A reinforcement learning algorithm detects more infected travelers than random border testing.
Since the beginning of the SARS-CoV-2 pandemic, I (and perhaps you, also) have been wondering why on Earth we’ve not been leveraging among the most hyped technologies in recent memory for something other than selling stuff. Those would be Big Data, artificial intelligence and the subfields machine and deep learning.
Finally, it appears, someone has.
An editorial and paper published this week in the journal Nature reveal that a team consisting of data scientists, epidemiologists and public health experts applied reinforcement learning to the problem of Covid-19 border testing. The framework was used by the overextended Greek government to determine when and how to reopen its borders to its economic lifeblood: tourists.
The government lacked the resources to test all travelers for exposure to Covid-19. Moreover, random testing or screening based on a visitor’s country of origin often misses infected travelers who are asymptomatic.
The researchers developed a machine learning algorithm dubbed Eva, billed as an efficient and targeted Covid-19 border testing approach. Previously, the Greek government relied on random testing based on travel history, an inefficient use of scarce resources. Eva went several steps further to parse demographic data gleaned from passenger information forms.
The results? Eva spotted 1.85 times as many infected but asymptomatic travelers as random surveillance testing, including up to four times as many during peak tourist season.
“Our results raise serious concerns on the effectiveness of country-agnostic internationally proposed border control policies that are based on population-level epidemiological metrics,” the researchers reported. “Instead, our work represents a successful example of the potential of reinforcement learning and real-time data for safeguarding public health.”
Part of the reason AI and machine learning have failed to make much of a dent in the pandemic is lack of data stemming in part from stricter data privacy rules. That’s understandable. Much untapped Big Data needed to train machine and deep learning algorithms is held by governments and corporations.
As the editors of Nature noted, data privacy frameworks are needed so relevant data can be shared with researchers.
The creators of Eva were mindful of this requirement, developing their reinforcement learning algorithm in compliance with the EU’s General Data Protection Regulation. GDPR establishes rules for collecting data and obtaining consent to store and use personal information.
That’s a good first step, but data privacy rules must be strengthened further before we can enlist Big Data and AI in the fight against Covid-19. “There must also be a focus on transparency about how algorithms are designed and what data are used to train them,” the Nature editorial correctly notes.
The public health benefits of using trustworthy algorithms to help suppress the pandemic probably outweigh the risks to data privacy. After all, we frequently hit the “Agree” button on “Terms and Conditions” without a second thought when downloading an app.
The Greek government’s machine learning experiment represents a promising first step toward the appropriate application of technologies with the potential to do far more than sell stuff.
This article was originally published on EE Times.
George Leopold has written about science and technology from Washington, D.C., since 1986. Besides EE Times, Leopold’s work has appeared in The New York Times, New Scientist, and other publications. He resides in Reston, Va.