[I’m trying to resolve confusions I have brought on when I realize that people forget most of what they learn by default. Yet we still seem to get better at things. I bring up a model of passively learning certain things and that people who are successful at certain fields are using better learning strategies. Probably still wrong and confused, but <feels good to write a new blog post>.]
I’ve noticed myself confused about how to think about growth. I have two conflicting lines of thought: (1) Most of what we learn is not retained, and yet, (2) across many domains, we seem to be able to build off prior knowledge and advance anyway. What’s going on here?
Well, first off, our default methods for learning are sub-optimal. As I’ve pointed out before, there is a pernicious illusion of understanding that can creep up on you. Reading a textbook can feel like you’re “studying”, but unless you’re actually testing yourself along the way, you’ll be wasting your time. Similarly, coming up with a math proof is far more difficult than verifying that is correct, but, internally, they can feel like the same thing. (See Conceptual Similarity Does Not Imply Actionable Similarity and Recognizing vs Generating for more examples.) Furthermore, it seems like our ability recall decays exponentially, meaning that, in a short amount of time, we fail to remember much of what we learn.
Anecdotally, it’s often the case that people freely admit they remember little of what they learned in the past. People often say things like, “Ah, it’s been a while…I don’t really remember” and this is a socially acceptable excuse for not being able to recall many details. I’ve had difficulty reviewing things like notes I’ve taken in different classes. I can remember that there was a time when the notes made more sense, but they don’t make sense now. The memory of fluency is there, but not the fluency itself.
This illusion of understanding paired with the decay seems to imply that most of what we read proves to be largely inaccessible in the long run, let alone actionable. If you believe strongly in having what we learn influence what we do, this is worrying. This means the quality your information diet would seem to have little effect on your long-term outcomes.
There are once again considerations involving fading novelty here. For example, though re-reading a book might better help solidify its ideas in your head, if you’ve (perhaps mistakenly) marked something as “already read”, then you’ll be driven to other, newer / shinier options.
What all this seems to point to is that the cards are stacked against us when we’re trying to learn anything at all. So what’s actually going on when people do in fact learn things and improve?
One thing is that we do have a good idea of what it takes to improve our recall. Spaced repetition, deliberate practice, and retrieval practice are all well-documented interventions which outperform our defaults when it comes to building more explicit knowledge. So perhaps people who see a lot of growth of this type are those who’ve managed to discover these strategies and use them?
It does seem like there are generalized principles for learning well which stay roughly the same across different disciplines. For example, ensuring that your practice environment is very similar to your performance environment, e.g. how a magician wants to practice their tricks in front of spectators, rather than just in front of a mirror or how a driver wants to practice driving on the actual road, rather than just in the parking lot.
That seems to condemn everyone else, though, to a fate of never learning much at all. Yet, most people seem to be able to improve at some things, simply given time.
Something we often observe is the Pareto Principle—most of the impact can often be traced back to just a few of the contributors, rather than being more equally spread amongst all of the contributors. So perhaps it’s okay that we forget most of what we learn if it turns out that the small amounts we do remember end up being responsible for our growth. Taken at face value, though, this seems too optimistic. Anyone who’s kicked themselves for forgetting a crucial formula on a test knows how unreliable our memory can be, if left to its own devices.
But what does seem interesting to tug at is the notion of experience. We seem to develop better models and expectations for how things work by virtue of just interacting with them for a while. This is largely a passive process that doesn’t look like the intentional process I described earlier for maximizing our learning. The sort of benefits afforded to us by experience feel more implicit than the explicit knowledge I had in mind when I talked about growth in the passages above. It’s the difference between “I’ve had similar experiences 100 times before, so I think the result will look like this” and “I’ve got a model with gears that explains why the result will look like this”.
So perhaps the original way I was thinking of growth was mistaken. Maybe explicit knowledge doesn’t capture most of how people grow and there’s this other class of implicit knowledge— things we know but can’t clearly recall or explicate. And perhaps implicit knowledge can explain how we grow.
If this is the case, then a question to investigate is “For a given field, to what extent does success appear to come from implicit vs explicit knowledge?”
I think Implicit knowledge is important for “performance” areas like sports, martial arts, sales, and public speaking. These are situations where there is some given setting with a set of standard rules, the goal does not vary much, and you’re interacting with other humans. You don’t have time in a performance to pause and consult detailed references. Doing so would break the flow of the performance. Thus, experience, with its non-verbal models, is what you rely on.
But that’s not everything. There are also areas where explicit knowledge plays a bigger role. Much has been written about “mathematical intuition” and the insights that seem to come only from experience, but I think this misses the obvious—at its core, you still need to be able to have a strong command of how concepts link to one another in order to make progress. And the manipulations used in proofs seem like explicit models, requiring recall. These types of manipulations and models seem to span most sciences, where what you need models of is something that’s rather alien-looking, compared to the sorts of phenomena our brains have been dealing with for most of our evolutionary history. Models in the sciences have precision or a sense of discreteness that is often absent from implicit models; it is difficult to port our gut feelings into this framework.
From this, a potential answer emerges: implicit knowledge can be thought of as being effective at modeling humans, relationships, conflicts, and other mainstays across most of human society. Explicit knowledge is effective in domains where our intuitions and non-verbal models fail us. This shows up when looking at things we have not had to model for most of human history (EX: gears-like models of nature) and requires a different type of knowledge.
Thus, we should expect to see more passive growth in domains where experience can provide better models over time of the environment and less passive growth in domains where experience is a mistaken guide.
More speculatively, perhaps this is why people seem to think of themselves as not having a “knack” or “talent” for the sciences? If our default strategies for learning lend themselves less easily to the optima when it comes to gaining explicit knowledge, their suspicions might not be far off. Rather than a knack, though, perhaps what they lack is a good set of strategies for improving?