the seasonal flu, while the Robert Koch Institute in Berlin used a model that takes into account containment measures by governments, such as lockdowns, quarantines, and social distancing prescriptions to show that containment measures can be successful in reducing the spread. Elsewhere, startup Metabiota offers an epidemic tracker and a near-term forecasting model of disease spread, which they use to make predictions.