Critique of Article: Why Tacit Knowledge is More Important Than Deliberate Practice

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.

A reader recently shared this article with me and asked for my thoughts: Why Tacit Knowledge is More Important Than Deliberate Practice on Commoncog.com.

I have two thoughts. First, I’ve previously been asked about one of the referenced articles “The Problems with Deliberate Practice” and I think that article itself has problems, which I’ve addressed in a critique here.

Second, and more importantly, my take on tacit/implicit knowledge (edit: versus deliberate practice up an explicit skill hierarchy) is that – while I do believe it exists and becomes important to focus on as one gets into a very advanced level in a field – I don’t think it’s a productive focus for beginners (i.e., students). The visual I have in mind is a skill tree where the trunk is narrow and consists of comprehensive foundational skills that are easy to enumerate explicitly, but as you get up really high in the tree and go out all the way out to the leaves, there’s a combinatorial explosion of skill combos/contexts and the branches kind of “fade away” in the sense that it’s really hard to explicitly enumerate all the individual skills to be learned. (It’s really a DAG, not a tree, but I’m just going to continue calling it a tree here because it helps visualize the situation.)

We have come face-to-face with this problem while thinking about how to develop courses for SAT math and competition math. You can enumerate lots of the foundational skills and get students up to a decent level of performance, but as you get into the really advanced problems that appear later in the test, the combinatorial explosion creates a lot of friction and eventually you have to transition away from atomic topics and towards a dispersed “ether” of knowledge that you absorb by grinding a large volume of problems LeetCode-style (edit: though you can and should still create atomic prerequisite topics out of any categories of problems where solutions have similar-ish structure, because the more tacit/implicit knowledge you make explicit, the more students will be able to learn it who would otherwise not catch on). We are working on building this additional component into our system for our test prep courses.

The transition has to happen eventually, but I think many educators/education resources screw it up in the following ways:

• Failure Mode 1: Forcing the transition too early. A single teacher might give up on enumerating foundational skills because it becomes tedious and they've reached a point that's "good enough" to be acceptable to parents and admins, but a team of content developers who are in it for the long haul (not just the next semester) and being held accountable and compensated on the basis of the actual content development work they do (not just "taught a semester-long class that was deemed acceptable quality by parents and admins") can go much further in a coordinated effort.
• Failure Mode 2: Having students grind advanced problems without having the foundational skills in place. Again, for a single teacher with multiple classrooms of students each day, it's impossible to manually track every student's knowledge profile to a high degree of precision, figure out what problems are appropriate for them, deliver a different specially-chosen problem to each student, and provide immediate feedback to each student -- but this can be done by sufficiently skilled software developers who were former teachers/coaches, are being compensated for solving the problem in an absolute sense (not just "good enough for parents and admins"), and are in it for the long haul.
• Failure Mode 3: Not tracking the student's knowledge profile well enough in the dispersed "ether" of knowledge and not calibrating the difficulty of problems well enough to the student's level. Same issue as above: too much work for a single teacher to do manually to a high degree of precision, but it IS possible to automate the work if you have sufficient skill, compensation, passion, and are in it for a long time horizon.

In particular, I think many people underestimate the amount of tacit knowledge that is obtained by layering more advanced explicit skills on top of previously learned skills. There’s a saying that “calculus is where students learn algebra” that I think speaks to the amount of tacit algebraic knowledge that is obtained by going through calculus. Calculus forces you to grapple with algebra in progressively more complicated contexts and this forces a deeper level of understanding.

Sometimes people will argue that after learning algebra, students shouldn’t move on to calculus and should instead expand “sideways” and do enrichment projects to get a deep level of algebraic understanding. I think that’s completely false and inefficient. The most efficient way about this is to climb as far as you can up the tree mastering explicit foundational skills, move fast and absorb tons of tacit knowledge surrounding lower-level skills as you continue layering more advanced skills on top of them that pull them together in various ways.

Eventually you reach the end of this foundational skill tree (not a hard limit of what’s possible to encode as explicit knowledge, but the limit of how far whoever built the tree was willing to build it out). And then you transition away from atomic topics and towards a dispersed “ether” of knowledge that you absorb by grinding a large volume of less-structured practice experiences, because you’ve run out of road and that’s the only way you can continue forward.

(Edit: Likewise, many people underestimate how many domains have foundational skill trees. Sometimes the known foundational skill tree is quite tall and you can climb it for a while before you exhaust it. Other times the known foundational skill tree is relatively short and you end up having to switch over to tacit knowledge acquisition a lot sooner. But there’s still a skill tree to be climbed.

For instance, full immersion in a language eventually becomes helpful to reach a high level of fluency, but unless you are a child learning your first language and you have no choice but to fully immerse, you wouldn’t do that from the get-go on day one, right? It’s more efficient to learn foundational words and grammar rules in a structured manner first before making the leap into an immersion experience that pulls together all those prerequisites. Otherwise, your learning efficiency tanks in the immersion experience due to overwhelm. It’s an instance of the expertise reversal effect.

Of course, structured learning can be frustrating and legitimately less efficient for outlier students in a class of more typical students who need to move much slower. But if you de-tether the structured learning experience from the class and instead tailor it to each student’s individual pace, then the structured approach becomes a very efficient way to acquire the foundational knowledge you need to keep up your learning efficiency in the immersion setting and avoid getting overwhelmed.)

Edit: Some Final Thoughts

Tacit knowledge acquisition is interesting and I’m glad that Commoncog is writing about it, but I don’t think it’s necessary to shoot down DP in the process. They strike me as two parts of a single pipeline (climb foundational skill tree –> switch to tacit knowledge acquisition).

I tend to focus my writing on climbing the skill tree because, in my experience, that seems to be where most people get stuck in developing expertise. I continually see people failing in their endeavors due to missing skills that could be developed efficiently by climbing a known skill tree. I also see lots of room for improvement in known skill tree offerings, so that more people are able to climb them and are able to climb faster. Basically, I see a ton of additional juice to squeeze in the skill tree portion, so I’m focused on squeezing it.

But I’m sure there is plenty of room for improvement in the tacit knowledge acquisition pipeline as well, and I’m happy to see juice squeezed from that as well.

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.