Coronavirus: How Mathematical Optimization Can Help Organizations Overcome Today’s Unprecedented Challenges

Author: Juan-Carlos Mani
Date: 4/21/2020

We tend to overuse the word “unprecedented,” but, when describing the coronavirus pandemic that is currently engulfing the world, “unprecedented” is perhaps the only word that is accurate.

Everything about this coronavirus pandemic is unprecedented: the scope and severity of the health crisis, the speed and scale of the economic downturn, the upheaval and uncertainty in our financial markets, and the disruption in our global supply chains.

Governments and businesses around the world are reeling from the fallout from these unprecedented challenges and changes, and scrambling to find ways to contain the spread of the virus and cope with its devastating economic impact. Some countries are now starting to cautiously reopen their respective economies (which is a much more complex task than shutting down these economies), and are trying to determine how far this easing of restrictions should go and how fast in order to avoid a resurgence of infections.

What is the best way for organizations today to navigate the uncharted waters in which they find themselves? How can they move forward when their historical data, their plans, and their predictions have all been essentially rendered meaningless?

The answers to these questions are not simple, but one thing is for sure: To react and respond efficiently and effectively to the unprecedented disruption caused by the coronavirus pandemic, companies and businesses today need to leverage the right technological tools. And, without a doubt, mathematical optimization is one of these.

Harnessing the Power of Deductive AI

On a fundamental level, we can divide AI technologies into two different categories: those that are underpinned by inductive reasoning and those that are underpinned by deductive reasoning.

The majority of AI technologies out there (most notably machine learning) employ an inductive reasoning approach. This means that – by using data to train and shape an underlying mathematical model – these technologies are able to find patterns and trends, generate predictions, and infer probabilistic solutions to problems.

The quality of these “inductive AI” solutions depends on the accuracy, timeliness, and completeness of the data (rather than on the design of the underlying mathematical model).

• Although machine learning and other inductive AI solutions are critical tools for organizations today, they do have some significant shortcomings as:
  Users do not have a causal understanding of what is happening in the system – they only can attain an inferred understanding and answers based on data.
• The underlying system and model have been trained and built on data, but when the conditions in the real world and the characteristics the system change dramatically, the answers provided by these solutions are not reliable.

And this second point is precisely what is happening now with the coronavirus pandemic: No inductive AI system has ever seen such an unprecedented disruption, and thus the underlying model and system (which are based on data) are not capable of generating reliable and robust solutions to the problems of today.

In contrast, “deductive AI” technologies utilize a deductive reasoning approach. This means that – by building an underlying system and mathematical model that is a digital twin of the real world – these technologies are able to generate deterministic, accurate, and optimal solutions to real-world problems.

Mathematical optimization is a perfect example of a powerful deductive AI technology, as it:

• Captures the key features of real-world business problems (including their decision variables, business rules or constraints, and business objectives) as mathematical models. When real-world conditions change (as they invariably do on a day-to-day basis), users can easily adjust and adapt the underlying mathematical model accordingly by adding new decision variables and constraints that reflect these real-world changes.
• Utilizes a mathematical optimization solver, which reads the mathematical models, processes all available data, and rapidly generates globally optimal solutions.

So, even in the face of the unprecedented disruption caused by the coronavirus pandemic, mathematical optimization technologies have the flexibility and robustness to handle all the sudden and significant changes in the real-world business landscape – and still deliver optimal solutions. Companies and governments can use these solutions as the basis to make optimal decisions and drive optimal business outcomes – empowering them to react and respond to change and disruption in the most efficient manner possible.

Mathematical optimization technologies provide organizations with a system that models and encompasses the unprecedented business conditions and challenges in our world today, and they can rely on this system to consistently generate optimal solutions – no matter how profoundly the surrounding business landscape continues to change.

Navigating Through These Unprecedented Times

I do not mean to suggest that mathematical optimization is the panacea for all the unprecedented challenges that the world is facing today. To cope with the healthcare and economic crises that we are experiencing, we will need a full arsenal of cutting-edge technological tools – and mathematical optimization is definitely one of those tools.

Mathematical optimization gives businesses and governments the power to capture their real-world business problems – from supply chain network design to resource allocation, production and logistics planning, strategic investment, energy distribution, and many, many more – as mathematical models and then find the best possible solutions to those business problems.

I also do not mean to suggest that mathematical optimization should be used in isolation. On the contrary, some of the most powerful AI applications are those that combine different deductive and inductive AI technologies, like mathematical optimization and machine learning. Mathematical optimization can and should be used to augment other AI technologies – especially in times of volatility and disruption.

To overcome today’s unprecedented challenges, we will need to leverage all of our most powerful and state-of-the-art technologies and also tap into our inherent ingenuity, resourcefulness, and resilience as human beings.