Ideas

The Three Kinds of Tacit Knowledge

Feature image for The Three Kinds of Tacit Knowledge

Note: This is Part 3 in a series about tacit knowledge. Read Part 2 here.

There’s an old quip about taxonomy being the lowest form of science, which has a grain of truth to it — taxonomy isn’t a particularly interesting research activity. But the names we use for things matter if we want to have productive discussions about them.

In the weeks since I started writing my series on tacit knowledge, a number of friends and readers have reached out to me to discuss all the forms of tacit knowledge that they’ve experienced in their lives. These discussions became very confusing very quickly, mostly because the definition of tacit knowledge I used in this series — “knowledge that cannot be captured through words alone” — wasn’t a particularly clear one.

I’ve tried to sidestep the task of defining tacit knowledge in my previous post, where I wrote that I was mostly interested in ‘expert intuition’ as a form of tacit knowledge. I also argued that Commonplace is a practitioner’s blog, and that I was focused on instrumental outcomes. But still the discussions kept coming.

So I’ve given in. This is a short post that summarises the best book on the topic — Harry Collins’s Tacit and Explicit Knowledge, which I read on the recommendation of reader Stephen Samild. (Stephen, if you’re reading this, thank you!) Collins’s book has got a lot going for it — he extends Michael Polanyi’s original definition, examines philosophical, pedagogical, and AI treatments of the subject in the decades since, and then synthesises a set of definitions that I think is quite complete. I wish I’d read it in the beginning instead of Polanyi’s book, but there you go.

This post is short because it isn’t explicitly useful — the goal here is to clarify definitions, after all. But I hope that you will walk away from this piece with a better understanding of ‘tacit knowledge’, and that it’ll result in clearer discussions for you.

Collins’s Categorisation

According to Collins, there are three kinds of tacit knowledge:

  1. Relational tacit knowledge — Knowledge that is tacit because of the way people relate to each other. Sometimes people do not make their knowledge explicit because they do not know how to do so, other times, they do not make their knowledge explicit because they do not want to do so.
  2. Somatic tacit knowledge — Knowledge that is tacit because it has to do with the human body. Think of executing a tennis backhand, playing a guitar, or riding a bicycle. This knowledge is tacit because it has to do with the embodied nature of the skill involved.
  3. Collective tacit knowledge — Knowledge that is tacit because it is embedded in our social environment, and we do not know how to make it explicit in a way that doesn’t involve socialising.

Let’s examine each type in turn.

Relational Tacit Knowledge

Relational tacit knowledge is the tacit knowledge we’ve been talking about in this series. It is ‘relational’ in the sense that it has to do with the relationships between human beings; it is knowledge that is kept tacit because:

  1. People do not want it to be widespread. (e.g. it is transmitted only through selective apprenticeship, or kept within secret societies).
  2. It is incredibly time consuming or costly to explicate. This is essentially the nature of expertise we’ve been talking about in Parts 1 and 2 of this series.
  3. It is held by people who are not capable of explicating it, perhaps because they are bad at teaching. (Woe to the skill whose last practitioner happens to be a horrible teacher!)
  4. Or it could be that practitioners themselves are not aware that certain parts of their knowledge are absolutely critical to their success, and so are unable to articulate what it is that they really do.

In principle, all relational tacit knowledge may be made explicit, with enough time and effort. The key clause in that sentence is ‘enough time and effort’ — for instance, it often takes Naturalistic Decision Making researchers weeks or months before they successfully explicate the nature of the expertise they are studying.

Collins himself has written a fair amount about scientists who were attempting to build a new kind of laser — the transversely excited atmospheric pressure carbon dioxide laser, or TEA laser. He notes, of that pursuit:

My study (…) showed that the scientists failed if they used only the information published in scientific papers. These papers included those which supplied details as intricate as the cross-section and machining instructions for the electrodes and even the manufacturers’ part numbers for bought-in items. It showed, however, that only those who spent some time socially interacting with others who had already built a working model could succeed.

It seems that the act of building a working TEA laser involves all sorts of tiny details that the original creators did not think to explicate. And even if they did make an effort to do so, Collins is not sure that the results would have been any different — some of these things were tiny tweaks that everyone in the lab group understood, but nobody had thought of as important knowledge.

Here's Collins again, on a different group of scientists:

My 2001 study of scientists trying to measure the quality factor, or Q, of sapphire, backed up the earlier work by showing that measurements of the quality factor of small sapphire crystals were so hard that only one group of scientists in the world were able to achieve them until a member of the successful Russian group spent considerable time in the laboratory of a second group, in Glasgow, who finally managed it after a week or so of interaction.

Collins appears to enjoy using the laser example and the sapphire example because the scientists were working from detailed specifications published in peer-reviewed journals … and they were working in physics and material science respectively — both of them ‘hard sciences’! Collins’s point: if it is difficult to explicate knowledge in the ‘hard sciences’, then we should expect no less in other domains of human knowledge.

I look at stories like this and think that a certain amount of relational tacit knowledge will remain tacit forever, because it is simply too troublesome to make it explicit. Why bother writing directions in minute detail, after all, when you can have groups of scientists talking to each other and learning by osmosis instead?

Somatic Tacit Knowledge

Somatic tacit knowledge is Collins’s fancy term for embodied knowledge — that is, things that we learn to do with our bodies. This knowledge is tacit since you can really only learn to ride a bike or throw a curveball with actual physical practice.

The Inner Game of Tennis deals with the nature of somatic tacit knowledge quite well, I think; the author, tennis pro Timothy Gallwey, wrote this beautiful passage on the nature of technique that still haunts me:

Tennis was brought to America from Europe in the late 1800s. There were no professional tennis teachers to teach technique. The best were players who experienced certain feelings in their swings and tried to communicate those feelings to others. In the effort to understand how to use technical knowledge or theory, I believe that it is most important to recognize that, fundamentally, experience precedes technical knowledge. We may read books or articles that present technical instructions before we have ever lifted a racket, but where did these instructions come from? At some point did they not originate in someone’s experience? Either by accident or by intention someone hit a ball in a certain way and it felt good and it worked. Through experimentation, refinements were made and finally settled into a repeatable stroke.

Perhaps in the interest of being able to repeat that way of hitting the ball again or to pass it on to another, the person attempts to describe that stroke in language. But words can only represent actions, ideas and experiences. Language is not the action, and at best can only hint at the subtlety and complexity contained in the stroke. Although the instruction thus conceived can now be stored in the part of the mind that remembers language, it must be acknowledged that remembering the instruction is not the same as remembering the stroke itself. (emphasis mine)

In the same chapter, Gallwey relays the story of teaching a student without explicit instruction: “I was going to skip entirely my usual explanations to beginning players about the proper grip, stroke and footwork for the basic forehand. Instead, I was going to hit ten forehands myself, and I wanted him to watch carefully, not thinking about what I was doing, but simply trying to grasp a visual image of the forehand. He was to repeat the image in his mind several times and then just let his body imitate.”

It worked. Gallwey concluded: “I was beginning to learn what all good pros and students of tennis must learn: that images are better than words, showing better than telling, too much instruction worse than none, and that trying often produces negative results.”

When teaching somatic tacit knowledge, words don’t really help as much as imitation does.

Collective Tacit Knowledge

Collective tacit knowledge was what I was alluding to, last week, when I wrote:

In 1988, Taiichi Ohno published The Toyota Production System, which laid out the ideas behind lean manufacturing for the first time. He did so after resisting codification of those principles for many, many years; Ohno feared that the publication of TPS would destroy Toyota’s competitive advantage. But it turned out that the publication of Ohno’s book by itself did not lead to widespread adoption amongst Toyota’s competitors. More bizarrely, competitors who poached TPS experts away from Toyota were not able to replicate Toyota’s system in their company. It appeared that TPS required an organisational culture to develop alongside the implementation of the system itself. This phenomenon sparked off an entire subfield of research into organisational tacit knowledge.

We learn certain things as a result of socialisation. For instance, we may learn to drive a car domestically, according to the traffic laws of our country, but when we travel abroad — to Vietnam, say, where the traffic behaviours are crazy — we adapt quickly to the norms of the new place. We pick it up on the go; this requires little explicit instruction.

Collins also talks about this in the context of bike riding; he points out that bicycling is somatic tacit knowledge, but when we zoom out, we realise that a good bit of it is collective tacit knowledge as well:

Negotiating traffic is a problem that is different in kind to balancing a bike, because it includes understanding social conventions of traffic management and personal interaction. For example, it involves knowing how to make eye contact with drivers at busy junctions in just the way necessary to assure a safe passage and not to invite an unwanted response. And it involves understanding how differently these conventions will be executed in different locations. For example, bike riding in Amsterdam is a different matter than bike riding in London, or Rome, or New York, or Delhi, or Beijing. (For example, in China, bicycles are ridden at night, without lights, in ways that would be considered absolutely suicidal in the West.) Then again, even in one country there are different settings for riding — riding in the country and riding in the town — mountain biking, riding to school or work, racing, or riding as a display of skill.

Similarly, while the Toyota Production System could not be explicated and transferred between organisations easily, a new employee inducted into Toyota would rapidly internalise what is necessary for the system’s success. We use words like ‘culture’ and ‘organisational behaviour’ to capture this ineffable property; Collins argues that such knowledge is the most tacit of the types of tacit knowledge that exists. He argues that even if we train machines to learn to drive, we cannot expect machines to adapt to driving patterns in the way that humans can when they travel abroad. The human would immediately evaluate the social context they are embedded within; the machine would not.

But my favourite example of collective tacit knowledge is this one, from Star Trek:

In one episode of “Star Trek: The Next Generation,” Dr. Crusher shows Lieutenant Commander Data—a clever android—the steps of a dance. She shows them only once. Most of us would need to practice and practice under Dr. Crusher’s guidance before we could repeat the steps with her skill; for us it would be a case of acquiring somatic tacit knowledge. The scene is amusing, because this is what we are led to expect, yet Data is immediately able to repeat the steps at speed and without fault; he has the kind of quick brain that could also master bike balancing from Polanyi’s instructions. If Data has a somatic limit, it is less restrictive than that of any human.

Dr. Crusher, impressed by the astonishing facility exhibited by Lieutenant Commander Data, then tells him that he must simply repeat what he has done with some additional improvisation if he wants to dance with verve on the ballroom floor. This is where “Star Trek” goes wrong, because it shows Data managing improvisation as flawlessly as he had managed the initial steps. But improvisation is a skill requiring the kind of tacit knowledge that can only be acquired through social embedding in society. Social sensibility is needed to know that one innovative dance step counts as an improvisation while another counts as foolish, dangerous, or ugly, and the difference may be a matter of changing fashions, your dancing partner, and location. There is no reason to believe that Data has this kind of social sensibility.

Collins insist that this kind of knowledge is impossible to explicate. Whether it is really impossible — and therefore out of reach for an AI! — is beyond the scope of this article. Collins thinks that it is; I do not. Nevertheless, it is useful to know that such knowledge exists; we take it for granted because it comes so naturally to us.

Wrapping Up

Collins makes two more observations that I thought were pretty interesting.

  1. In principle, relational tacit knowledge is explicable, even if challenging; somatic tacit knowledge less so, and collective tacit knowledge ‘impossible’. But:
  2. When it comes to human behaviour, the ease with which we pick up the three forms of tacit knowledge is reversed! It is easiest for us to pick up collective tacit knowledge — we begin socialisation from birth; we don’t even notice when we absorb social context from our surroundings. In comparison, it is slightly harder to pick up somatic tacit knowledge, and it is most difficult to pick up relational tacit knowledge, since this hangs on the nature of expertise, and it the depends on the ways with which we relate to each other.

In practice, many things that we do involve all three types of tacit knowledge — car driving, for instance, involves relational tacit knowledge, since it is a skill that must be acquired from others; it involves somatic tacit knowledge because it is something that we do with our bodies, and finally it involves collective tacit knowledge, since we pick up on norms whenever we are driving, wherever we are in the world.

Read Part 4 — How to Use YouTube to Learn Tacit Knowledge.

Originally published