Analyst’s Pendulum (Reasoning)
A Concept That Is Clearly Not Elementary, My Dear Watson
Logic and analytics are fairly synonymous. Not completely, but close enough that bringing in aspects of reasoning is not a very big stretch. That said, little is quite as confusing to young analysts. At least that is what I am inferring…
Deductive vs Inductive Reasoning
Before we get too far, let’s shoot the elephant. You may be thinking, wait that is an hourglass, not a pendulum. They are in fact synonymous in the nomenclature of time keeping. This makes the title of The Hourglass Pendulum by Arthur Conan Doyle a bit redundant.
It is also a nice transition to the more confusing title of its main character Sherlock Holmes. He is known as the master of the “Science of Deduction”. Meanwhile, he is more accurately described(or at least more typically engaged) as the master of the “Science of Induction”. It doesn’t help that deduce and infer are considered synonyms as well. Why is this so confusing?
This infographic does an excellent job of laying out the differences of our “top down” vs “bottoms up” approaches. They even put Sherlock on the right side. Unfortunately, Aristotle covered both approaches, so why he was chosen for the left hand side is not so clear. Incidentally, Google this type of infographic and you will find examples with Holmes on the left as well…
I would also add that analysts, unlike many other scientists work largely to disprove (not confirm) theories. But I digress…
Deduction starts from the top, with clear rules. Those rules are then applied to some observation to predict an outcome. Induction begins at the bottom with a set of observations. A set of rules is then inferred from those observations. Deduction flows from theory. Induction produces it.
Deduction is predicated on truth. Rules (hypotheses) are either true or false. Induction is a built on probability. Rules are either likely or not. These probabilities are quantified using theorems like Bayes. Invariably, the situation flips (much like our hourglass). Rules generated via induction becomes rules applied through deduction.
Don’t get confused by questions of whether you are deducing or inferring. In the English language at least, these terms don’t actually define any real distinction. Some people believe they imply it, but that is a stretch. If you are ever asked if your statements are inferred, be aware that the individual asking is likely well out of their depths. Almost all analytics is inferred.
Honestly, almost all analytics falls in the category of Abductive Reasoning. For readers of this series, you will remember there are no absolutes (or at least no more than one) and the world is rarely binary. In other words, the existence of a third is hardly surprising. But then it also isn’t very unique either.
Abductive Reasoning is just Inductive Reasoning with a little more ambiguity. It is still bottoms-up, observation-based, and probability-driven. Why someone felt the need to create a whole new label… well that is another subject entirely.
Reasoning is a natural component of analytics. Understanding the terminology for the processes you use isn’t critical… or this would never have been so hard. On the other hand, other disciplines cannot always say the same. When you see references to deductive and inductive reasoning in other sciences you will have a better sense of how they map to the common processes in analytics. It is not so much a swing of the pendulum as recognizing a pendulum by another name.
Thanks for reading!
Start at the begining: