Using Data and AI To Manage a COVID-19 Response
Historically, the approach to mitigating virus spread has been contact tracing and this is where AI can play a role.
Black swan events like the COVID-19 pandemic have the potential to stress public health infrastructures and civic systems to the point of collapse. Because the virus primarily spreads via direct, and even indirect, human contact the potential scale for the spread, as we are seeing, is significant.
Two techniques can be applied to curb the spread of a virus:
• Prevention/Cure: Medical solutions driven by vaccines (prevention), and therapeutics (cure).
• Mitigation: Arresting spread through human behaviour, such as practicing hygiene and social isolation/distancing.
Due to the absence of prevention and cure methods, mitigation is the primary way the pandemic is currently being managed.
Historically the approach to mitigating virus spread has been “contact tracing” – tracking the movements of infected people to identify who came in contact with them, based on interviews and other mostly manual methods. With the speed and scale of this pandemic, it is hard, if not, impossible to adequately map the spread in this way.
A more effective option is to use data and artificial intelligence (AI). AI powers two foundational tasks of pandemic mitigation: (a) infection tracking and (b) infection spread prediction, by enabling three foundational goals.
• Measuring social isolation by observing individual mobility
• Identifying clusters of more than a certain number of individuals and identifying the corresponding locations
• Risk assessment of individuals and locations, at scale, by understanding the movement of infected individuals
It turns out data and AI are being widely used to battle COVID-19, particularly in countries that adopt a scientific approach to public health. Data scientists are creating machine learning models to predict infection and mortality rates which drive resource need and allocation planning.
The key to the successful use of AI relies on the data attributes, known as “features”, that are fed into the models – If this data is inaccurate or lacks scale, the ability of the model to predict outcomes will be poor.
At Mobilewalla, we work with major healthcare companies, hospital systems, and the US Government, who play significant roles in pandemic management. Using sophisticated AI and machine learning techniques, we create the feature artifacts that our partners leverage in their spread prediction models.
India is also attempting to use technology to manage the COVID-19 crisis – a prominent example being the Aarogya Setu app. A worthy effort, it could serve as a useful consumer tool to receive current COVID-19 information, as well as minimise risky behaviour.
However, it is important to understand that the app by itself is simply a front end to information delivery – its effectiveness is only as good as the information fed through it.
It is vitally important to understand that the app itself is not producing that information — the data the app uses is coming from a variety of external systems that have been put in place to monitor and track the infection. It is not clear to me that these systems are producing reliable information at scale.
Given that a key purpose of the app is to provide information that allows individuals to manage their own risk of exposure, the worst that can happen is the app disseminating data that instils a false sense of security in users. If inadequate information is fed in, this possibility cannot be ruled out.
While it is impossible to ascertain the quality or the scale of the information fed into the app, any user should be asking the following questions
- What is the scale of adoption of the app? For the app to be useful, the network of users has to be substantial
- Are users truthful in their disclosures? The app requires certain key disclosures from users, such as their infection status. In India, where social stigma still plays a key part in social interaction, one might question the likelihood of truthful disclosures at scale. Faulty disclosures will result in misleading estimations of risk
Knowing the GIGO (garbage in garbage out) principle applies to any information portal, the creators of the app need to take special care that high-quality information, at scale, is its foundation. Countries need to be taking a holistic approach to fighting the spread of COVID-19.
Social isolation increased testing to identify infected individuals and tracking the spread through contact with infected individuals is where countries are currently focusing their efforts. As this identifying and tracking need to be done at a mass scale, AI can be a powerful tool.
(Anindya Datta is CEO & Chairman of Mobilewalla, a global player in consumer intelligence solutions. This is a contributory piece and the views expressed are the author’s own. The Quint neither endorses nor is responsible for them.)
Subscribe To Our Daily Newsletter And Get News Delivered Straight To Your Inbox.