The Covid Diaries are written by Xinova innovators from around the world to share expert insights that advance understanding of the global pandemic in all its sprawling complexity, and come up with ideas to solve it one problem at a time. In this article, Richard Hartung shows how a Singapore start-up re-applied its AI technology to track Covid-19.
As Covid-19 spread rapidly around the world in the past few months, Singapore-headquartered artificial intelligence (AI) start-up SQREEM looked at how it could use its expertise in finding customers for financial services companies and other multinationals to help governments reduce the risk of Covid-19. It quickly developed an innovative solution that is being used in South Africa and is being considered by other governments as well.
Adapting an AI Marketing Solution to Covid-19
Until quite recently, SQREEM had focused on using cognitive AI algorithms to enable banks, insurance companies, automobile manufacturers and other clients to understand and predict human behaviour. These clients use its insights to understand customer journeys, map behavioural intent to purchase, identify hundreds of thousands of potential customers, and target their media placement
As the number of Covid-19 cases started climbing this year, SQREEM rapidly pivoted and adapted that same methodology of its AI for use in contact tracing for Covid-19. It has fused layers of data so that it can identify, reach and intercept individuals who are or may be at risk of becoming infected. SQREEM’s global data coverage, AI, media infrastructure and insights into consumer behaviours enable it to deliver messages directly to individuals, clusters of individuals or people who have been in a location of interest.
“It starts with the person who is infected,” SQREEM CEO Ian Chapman-Banks explained. “The government gives us their address or device ID. We put a bubble on the address.”
SQREEM then looks backwards for about two weeks to identify people who have been in contact with the infected person. SQREEM uses data from apps on the infected person’s phone, which “ping” people at least once a day and sometimes 50 or more times per hour, to trace the routes the infected person has taken. It then merges social, web visit, GPS and behavioural data – all anonymised – to produce a timeline showing places where the infected person spent at least 30 minutes.
Even then, it is not possible to track a journey minute-by-minute, and there are gaps.
“What we do,” Chapman-Banks said, “is we use our AI engine’s behavioural data to fill in the gaps. Once we triangulate where they work and live, and places they go, the behavioural data kicks in.”
SQREEM can trace the journeys they made with at least 93 percent accuracy and then matches those journeys with other devices to identify people the infected person interacted with along the way.
SQREEM uses the vast amounts of anonymised social data from the apps of other people who were near the infected person to call them or send messages via WhatsApp, Facebook or SMS. The messages tell the persons that they were potentially in contact with somebody who has Covid-19 and ask them to exercise caution or call a Covid-19 healthcare phone number, depending on what the government prefers. If the infected person went to Starbucks for coffee, for example, SQREEM can send a message to anyone who was in the Starbucks shop around the same time to let them know they may have come into contact with a person with Covid-19.
In one case in South Africa, for instance, the government identified a woman who was infected. By the time she was identified, her husband had gone off to work in a mine in another part of the country. SQREEM tracked the husband to the specific mine, used device IDs to find nearly a thousand people he had been in contact with, and sent each of those thousand people a message letting them know they were at risk.
Along with using “backward tracking” to identify people who interacted with the infected person as well as situational concerns and social convergence, SQREEM can use “forward tracking” to forecast health concerns and newly emerging trends.
The advantage for a government is that it can trace and contact huge numbers of people more effectively. By way of example, Chapman-Banks said that if there are 1,500 cases a day and each of those infected persons has been in contact with 100 people, there would be more than 150,000 calls to make to contact everyone. Realistically, he observed, “you can’t call 150,000 people.
While identifying the people at risk is clearly a huge computational task, Chapman-Banks said SQREEM can identify millions or even hundreds of millions of people and has no limitations on the number of messages it can send. “Every app on the phone sells the data, anonymised, at a marketplace,” he explained. “The GPS location is up for sale, traded on an open exchange. We buy all the GPS signals for a country. We load them to a map. We’re one of the biggest ad exchanges in the world.”
And while it might seem like analysing so much data would run into concerns around privacy and personal data protection laws, no government data is needed and SQREEM only uses anonymised data that is already available.
Along with supporting the government in South Africa, SQREEM has been in contact with governments in countries in Asia Pacific and Europe to explain how the solution works. If any of them decides to use the solution, it would be adapted for their specific requirements. Whereas authorities in some countries instruct people to go to a specific healthcare centre to be tested, for example, those in other countries ask people to self-quarantine at home for 14 days and call a healthcare provider only if they develop symptoms.
“It’s a nice solution,’ Chapman-Banks said. “It works.”
And amidst challenges in a multitude of countries that are having difficulties tracking potential infections, it may well save thousands of lives.
Bio: Richard Hartung has over 20 years of experience in the payments and consumer financial services industry with extensive experience in the Asia Pacific region. In May, 2002 Richard founded Transcarta, which focuses on assisting financial services companies with strategy, payments training programs, operations process enhancement, merchant acquiring, market entry research and other business practices. He is also a freelance writer for Today, gtnews, Challenge, The Asian Banker and OOSKAnews. Richard has a BA from Pomona College and an MBA from Stanford University. He is active in community organizations, including on the boards of the Metropolitan YMCA and the Jane Goodall Institute (Singapore).
311 ½ Occidental Ave S
Seattle, WA 98104
+1 (206) 800-2678
Xinova Japan GK
Yaesu Mitsui Building 6F
2-7-2 Yaesu, Chuo-ku
Tokyo 104-0028 Japan
Xinova Innovation (Beijing) Consultancy Co. Ltd.
Room 2494, Level 24, Tower 3
China Central Place
77 Jianguo Road
Beijing, 100025 China
Xinova Asia Pte Ltd.
331 North Bridge Road
#22-01, Odeon Towers
+65 6408 5080
10th floor, Golfzone Tower
Seoul 06072, Korea
+82 2 6952 8840
Xinova Australasia Pty Ltd.
223 Liverpool St
Darlinghurst, NSW 2010
+61 2 8093 4850
Suite No 208, 2nd Floor
Survey No 414
Govinda Reddy (GR) Layout
Karnataka, 560076, India
Erottajankatu 5 A 4
Affiliate offices in Tel Aviv & Vienna