By Gordon Rugg
There’s a widespread belief that there’s a useful distinction between “natural” learning and “artificial” learning.
There’s also a widespread belief that Elvis is still alive.
In this article, I’ll explain why the natural/artificial distinction is worse than useless in the context of education, and I’ll describe a more useful, solidly-grounded set of categorisations.
Natural and unnatural skills: The handaxe and the razor
Anyone with an armchair and an opinion can argue forever about definitions of “natural”. The Internet shows that a lot of people have armchairs, and opinions, and a disturbing amount of free time.
If you were trying to work out the meaning of the word “natural” from the surrounding context in which it was used in Internet discussions, you’d probably conclude either that it had something to do with holistic vibrations and healing, or that it had something to do with tradition and a dislike of new-fangled innovations. Either way, it tends to be tangled up very tightly with the writer’s ideology and beliefs, and to mean almost exactly the same as “what I approve of”.
The situation is slightly different with the concept of “natural learning”. Slightly, not hugely. The framing of “natural versus artificial” looks superficially plausible, but the way that arguments about this have been going on for years without much visible progress are a strong hint that starting from here isn’t the best way forward.
I’ll resist the temptation to wade into the swamp of arguments about what “artificial” or “natural” should mean, since they’re a classic example of what goes wrong when people try to force the world into two either/or pigeonholes. The two-pigeonhole approach is usually a recipe for decades of fruitless argument about definitions.
Instead, I’ll look at some other, richer, categorisations, which demonstrate how a different way of looking at the topic can be much more productive.
First, though, a brief disambiguation: The term “artificial learning” is also used in the computational field of Artificial Intelligence, where it refers to computers “learning” for themselves. This is a completely different concept from “artificial learning” in the sense used in this article, namely a term often used to describe formal school-based human learning with human-made artefacts.
Nonverbal versus verbal instruction
A useful, cleanly-delineated way of categorising learning is whether the learning involves verbal instruction or not. Yes, this distinction is binary, but the categorising principle lets you make much clearer distinctions than is the case with “artificial” versus “natural”.
Some learning involves no verbal component; an example is learning by preverbal babies. There’s been a lot of work on what preverbal babies learn, and how they learn it, with fascinating results. It’s a fine, rich literature.
It’s clearly also possible for older children, and adults, to learn non-verbally. If you’re sceptical about this, then try watching some tutorial film clips on topics such as making pottery, with the sound turned off. You can learn a surprising amount just by watching. I’ll return to this issue in a moment.
It’s easy to learn some things non-verbally, but there are other things that are prohibitively difficult to learn without words.
The examples of making flint handaxes and of shaving with a straight razor bring the relevant issues into stark relief. In both cases, you’re dealing with extremely sharp edges in close proximity to major blood vessels. The flakes that you produce when working flint are often sharper than the best straight razors. Both these activities concentrate the mind to a remarkable extent.
So how do people use language in the context of learning these skills? The answer is that they tend to use it for two major purposes.
The first is rationale, i.e. explaining and asking about the reasons for doing something in a particular way. These reasons can be about making the task easier, but they’re often about avoiding a dangerous outcome, such as having the flint shatter in your hand, or nicking a blood vessel when shaving your throat.
The second is teaching heuristics, i.e. useful rules of thumb for structuring what you do. For instance, a useful heuristic about shaving with a straight razor is that the first shave with one is much the worst. This gives the novice a useful set of calibrating expectations, so the novice is much less likely to give up too early.
Since rationales and heuristics often involve outcomes that are rare and/or dangerous, there are a lot of advantages in explaining them verbally, as opposed to showing the learner real examples of something going drastically wrong.
This may be the reason for handaxes being an extremely conservative technology. They were used in much the same form for hundreds of thousands of years, across changes in species. Then, about thirty or forty thousand years ago, modern humans started producing a much broader and faster-changing toolkit. The reasons are hotly debated, but one plausible explanation is that handaxes were produced using only non-verbal learning. This would heavily bias the learners towards close imitation of what they were seeing, without a mechanism for safely investigating what would happen if they diverged from that template.
Implicit and incidental learning
Nonverbal learning isn’t a single homogeneous category. There are various separate types of learning that can occur without a verbal component.
Implicit learning is one type of nonverbal learning that has been extensively studied. In implicit learning, the student learns something by a form of nonverbal induction through seeing examples and being told the category to which each example belongs. (There are other forms of implicit learning, but for brevity I’ll focus on this form.) The classic example is learning how to tell the sex of a newly-hatched chick, by seeing chicks and being told whether each one is male or female.
In implicit learning, the student typically learns the skill without ever being able to put it in words. Typically, this type of learning requires huge numbers of training examples, and a lot of time (in the order of thousands of examples and weeks of learning).
Incidental learning is also nonverbal, but is otherwise very different from implicit learning. Incidental learning occurs, as the name suggests, incidentally, i.e. while the student is focused on something else. A classic example of this method being incorporated into training is when an apprentice is given the job of sweeping the workshop floor, thereby giving the apprentice an opportunity to observe incidentally how the experts work, how they use their equipment, etc.
Incidental learning can occur after just one example – for instance, the apprentice learning to check whether a worked surface is straight by sighting along it. This is very different from implicit learning, which typically requires a lot more resources and time. Also, the student may be able to put their incidental learning into words, even if it is something that they learnt non-verbally – again, very different from implicit learning.
So where does that leave us?
In summary: Arguing about what’s “natural” and what’s “artificial” isn’t very productive.
Looking at the literatures on nonverbal versus verbal learning, or on implicit learning as opposed to incidental learning, or on rationales and heuristics, is a much more productive approach, that gives teachers specific, practical, evidence-based categorisations which they can use to select the best combination of methods for their needs.
That’s a good, positive note on which to end, so I’ll quit while I’m ahead.
I hope you’ve found this useful.
You’re welcome to use Hyde & Rugg copyleft images for any non-commercial purpose, including lectures, provided that you state that they’re copyleft Hyde & Rugg.
There’s more about the theory behind this article in my latest book, Blind Spot, by Gordon Rugg with Joseph D’Agnese
Binary categorisation in general:
Categorisation and gender: