The EconomistMarch 28th 2020 BriefingThe pandemic and the state 172
1kcdcsays that system was too slow, and it
has now automated the request process, al-
lowing contract tracers to pull data in auto-
matically through a “smart city” dash-
board. This data-request system was put
into operation on March 16th. Korean news
reports say that the automation has re-
duced contact-tracing time from 24 hours
to ten minutes
It might also be possible to do some-
thing similar from the bottom up, thus lim-
iting government snooping. Start with an
app that sends coherent health and travel
data to a central registry, as China’s Health
Check purports to. Then add sufficiently
smart and powerful number-crunching for
the system to be able to find all the places
where two people’s histories cross. When
someone gets sick, the system can then
alert all the other users whose paths that
user crossed. Because the infrastructure
would be separate from that of the spooks,
it could be much more open, scrutable and
trustworthy.
Such approaches, though, face serious
problems. The number of people an infec-
tious person actually infects will almost al-
ways be much smaller than the number
they encounter. Sean McDonald, an expert
on public health and digital governance,
says a system which alerted all the people
that an infected person had been near over
the past week could lead to a demand for
tests that would entirely overwhelm the ca-
pacity available in most countries. If the
relative risk of, say, walking past someone
on the street and drinking from the same
water fountain an hour apart were known,
and if the data picked up such niceties,
things might be different. But they are not.
An alternative to too much testing
would be not enough. Annie Sparrow, an
epidemiologist who advises Tedros Adha-
nom Ghebreyesus, director-general of the
who, points out that modellers without
field experience tend to misunderstand the
psychology of testing. The stigma associat-
ed with a disease, she says, is likely to out-
weigh the rational pull of keeping oneself
and one’s family safe. And both Dr Sparrow
and Mr McDonald point out that any sol-
ution which relies on smartphones and in-
ternet access inherently ignores the half ofthe planet which does not have internet ac-
cess. Mr McDonald says he would prefer to
see the data wizards apply themselves to
easier problems such as optimising the
supply chains for medical goods like masks
and ventilators.Big brother is contact-tracing you
Google says that, having heard epidemiolo-
gists make such points, it is not planning to
use the location data it collects to do con-
tact tracing. The data-collection mecha-
nisms built into products like Android or
Google Maps are “not designed to provide
robust or high-confidence records for
medical purposes and the data cannot be
adapted to this goal”, the company says.
Facebook says something similar. Both
companies can be assumed to think that
talking explicitly about how well they
might be able to do such things would raise
concerns about privacy.
What Google and Facebook will not do,
though, the government of Singapore is
quite up for. Its Government Technology
Agency and health ministry have designed
an app which can retrospectively identify
close-ish contacts of people who comedown with covid-19.
When two users of this new app, called
TraceTogether, are within two metres of
each other their phones get in touch via
Bluetooth. If the propinquity lasts for 30
minutes both phones record the encounter
in an encrypted memory cache. When
someone with the app is diagnosed with
the virus, or identified as part of a cluster,
the health ministry instructs them to emp-
ty their cache to the contact-tracers, who
decrypt it and inform the other party. It is
especially useful for contacts between peo-
ple who do not know each other, such as
fellow travellers on a bus, or theatre-goers.
The app’s developers have tried to as-
suage concerns about privacy and security.
Downloading it is not compulsory. Phone
numbers are stored on a secure server, and
are not revealed to other users. Geolocation
data are not collected (though Google’s
rules governing apps that use Bluetooth
mean that they will be stored on Android
phones running the app). They are plan-
ning to publish the app’s source code and
make it free to reuse, so that others may
capitalise on their work.
Singaporeans trust their government.
Since TraceTogether was released on March
20th it has been downloaded by 735,
people, or 13% of the population, according
to government data. Several Singaporeans
your correspondent spoke to one overcast
day in the business district were unaware
that they could be prosecuted for refusing
to hand over their data to the health minis-
try. But they had no intention of frustrating
the authorities. “I’d rather be responsible
than irresponsible,” said one trader.
In an attempt to get past the uproar
about the security services tracking the in-
fected, Israel’s Health Ministry has
launched a similar app that allows people
choosing to use it to see if they have come
into contact with other users who subse-
quently took ill. The government says that
the app, which uses open-source software,
does not share data with the authorities.
The whoMyHealth app, also open-source,
might in time take on a similar contact-
tracing function.
This patchwork of global systems pre-
sents its own challenge: how to make themWays forward
Data tools for the covid-19 pandemicSource:The EconomistCivil Where it’s
Application Purpose Data source liberties risk happening
Quarantine Knowing people are where they should be GPS data sent from bracelet or phone Medium Hong Kong , Taiwan,
enforcement Singapore, China
Contact tracing Knowing whose paths have crossed Top down: Government takes data from platforms High Singapore, S. Korea
Bottom up: Phones provide data to each other Low Worldwide
Flow modelling Knowing how many people pass through Mobile-phone-tower data Low Google, US,
places, and how quickly probably more
Social-graph making Knowing which people tend to meet repeatedly Mobile-phone-tower data with machine learning High Nowhere known