Chris Dalton's Personal Development Blog

Inference /ˈɪnf(ə)r(ə)ns/ n.  a conclusion reached on the basis of evidence and reasoning

Blake’s engraving of Chaucer’s Canterbury Tales – metaphors everywhere!

A lot of inference goes on in management education, but I wonder how much of it is rigorous or even does any good. Here are a few thoughts on this topic.

1. Inductive reasoning

Imagine that you are given a map of an inland territory, part of a larger land mass (maybe a continent) and asked continue drawing the map. If you only use what is already there, you would have to extend by extrapolation to continue the drawing outward. Each time you drew something, sure, you could go and check, and any new data could become part of the further drawing. This is inductive – your test of accuracy is in the form of further observation (trial and error). If you were incapable of learning from this, you would be restricted to the same simple protocols of feedback (that is, you could make corrections to your map but not to your method of map making). Does this mode of inference equal learning how to learn? Not much. Arguably, only in respect to the meta-level skill of getting better at a process of trial and error (i.e. if placed in a different situation that required blind trial and error, your years of map-making like this might have resulted in an increase in speed of your trial and error method). This is often what happens to managers as they acquire skills during their careers.

2. Deductive reasoning

Given the same starting point of having to draw a map from a fragment, you notice that on the partial map there is a river. If you know that water always flows downhill (knowing this doesn’t restrict you to inductive reasoning for map making, though you might not use this knowledge), you could perhaps predict and then draw the likely course of the river into your new map on the basis of other extrapolations. In other words, you draw through a mental process of “if…, then…”.

The accuracy of your map (your prediction, or “then”) is still subject to verification by observation, but is now based on application of a covering rule – it contains a test of a hypothesis. Similarly, you may apply other rules, such as that rivers flow into other rivers and eventually into the sea, and this thought also becomes part of an imaginary map to be tested against experiment. The application of a premise established earlier in time (a priori) and independently of experience is deductive inference. Of course, there is a possibility (sometimes an aim) that a hypothesis is not matched by data. Assuming you can trust the data (i.e. your senses), you now have a route to amendment of either the hypothesis, or of the covering rule that generated the hypothesis. Is this learning to learn? The same argument could be made as for learning in inductive inference, that a you get better at applying deductive hypotheses in other contexts and this is a sort of learning how to learn. Inductive and deductive practices often go hand in hand with in practice and the boundary between them is actually only arbitrary. In either case, learning is correction of error in terms of a specific response from within a given set of alternatives (context learning), and not a correction of error in terms of change in set of choices.

The problem for deductive reasoning is that the ‘then’ is only as sound as the premises informing the ‘if’ premise itself. The best ‘ifs’ are those that express fundamental principles, but this is far from easy in managerial situations. The device, or reasoning, that can make the leap from date to theory is called abductive (sometimes retroductive).

3. Abductive reasoning

Ok, you are equipped with a keen eye for observation and a decent education and you are given the map-making task. How do you proceed? Yes, you could dive in (as many managers think they ought to) and set about solving the problem you’ve been set (whether by trial and error, or by prediction based on rules you have been taught), but neither of these will lead you to higher level learning. Induction is clear and simple and will suffice for a bit. Deduction is clever and structured, and will work as long as the premise holds up. But neither one produces anything novel. Neither is creative. And neither one leads to deeper understanding of the world as it actually is. For that, you need also abduction.

Abduction is deliberately taking the explanation for one set of phenomena and asserting this also as explanation for the data you have. It’s a more complex and artful form of reasoning as it contains a leap, and crucially demands an understanding of a deeper pattern (i.e. a knowledge of what connects otherwise disparate forms).

Pattern.

This may all sound a bit woolly, but for many great scientists, abduction is how they explore new ground. It is guessing, but educated, informed guessing is a good thing. For example, Albert Einstein and Richard Feynman each used such informed guesswork to bridge gaps from data to theory.

I like to characterise abduction as use of the logic of metaphor, and few things have the ability to drive the imagination, or kick-start the generative process of creation, than the bringing together of two unalike things in order to see how they are alike.