Will Machine Learning Help Health Care and Policymakers Solve the Pandemic?
The explosive rise in COVID-19 cases has produced immense amounts of data, but harnessing the value of this information to improve treatment and deliver better predictive models of the virus’ potential spread has proved challenging. Various groups, however, have begun using artificial intelligence and machine learning to better organize and learn from this data to inform future policymaking and more accurately predict the spread of disease.
Harvard University’s Kennedy School, the Stanford Institute for Human-Centered Artificial Intelligence and the United Nations recently collaborated on a platform to produce a decision-making tool that initially will focus on digital contact tracing of coronavirus infections. The platform, which could be operational by September, will identify secondary and tertiary effects of workforce availability and product/supply shortages across critical infrastructure sectors.
The team, led by Harvard University researchers, proposes to use AI to identify COVID-19 outbreaks before they get out of control. A research paper describes an early-warning system that uses social media posts and internet searches, along with mobility data from smartphones and information from smart thermometers and other sources to inform public policies on reopening and social distancing. When all data sources were applied, the model showed exponential growth roughly 2-3 weeks prior to comparable growth in confirmed COVID-19 cases and 3-4 weeks prior to comparable growth in COVID-19 deaths over the last six months.
The algorithm, the researchers noted, could help guide smoother and safer reopening of states that have enacted shelter-in-place orders and other restrictions on the public to try to stem the spread.
The Harvard-led Collective and Augmented Intelligence Against COVID-19 (CAIAC) will join similar efforts by AI software provider C3 and the Allen Institute for AI’s COVID-19 Open Research Dataset (CORD-19) to organize raw information from statistics, data and scientific journals and analyze it to support better policy decision-making on COVID-19.
CAIAC leaders believe their work will help policymakers worldwide not only in responding to the COVID-19 pandemic, but also when the next major global health crisis occurs.
Meanwhile, an international team of scientists has developed a model that could predict COVID-19 outbreaks two weeks before they occur, giving time to put effective containment measures in place, according to a recent New York Times report.
Despite these innovative efforts, it’s likely to be awhile before we’ll know the effectiveness of these AI-based approaches. And there are various factors that no algorithm could account for in predicting future public behaviors, such as the major nationwide protests and mass gatherings that took place after George Floyd’s killing that may have seeded new outbreaks, despite precautions taken by protesters.
Likewise, it’s too early to know whether such efforts can accurately inform public policy. As some reports have noted, social media and search data aren’t the most reliable predictors and search engines like Google can become less sensitive over time — the more familiar with a pathogen people become, the less they will search with selected keywords.