Epistemic systems

One topic that often comes up when we meet family is that of children – education and career. And the discussion morphs into passion and purpose. And something that everyone of us desires of ourselves and our children – what if education and careers truly align with passion and purpose. The ikigai infographic was doing the rounds not so long ago.

Which got me thinking: why is it so that our economic and social systems require us to make career defining choices by the late teens and make an initial career commitment by early 20s? Of course there are outliers – most of us have heard inspiring stories of people who have successfully shifted careers at different stages in their lives. And we all know that these are outliers. It is apparent that the ‘systems’ have evolved to a state of equilibrium that expect individuals to start generating economic value from their early 20s onwards. However, we economies grow and become richer, societies can afford the luxury of allowing the incoming generations to delay their entry into the workforce, and in the process, explore the alignment of passion and purpose with vocation. No wonder ikigai has been trending of late.

All this leads us to an interesting observation: this is an example of emergent systems that seek their equilibrium over time. In the absence of a central authority, the systems continuously learn through feedback loops until they reach a dynamic equilibrium. Understanding and modeling these systems is an interesting area of study. And for the longest time, humans have attempted to model and predict these kind of systemic behavior – there are two broad methods: aleatory and epistemic.

Aleatory systems assume that you can define a system completely and then quantify the uncertainty of random events that affect the outcome. Simplest example: throw a fair 6-sided dice and estimate the probability of a 6. Some argue that this is too simplistic.

Epistemic systems attempt to study the inherent complexity of any system and try to model that as well. These would start by trying to model the dice itself – how many sides does it have? And that in turn, drives the probability of throwing a 6.

When we make a choice between aleatory and epistemic systems – we are trading-off the modeling complexity for a better understanding of the system. However all too often, we forget this simplification and fall into the trap of Reduction Fallacy.

And so, the next time we seek to understand and explain some of the conventional wisdom around us, it will do us good to appreciate that the complex systems have evolved to attain their states of dynamic equilibrium. And avoid falling into the trap of taking an aleatory view – at the very least, be aware of the trade-offs at play.

That begs another question: what if at all, does this mean in an Enterprise context? We tend to assume that enterprises are not self-emergent systems but directed structures, defined by a common goal and governed by rules. And most of us would agree that the modern corporation in today’s marketplace is anything but that – the complexity is only accelerating all the time. It is even more exciting that as the complexity increases, we have the data, methods and systems in place to better tackle this complexity – those who can truly exploit  Data and Decision Sciences to better understand and model epistemic systems will be able to extract the real value from Data.

I will end this with a simple example: forecasting has traditionally been a time-series problem (a classic aleatory system). Epistemic uncertainly (e.g. a supply chain shock due to an unpredictable event)  are treated as external events and not factored into forecast accuracy. The thinking is that this very hard, if not impossible, to factor these uncertainties. And that often creates a mental block – imagine the value that can be created if you have an early warning system that can sense Anomalies and develop scenarios to mitigate them.

Nassim Taleb made a fortune doing exactly that – perhaps we should think along these lines as well.

Side note: there is a relatively obscure debate going on between Taleb and Nate Silver on what a forecast ought to model: current uncertainty (Silver) or future uncertainty (Taleb). This is more just an academic debate – forecasting (Sales, Supply Chain), predicting outcomes of events (customer churn) all need to think along these lines.