Simulation: From Impossible to Probable
More data and more computational power have finally triggered the breakthrough of artificial intelligence (AI) in analytics. AI methods have evolved from three different domains including (see for example, McKinsey’s “An executive’s guide to AI,” link): Statistics (e.g., K-means clustering as a popular statistics-based unsupervised machine learning algorithm), operations research or OR (e.g., support vector machines or SVMs) and bio-computational science.
Mimicking nature: From neural networks or deep learning … to simulation
Bio-computational methods loosely mimic biological phenomena. Neural networks are an example of such AI. They are now called “deep learning” and were originally derived from our understanding of how the human brain works (MIT, link). One powerful application of computational AI in general is in simulation. In his now famous and highly influential classic of modern AI, “The Science of the Artificial” (link), Herb Simon, a 1978 Nobel laureate (link) and 1975 Turing Award winner (link), lays out a foundation for systematically harnessing the power of information processing to design and engineer better performing systems.
Decomposition, hierarchy, and adaption
Simon points to a few “tricks” for decomposing the complexity of systems to improve our understanding of them. They include thinking in hierarchies, building blocks, and adaptive interactions. A system can be decomposed into a hierarchy of “chunks” or “building blocks” of simple behaviors that interact and adapt in a complex environment. Simon uses the human body as an example: It is a very complex system with many elements functioning simultaneously, yet it can be decomposed hierarchically “into organs, organs into tissues, tissues into cells” (p. 186).
80/20 efficiency of simulation-based computational explanations
Very often, it is impossible to test entire systems. It may be too dangerous, expensive, time-consuming … even illegal. With simulation it is possible to build an abstraction of the truth that is focused on only those key “building blocks” and interactions that are relevant to the investigation. If it is done scientifically and rigorously – think of clinical trials for new medication, such as vaccines – results will be highly probable. In such, simulation embodies the Parento principle, a business efficiency measure, which states that, for many events, roughly 80 percent of the effects come from 20 percent of the causes.
Read about how we put Simon’s tricks to work to create true business advantage in: Schlueter Langdon, C. 2020. Simulators in Business: Testing the impossible, discovering the probable. Working Paper (WP_DCL-Drucker-CGU_2020-05), Drucker Customer Lab, Drucker School of Management, Claremont Graduate University, Claremont, CA