02【原声】疫情特别节目：Tracking and Tracing

Part 1: Introduction

Brief: The Challenge of Deciding under Uncertainty

00:00/22:52 

Hello listeners of Himalaya. In this episode, we look at how Big Data and AI are crucial tools to monitor and contain the COVID-19 pandemic. It’s all about using data to better know where the virus is and how quickly it spreads.

Do you remember from the previous episode? In the early stage of the global pandemic, public health officials were able to realize the enormity of the challenge and implement draconian measures. They were aided in their decision-making by Big Data and AI that uncovered hidden patterns in data. 

So governments around the world put in place “lock downs” of society, bringing much of the economy as well as social life to a standstill. That was necessary to regain a modicum of control over the pandemic. And it worked – especially compared to countries that reacted too late or thought that they could avoid some of the lock-down measures altogether.

China locked down in early February, followed quickly by South Korea, but Italy was relatively late putting in place measures. It suffered dire consequences, as the virus spread wider and faster in Italy than in many other Asian and European nations. Similarly, Great Britain initially followed a different strategy and thought it could avoid a drastic lock-down, only to see infections skyrocket and overwhelm the health care system. 

Thousands of people died that otherwise would have lived. Only weeks later, the US, Russia and Brazil acted similarly irresponsibly. But the virus can’t be out-bluffed; it can only be beaten with comprehensive and tough measures, informed by data and its analysis.

Without a vaccine, the virus cannot be eradicated; the best that nations can do is to get the level of infections low enough that the health care system – the hospitals and pharmacies, the doctors and nurses – aren’t overwhelmed by demand for care. This strategy is called “flatten-the-curve” and has been implemented throughout the world.

The initial lock-down, however, actually was the easy part. The real challenge is awaiting nations around the worldafter they have weathered the first wave of the pandemic and gotten the situation under control. 

Because eventually, life needs to return; factories need to restart, shops need to reopen – and so do restaurants, and schools, and public transport. We can stop for a bit, but we can’t stop for too long. Otherwise the damage to our economy and our society is so strong that it may even outweigh the cost and consequences of the virus.

This opening requires an immensely careful balancing act. Open up too little, and you damage society; open up too much and the virus may spread out of control. So the core task is to restart the economy and when the number of infections increases too far, quickly reduce public life again.

Any meaningful restarting of the economy and the society requires an enormous amount of difficult decisions. On an individual level, people have to decide how they go about their lives, how the interact with others etc. And public health officials have to decide when to open factories, public transport and schools, and when to clamp down again. 

If in the first phase, one needed to know roughly the ballpark of the pandemic because decisions could be crude, now one needs detail and precision. At the core of get the pandemic under control lies decision-making, and a central tool to aid humans in decision making today is Big Data and AI.

The key idea is that we can track with whom we have come in contact, so that if we are found to be infected, everybody who has been in touch with us over a given period, say the last two weeks, can be informed and take measures to limit any potential spread of the infection.

Tracking human contacts in epidemic outbreaks is not new. In fact, in many nations around the world public health authorities are manually tracking contacts after an infection to limit the spread. This means that public health officials go out to where a new infection happened, interview the infected person, and ask her who she had come in contact with and for how long during the past weeks. 

Then public health officials will get in touch with every one of those that have come in contact. They will call them or make personal visits, ask them about their memory of meeting the infected person, asking them about their health and who they had come in contact with. It is an enormously laborious and time-consuming process.

Even for large and well-funded health care systems, the capacity for manual tracking is very limited. In Germany, for example, authorities can manually track about 200 people at the same time; that is minuscule for a population of over 80 million. 

Consequently, manual tracking worked for example for Singapore when the number of infections were in the dozens but collapsed when infections mushroomed to many hundreds. Manual tracking simply does not scale well.

In contrast, tracing and tracking with digital tools, such as a smartphone app scales well and can be done for an entire nation. It provides timely and useful information to individuals about whether they have come in contact with an infected person or been in the neighborhood with many infections – so that they can make better decisions about whether to get tested or self-isolate. 

And it can provide better aggregate but detailed information to public health authorities. In short, the promise of tracking apps is that with them and with the expertise of Big Data and AI that’s built into them we regain some of the normality of life back and we can restart our economy and our society without losing track of potential infections. It’s our path away from the pandemic until we have a vaccine.

Part 2: How Tracking Apps work

Brief: Tracking Looks Simple, but the Technology Behind it is Substantial

07:38/22:52 

Many of you may have used apps in this pandemic to track cases and see how the situation evolves day by day. That’s very helpful, but have you ever wondered how it works, and what the technology is behind it? And what other nations do?

On a technical level, there are two different approaches to tracking apps. Both approaches rely on apps on smartphones, but then the technical details differ.

The first, the assessment approach, pioneered in China, is based on apps developed by the big digital platform companies Alibaba, Tencent, and Baidu. Users enter their information, then the Department of Health and Disease Control makes an infection risk assessment based on the user’s ID information whether the user is a confirmed case, a suspect case or has come in close contact. Then, data of telecom operators and travel data is used to analyze a users’ travel movements. 

Based on the risk, the user is assigned a color code that informs the user but can also easily be checked and informs public health decision making. It also determines what the person can do and how mobile the person can become. Exact risk settings vary from region to region, but the principle is the same. And I am sure you are all very familiar with how it works on a daily basis.

This approach is strict and in general quite cautious. It was developed in an amazingly short period of time and has helped very substantially to track and contain the virus. But it may make some people uneasy about their privacy.

The second approach, which I call the tracing approach, is pushed in Europe and North America. It tries to address the privacy issue but is technically far more challenging – and has been marred in political controversies. 

The idea is that the app should do everything automatically, and when one person is confirmed to be infected and enters that information in the app, the app will inform everyone who has been in close proximity to that person of the potential infection risk – and perhaps also provide information to public health authorities.

A typical such tracking app has three parts: keeping track of close contacts; informing potentially infected individuals, and – at least in some apps – providing information about the spread of infections to public health authorities. Let’s look at each of these three parts in turn.

First, keeping track of close contacts. For the virus to spread, an infected person needs to be close to another person. It may not be something that individuals realize themselves. Just think about it: You are standing in line for some food or at the post office and the person behind you is too close and if infected might infect you as well. You would not even notice. 

So how would an app do this? To be better than humans, the app will need to be able to measure distance to other humans and the duration of encounters. A proxy could be location information, although that is somewhat imprecise and also requires a lot of data to flow. Easier and simpler is using smartphones connecting via Bluetooth beacons.

We all are familiar with Bluetooth technology. It lets us connect our smartphone to earphones or loudspeakers or other devices; it enables contactless payment and entry into buildings. Normally, the Bluetooth protocol requires that two devices are “paired” before they can exchange information. 

But that is cumbersome. Fortunately, there is another Bluetooth protocol, called beacon, that enables our smartphones to send out ID data like little lighthouses; other devices listen to these transmissions. Importantly, we can measure the signal strength of the Bluetooth beacon and use it as a proxy for proximity – for how close people are. 

Together with timing the encounter, this gives tracking apps an opportunity not only to keep track of who we were close to, but also to estimate infection risk for each encounter using sophisticated Big Data analysis, and perhaps even machine learning.

The second part of a tracking app is to inform everyone who has gotten too close to an infected person. That is important because then individuals can decide to stay at home, or decide to get tested, or take other precautions. It’s valuable data that informs our decisions.

Once a person has discovered to be infected, potentially infected people can be notified. How this notification works – whether it is simply a text message or a public health order to self-quarantine, or even something more restrictive that limits one’s mobility (and thus the chance to spread the virus further) varies from nation to nation. It needs to, because it has to be in tune with different societal values and preferences.

The third part of a tracking app is to provide public health authorities with relevant information, for instance about the number of possible infections and the locality of possible clusters. If the second part of a tracing app does not communicate such information to public health authorities, they will not have this information available. Without such information, public health authorities have far less information to decide whether to ease the lock down or to restrict activities in certain regions again.  

Here, too, different nations have very different approaches. In some countries, for instance in Germany, the government would like to know about potential infections and clusters, but if the app would provide such information, the fear is few people would use the app (or leave their smartphones switched to flight mode or at home). 

So rather than having no app at all, Germany has opted for a very limited app that only provides information to potentially infected individuals, but not public health authorities. In contrast, in nearby France the government has designed an app that does provide public health authorities with such epidemiological data.

Some of the tracking apps also are quite limited in scope and scale because they use imprecise implementations of the Bluetooth technology, or because the app does not run all the time on the smartphone or drains battery so much that smartphones go dead during an average workday. For instance, in Australia the initially app was not very accurate; similarly, in Singapore, the app lacked precision, and so Singapore relied heavily on manual tracking.

Part 3: The Challenge of Digital Sovereignty

Brief: Technological Dependence is Troubling

15:55/22:52 

Tracking close contacts with an app can inform potentially infected people and greatly aid containment. But because this second approach is technically so much more challenging, the early tracking apps that came out in the West in 2020 were all pretty inaccurate and riddled with problems. 

Restrictions in the two dominating smartphone operating systems, Apple’s iOS and Google’s Android, crippled these tracking apps initially – they required a lot of user input and also drained the battery.

In May 2020, Apple and Google announced new versions of their operating systems that addressed the shortcomings and permitted tracking apps to be far more accurate and useful. But at the same time, Apple and Google also announced that they would enforce strict limitations on what the tracking app could do. 

In particular, they would not allow the tracking app to inform public health authorities of potentially infected persons. This was intended to protect the privacy of the users and build trust in the apps, but it also keeps public health authorities in the dark about the potential spread of the pandemic precisely at a time when through the partial reopening of the economy and society, much information about new local outbreaks are needed. It means that public authorities don’t have the relevant information to decide well.

This created controversy especially in Europe. The French government strongly criticized Apple and Google and announced that they would build an alternative app. Germany on the other hand switched away from working with France and joined the Apple and Google camp. 

Great Britain, originally also in the French camp, is reevaluating its position. The different choices in Europe will lead to very different national tracking apps, and potentially limit interoperability across borders – and thus perhaps unduly limit the ability of individual travel. Across the Atlantic, the US has yet to make a definitive decision.

Regardless of the details of the debates, it is quite shocking that nations cannot build the apps they need because the providers of the smartphone systems do not let them. Shouldn’t societies be able to choose what they need? 

In this context, the Chinese choice, initially born out of necessity and the need to act quickly, seems to offer at least independence and include the ability to implement data gathering and data analytics to fight the pandemic based on societal needs and preferences.

In this episode we examined the epidemic assessment and tracking apps that are in use in many nations around the world, how they are built on Big Data and can facilitate AI-based analysis for public health authorities and individuals to make better decisions and improve the chances to manage and conquer the virus.

Just consider how valuable early localized information about potential infections is: it helps local authorities to adjust local rules, but AI and data analytics will also lead to better predictions about how the virus spreads and when: novel patterns in the data are discovered, and new insights gleaned from. 

This will make it easier to differentiate risk, and to ensure that we limit the spread of the infection while also keeping life as normal as possible. With Big Data and AI there is a direct line from collecting data to analyzing it to making better decisions so that we can ensure even in these challenging times that we stay healthy.