In the Eye of the Storm
If we are indeed in the eye of the COVID-19 storm, let’s use the data at hand to chart the course we wish to sail once we have more clement skies.
I sit nestled behind my desk in the French Alps, looking through the window at the growing green grass, bright yellow Forsythias, and budding cherry trees that signal the first premises of Spring. Although most of our close friends and colleagues are either well or mildly infected by COVID-19, I find my spirits dampened daily by the storm of bad news raining down on social media. If we are indeed in the eye of the storm, let’s use the data at hand to chart the course we wish to sail once we have more clement skies.
The need to focus on data-driven decision making has never been more pressing in light of the Internet’s constant amplification of unsupported opinions, self-serving contributions, and collective anxiety. Our customers' data science projects are on standby, our marketing funnel is running on fumes, and I’m left to ponder whether the sunshine tapping on the window sill is a premonitory omen of even darker times ahead. What is the data saying, and how can we frame what we can and cannot learn from the data spread before our eyes?
A quick review of the literature reveals several stories of how organizations are enlisting Data Science to address the current crisis. Three short months ago the BlueDot AI platform flagged a cluster of “unusual pneumonia” cases in in Wuhan, China. Baidu developed an AI-based detection system that incorporates computer vision and infrared sensors to screen up to 200 people per minute for potentially infected cases. Alibaba has introduced an AI system that can identify the virus in chest CT scans 45 times faster than the human eye.[i] Google’s DeepMind AI research lab is employing deep learning to find new information about protein folding associated with COVID-19. [ii] Kaggle has teamed together with several leading AI institutes to release the COVID-19 Open Research Dataset Challenge to address ten baseline questions developed by NASEM/WHO on the current crisis.[iii]
As promising as these examples may well be, four fundamentals steps of the analytical method can help us better understand what the data is saying, and where the figures offer little more than sounds of silence. How does the context help us understand the roots of the problem, how can we quality the data at hand, what is the appropriate methodology to address the problem, and how can we transform the data into collective action? Data Science isn’t about crunching the numbers, but understanding how perceptions, predictions, evaluations can be used to create actionable insights from data.
Dig into the current context to reveal the roots of the problem we are trying to solve
Is the problem created by the outbreak of this new strain of coronavirus basically one of medical, economic or social well-being? Outside of our data pipelines, these problems inter-related, each contributes to the severity of the others. We need to explore why certain populations more at risk than others, why is there a shortage of tests and medical facilities, and why are our economic and social systems having trouble adapting to this challenge.
Once we’ve identified the nature of the problem under study, we can look for datasets to plot a course of personal or collective action. We can choose to abide by, or ignore, public policy and recommendations, or we use the data to try to nudge our families and communities to act for the common good. In either case the decision environments are stochastic rather than deterministic, the “answer” can’t be found in the labelled data at our disposal. The best we can do is enlist Data Science to provide probabilistic frameworks hat can reduce the risk, ambiguity and uncertainty of the decision-making process.
Qualify, rather than quantify, the data at hand
How pertinent are the abundant statistics on the global spread of Covid-19 provided by WHO, national governments and local authorities? Has the data on infections, hospitalizations, and morbidity been collected using criteria that allow us to draw meaningful conclusions from one country to the next? Given the differences in containment, testing and treatment strategies across continents, what data are we missing about the attitudes, behaviors, and perceptions of the crisis that need to be taken into account?
As both Kenneth Fields[iv] and Amanda Makulec[v] have pointed out, the form in which the data is presented is also a fundamental concern here. In communicating the variables, data sets and graphs, we influence the public’s perceptions of the problem. The issues cut to the heart of data visualization: each presentation of the data is conditioned by its context, representation techniques, use of metadata, and implicit narrative. Because datasets are never objective, we need to understand the extent to which the numbers elucidate the problem we are trying to address, how the data reflects how the problem is evolving, and whether we have enough information to predict or influence the outcomes of our action/inaction.
Choose a methodology that appropriately addresses the problem
The advantages of data-driven decision making here cannot be underestimated, despite the tendency of some of the world’s leaders to wish this crisis away. In Data Science we learn that the choice of appropriate methodology depends the nature of the decision environment, the type of data at hand, the number of available parameters, resource constraints, and the required accuracy, precision and recall. Yet accounting for trade knowledge is vital when addressing the Covid-19 pandemic for we cannot discount the diagnostic and therapeutic concerns with the symptoms, pathologies, reaction times, and protocols.
This need for practical wisdom demonstrates the limits of Data Science, the breadth of human intelligence cannot be reduced to simple algorithms no more than human decision-makers can be depicted as cold-blooded calculating machines. Dealing with the crisis will require ethical choices of socially acceptable courses of action: Who is ultimately responsible for choosing the appropriate course of action: the individual, the medical profession, or the government? To what extent can individual liberties be sacrificed in installing a surveillance state? Should we be sacrificing the many to protect the privileged few in saving our national economies? Is isolation, whether it be on a national or community level, preferable to that of engaging a global response to the pandemic? The appropriate methodology to address the crisis must weigh momentary considerations of financial value with the perennial need of human values.
Transform the data into collective action
If there is no direct link between data and action, we must explore how heuristics shape the perceptions, predictions, and evaluations that lead decision-makers or act upon the data. On an individual level, why do people react so differently to the wealth of data being published each day? If the Internet has contributed to the instantaneous and global circulation of information, why are there so many differences in our national perceptions of the severity of the crisis? Why have Singapore and China acted to adopt systematic testing, where Great Britain and France prefer only to acknowledge those that are manifestly ill?
How can so many countries opt to enforce national quarantines when the Netherlands has chosen to do the opposite and the United States is considering sending people back to work? Why are the politicians stage justify their decisions “in a state of war” pleading that communities bind together against the "common enemy of humanity”? How can we use cognitive biases and the behavioral sciences to encourage organizations and individuals to adopt socially acceptable behavior? Today more than ever, the value of our contribution isn’t in analyzing the data on the screen, but in feeding meaningful stories capable of inciting change.
Many have written before on the nature of decisive moments – fleeting instances in time in which complex social and economic systems can significantly evolve for better or for worse. We are living one of those moments, where change in our medical, economic and social systems in indeed possible. As Yuval Noah Harari commented, we can’t avoid sailing into a storm, but we have the choice of both how we weather it out and what course we will choose to sail once we have more clement skies.[vi] One advantage of sitting in the eye of the storm is our illusion of free time - which if used to look at the data, can help us navigate more wisely into the future.
Lee Schlenker is the Principal in the BAI http://baieurope.com and Professor of Business Analytics and Community Management. His LinkedIn profile can be viewed at www.linkedin.com/in/leeschlenker. You can follow the BAI on Twitter at https://twitter.com/DSign4Analytics
- Originally published in Towards Data Science
[i] Dickson, B. (2020), Why AI might be the most effective weapon we have to fight COVID-19
[ii] DeepMind (2020), Computational predictions of protein structures associated with COVID-19
[iii] Kaggle (2020), COVID19 Global Forecasting
[iv] Field, K. (2020), Mapping coronavirus, responsibly
[v] Makulec, A. (2020), Ten Considerations Before You Create Another Chart About COVID-19,
[vi] Harari, N.Y. (2020), The world after coronavirus
