Greece used Eva – AI software – to screen tourist arrivals for COVID-19 in 2020 | Archives, Greece, History & Science


ΑΤΗΕΝS – Reopening Greece to tourists in the summer of 2020, as the COVID-19 pandemic raged, posed a problem for the government: How to test all incoming travelers to see if they are infected is time consuming and difficult.

It is now reported that a Greek-American operations researcher working in data science at the University of Southern California – Kimon Drakopoulos – played a key role in using artificial intelligence (AI) to identify them.

The Nature website reported that a few months after the pandemic, he emailed Prime Minister Kyriakos Mitsotakis and the head of the Greek Advisory Board of Doctors and Scientists to ask if they needed further advice.

They did it – and quickly – and so within a few hours, according to the website, Drakopoulos got his answer. Yes sir.

Greece and most countries had not yet developed the capacity to test all visitors at this point, and there was no real track and trace method on cell phones, and rapid tests and screening procedures were still under construction.

Between August and November 2020, Greece – with contributions from Drakopoulos and his colleagues – introduced an AI system that uses a machine learning algorithm to determine which tourists should be tested for COVID.

The researchers said it was more effective at finding asymptomatic people than random tests or tests based on a traveler’s country of origin. They found that the system detected two to four times more infected travelers than it did when testing it at random, the website said.

You can thank Eva. This is the name of the machine learning system that Greece uses.

However, there have been a few concerns, including accuracy and privacy, during the digital age when spyware spread to phones, hackers can find out your personal information, and government surveillance is suspected.

In many countries, travelers are selected randomly or by risk category for COVID-19 testing, but Eva not only collected travel history, but also demographic data such as age and gender from the Passenger Locator Forms (PLF) required to enter Greece.

The characteristics were compared with data from previously tested passengers and results that were used to estimate a person’s risk of infection. COVID-19 tests were aimed at travelers at the highest risk.

The algorithm also ran tests to fill in data gaps and make sure it stayed up to date as the situation progressed, Nature said.

It wasn’t spread further – and was withdrawn after the trial in November 2020 – due to concerns that governments and companies that make the software are storing people’s data and sharing it with researchers.

“It is also not clear how consent to the use of this personal data can be obtained or how it can be ensured that this data is stored safely and securely,” it says on the website.

Eva was developed in consultation with lawyers who ensured that the program complied with the data protection regulations of the General Data Protection Regulation (GDPR) of the European Union, although it is also used by airlines.


The GDPR requires that security standards must be adhered to and that consent is required in order to store and use the data that would be passed on to an authority, even if the skepticism about the collection and storage of personal data has grown.

These worries also affect the ability to use data collection software and AI, especially in the EU, which is more regulated. The website recommends protocols for data exchange and data protection.

UK tech news website The Register also reported on Eva, saying she was using reinforcement learning, particularly multi-armed bandit algorithms, to identify which potentially infected, asymptomatic passengers are worth testing and, if necessary, quarantining.

It was also able to produce up-to-date statistics on infections that officials can analyze, such as early signs of COVID-19 hotspots emerging overseas, the website said.

Eva was deployed at all 40 entry points in Greece and travelers were asked to fill out a questionnaire indicating the country and region they came from, as well as their age and gender.

Based on these traits, Eva chose whether to test for COVID-19 upon arrival. At its peak, Eva apparently processed about 30,000 to 55,000 forms a day, with each form representing a household, and about 10 to 20 percent of households were tested, the report said.

The software is designed to help high-risk travelers without relying on individual nation test numbers that may fail to report infections, suffer from bias, or lag behind the actual spread of the virus, The Reg said.

Eva would use her own fresh real-time data from people arriving in Greece and try to keep the infected out of the general population in order to contain the pandemic while combing through the asymptomatics.

“First, in light of the latest information, Eva tries to maximize the number of identified infected asymptomatic travelers,” said the American-Greek academic team behind the software in an article in Nature.

“Second, Eva strategically assigns some tests to travelers for whom she currently has no precise estimates in order to better learn their prevalence” as a method of determination.

“There is a very interesting pattern that we observed and reported in our study that shows that an increase in prevalence, which we measure through our system, is followed a few weeks later in the respective countries by an increase in the number of cases”, said Drakopoulos and Vishal Gupta, USC assistant professors told The Reg.

“We had enough resources to test about 10 percent of the arrivals in the peak travel season and 20 percent in the off-season when the arrivals were lower,” they also said.

Eva ended the project because “when the tourist season was over, the number of international passengers arriving became very low and there was very little benefit in allowing non-essential travel into the country,” said Drakopoulos.

They did not disclose how many people were tested after being selected by Eva, citing data protection reasons. “Accessing more data would clearly improve performance, but compromise people’s privacy,” Drakopoulos and Gupta told the website.


Comments are closed.