“Simple and complicated contexts assume an ordered universe, where cause-and-effect relationships are perceptible, and right answers can be determined based on the facts. Complex and chaotic contexts are unordered, and there is no immediately apparent relationship between cause and effect, and the way forward is determined based on emerging patterns.”
-David Snowden, in Harvard Business Review’s A Leader’s Framework for Decision Making (in describing the Cynefin Framework)
This past Saturday, amidst the COVID-19 pandemic and widespread protests for the unjust death of George Floyd, the U.S. launched two astronauts into space. The vehicle was the SpaceX Falcon rocket with its Crew Dragon capsule and this event opened a new era of space travel. This journey, with Elon Musk as the privileged vanguard, is symbolic of the future with a partnership between SpaceX and NASA and implementation of AI as an essential part of its sciences. A close colleague posited about artificial intelligence: even if we launched SpaceX with an incredulous return of its booster back to Earth (including practice returns on a drone ship at sea), maybe AI is still not as good as we think if we cannot seem to make even reasonable predictions about the course of the COVID-19 pandemic.
I think the observation can be reconciled in understanding the difference between “complicated” and “complex”, both of these contexts reside between “simple” and “chaos” in the systems science’s order continuum; these terms, however, are often used interchangeably and thus create an understandable confusion. A complicated problem can usually be clearly defined and be deconstructed to its component parts (system equal to the sum of its parts), and each part can be analyzed (and improved) as part of the whole system (therefore “reductionistic”). The process is ordered and linear with an attainable equilibrium and the outcome is usually predictable or deterministic. Examples of complicated systems include: cars, planes, and cardiac pacemakers. A complex problem, on the other hand, is less clearly defined (therefore ‘fuzzy”) and cannot be easily deconstructed meaningfully into separate component parts as each part is interdependent with the other parts (therefore “holistic” with dynamic relationships), Here, the process is unordered and nonlinear without an attainable equilibrium and the outcome is highly unpredictable or stochastic. Examples of complex systems include the immune system, the stock market, and the heart. While the SpaceX launch and booster return is duly “complicated” as an engineering system with many decisions and parts, the inclement weather that preempted its original launch is a “complex” organic phenomenon with highly unpredictable elements.
It should not be a surprise that predictions are often inaccurate for this “unordered” pandemic with its myriad of nuances as it entails a mutable biological infectious agent as well as complex human physiology and crowd behavior. All of these aforementioned elements, therefore, render the pandemic a collection of very complex phenomena that is exceedingly difficult to predict (even with sophisticated modeling and deep learning) and would require continual real-time adjustments based on new raw data as well as human expertise and wisdom to synthesize predictions. Perhaps the role of data science and AI is ensuring that the pandemic does not become an utterly chaotic process with extreme uncertainty and disagreement, and attempt to maintain the pandemic as a “simply” complex phenomenon.