On Learning, Part 2
How and Why We Learn
Part 1, found here, sketches a general “theory of learning”. Here we’ll hold human nature up to this picture, so as to draw attention to a number of features which would otherwise be hard to recognize as attempts at, or failures of, the human learning process.
VII. Learning To Be Safe
How do humans learn? And how well does human learning map onto the general picture given in part 1?
I've already observed that human learning ranges from "active" to "passive" depending on the skill in question. Most skills will start off as being actively-learned, while later becoming passive once a good-enough "model" has been synthesized. Think of a baby learning to walk, or learning to drive. By the time you're an adult you will hardly think about your motor skills, or how to drive, speak, socialize, or navigate the world—learning does happen, but at this point will act much more like the "gradient descent" of AI.1
Active-learning reappears when we slow down and pay attention to something: when learning a new activity, say, or relearning some motor skills after an injury. "Attention", then, can roughly be said to determine where our learning is happening.
Active learning can also be brought about by surprise. Here I think of driving: when another driver does something unpredictable, your idle thoughts will suddenly vanish, and fear will pull your attention to the road. Time will seem to slow down. At the risk of treating the mind too much like a computer, we can see the fear as sending a kind of "interrupt signal" which grabs hold of all of the mind's "processors".
The takeover by fear surely comes at a cost, at the least because your current train-of-thought will be dropped and probably will be hard to reconstruct. But your instincts have already made the cost-benefit calculation: Maslow’s hierarchy-of-needs has been applied and physical safety has won out. The fear is an anticipation of pain, which in turn predicts physical harm, and this is to be avoided at (almost) all costs—the short-term is prioritized over the long-term.
Emotional pain works roughly same way: other peoples’ anger and judgment cause pain in us in the form of “shame”, which aims to direct our attention to the perceived offense out of fear of being stigmatized. We’ll usually perform a quick calculation to see who is in the right, and will back off if it’s not us—unless we feel disrespected in which case the need to preserve status will supersede.
Pain, then, exists to to assign urgency to a short-term learning process. It says: figure it out NOW—and act. The underlying ethic at work in our human instincts can therefore be summed up very succinctly: "safety first".
Pain is therefore the "error" term in human learning—the information which tells us what to avoid.2 In the short term we react instinctively; in the long-term we backpropagate our pain and fears against our internal actions and beliefs, “invalidating” our mental models, and demanding that we synthesize something better so as to reattain safety.
VIII. Breakdowns of Learning
It follows that the mind would continuously circulate the memories of painful experiences into consciousness, where they can be taken on by an active learning process. From the inside this would feel like being "anxious" about a threat; here a “racing-thought” flavor of anxiety has appeared as a rather predictable failure-mode of learning. Again this is fear prioritizing learning—and again its purpose is to figure out what to do: either to work out what went wrong, to attend to what you weren't paying attention to, to take note of what has changed in the world, or, if nothing else, to accept the existence of the threat as out of your control. The reaching of one of these conclusions is what “model synthesis” looks like in a human mind. And these alternatives will turn out to be very general: a fear will relax when it is able to transmuted into either action, attention, belief update, or acceptance.
If the mind, faced with a problem for which model-synthesis cannot converge, falls back to circulating the problem as a racing thought, you might think the right thing to do in this case would be to “accept” one’s unsafety and move on. If this isn’t happening it’s because the mind believes at some level that it doesn’t actually have to accept it. It’s holding out for model-synthesis, but something else is barring the way; whatever outcome would handle the fear is contradicted by some other part of the person's own beliefs. Here I’m thinking of the sorts of knots of anxieties and maladaptive beliefs around things like “body shame”. Other examples often involve an ideal which one is not ready to let go of, for example, it may be hard to give up on a dream of being a writer because of a deep attachment to the virtues of great writing.
When the obstacle preventing the learning process from completing is some other belief, it can be very hard to figure out what's going on. We humans are rather bad at "admitting what we actually believe"—most of the time we'll say what we think we're supposed to believe, or recite what we believed in the past. These, though, are no help at solving a problem. What is needed is a higher-order learning which reconciles the need for safety with whatever other belief is in contradiction (itself probably a need for safety in some other sense)—the two fears have to "brought together" and a synthesis has to be "solved for". That this is so hard to do can be seen as a direct consequence of the “memory” constraint in human learning. If we could bring all of our beliefs into “working memory” at the same time, we wouldn’t have these problems.
The survive/thrive trade-off leads to a similar failure-mode. A person who feels threatened will generally adopt some low-order "heuristic" model to ensure their short-term safety while continuing to work on a long-term solution in the background. Here we can identify "avoidance", "triggers" and "being touchy about subjects" on our map of human behavior, as well as trauma and PTSD, in a sense.3 But it is typical, too, for these low-order reactions to be fairly costly themselves. e.g. overreacting to some small offense will tend to embarrass you in some other way. And again, as we humans are not very good at all at acknowledging what we actually believe, we’ll often tend to treat the new injury as a new problem to be solved—but then you have two problems—then three—and a person can easily get in a state where they’re continuously hurting themselves in a feedback loop, invalidating every compensating action their mind can come up with.
Here we find “depression”. If a mind which has become so riddled with invalidations and bogged down in anxieties that deadlock has set in, all learning and activity become impossible. A mind in such a state will lose the ability to act towards any goal in particular, and will have to prioritize its own healing over anything else. This "failure mode” is quite predictable, in light of present view, and it follows—again at some risk of treating a mind too much like a computer—that human nature ought to afford us some "scheduled downtime" to combat the tendency towards major depressions. This is the line of thinking that led to my previous post, where I argued that this is plausibly the function of "seasonal depression"—but it will only do its job if you don't fight it!
A further peculiarity of learning in humans is that, if we try and fail to figure something out for long enough, we’ll give up. This is the outcome of “acceptance”, and is arrived at by defining ourselves a certain way, e.g. “I’m not a math person”, or “… a sports person”.
On one hand, this is a reasonable way to think, and serves to put a stop to a learning process which is futile, letting one “backtrack” in the search process. On the other hand we very frequently reach this conclusion when it isn’t actually true, and once you’re convinced of such a thing, it will become impossible; it is nearly impossible to act in a way which contradicts your own self-image. When you’re on the brink of this decision, too, it is going to become very hard to learn; wondering whether there’s something wrong with how you catch balls, say, is going to get in the way of paying attention to the ball.4
Or, from a different angle, this self-definition can come as a relief, giving oneself permission to stop trying at or stop feeling bad about something (as in “oh, I’m just ADD”.)
I call these acts “essentializing” oneself, meaning the assignment of some attribute to your inherent essence or identity rather than identifying it as a consequence of your actions. Essentializing is the eventual result when so much invalidation has been backpropagated into your mental model that learning becomes impossible—when there is no longer any apparent direction of “action” or “attention” in which to go. For some single skill this just leads to “giving up”, which is quite understandable. But if you have to give up at enough skills, the invalidation will fall onto one’s self—a maladaptive sort of model-synthesis where the belief you arrive at is “I’m bad at everything”—which will lead to anxiety and depression.
The flip-side of essentializing oneself negatively is to essentialize positively, wherein you come to believe that you are uniquely talented or special at something, or in general. In one sense this obviously is a source of self-esteem, and, evolutionarily, can be seen as how a person “specializes” into a role they’ve shown talent at. It is helpful, too, to get better at something—by raising the bar for what can validate you. But I am inclined to think of this also as “pride”, with a rather negative connotation: once you believe something should come effortlessly and passively to you, it becomes difficult and potentially shameful to have to switch back to active learning when you reach your limits. To become great at something, then, requires you manage this process carefully: you have to continuously raise your definition of yourself, without undershooting (and becoming complacent) or overshooting (and making it impossible to learn without so much shame and anxiety you want to give up.) Validation and invalidation will come in waves—you have to ride them without falling off.
At this point I've called a number of similar things by the name "anxiety". It will be helpful to enumerate them:
The anxiety of a known threat, which a person has not yet figured out how to feel safe from.
The anxiety of not being able to reconcile different beliefs into a higher-order one, and being left with fears which circulate forever.
The anxiety of being too concerned with whether you are good at something to pay attention to the thing itself, which will tend to feed back into itself by making you worse at it.
The anxiety of having so many of one's thoughts and actions invalidated that no action is possible without risking further invalidation. One then has no choice but to give up.
These form a progression from basic-unsafety to full-blown mental illness.
All of this, note, is a predictable consequences of the "learning process" playing out in the constrained environment of the human mind. All of this follows from learning. And to me this gives a very clear and actionable picture of “mental illness” as a failure of learning, with none of the haunted or tragic mystique, or medicalized sterility, we normally associate with the words “anxiety” and “depression”.
I come to see what we normally think of as mental illness as the “backward pass” of life, as much a part of a whole life as the “forward pass” of acting and enjoying itself. If the backward pass takes over, it’s because it’s not working, and some of those problems need to be let go of—because life is finite, time is always constrained, and no amount of safety will protect you in the end from death itself.
IX. Culture As Collective Learning
The difficulty of interpreting our own feelings and beliefs is what causes the a person’s normal handling-of-threats (1) to start to pile up (2), interfere with learning (3), and eventually to get so knotted up they impede basic functioning (4). This is a kind of feedback loop; if learning falls behind it can be expected to become more knotted up over time, and, well—this is very bad!
This is the failure mode of human nature, I believe. Much of human culture can then be seen as various efforts to combat this failure mode. Religion is one: all the way back around 400 A.D., St. Augustine wrote in his Confessions that "every disordered mind becomes its own punishment", meaning exactly this, and it is not much of a stretch to identify his description of his own state of "sin" with the tangle of fears and misapplied heuristics I've described above. Christian ethics then consist of heuristics which, if followed, will tend to guard against this state. Each person then has a spiritual duty to maintain their own mind, and, while being unable to think clearly is no excuse for wrong action, at the same time forgiveness is possible—for "they know not what they do". The whole institution of "confession" exists to facilitate this. And here we can also locate the meditation, prayer, and rituals of many other. traditions. Spirituality, it seems, is about this problem in particular.
Secularly, we find "art" filling these roles, either as an opportunity for self-expression, or as an aid in recognizing oneself in another's self-expression, or in its ability to bring catharsis, as in a ritual setting.5
But the most basic way we handle fear is by talking about it—by processing our feelings in conversation with other people. Observe how a person with something bothering them will continuously bring it up in conversation, or, if they can't talk about the actual thing, will complain about something else so as not to let their feeling go unnoticed. So much of small talk, too, is just comparing notes about threats: did you see the news..?—did you hear what happened to...?—what's with these fucking LED headlights?
What it takes to "open up" a knotted mind is emotional safety, and the best way to feel safe is in the company of people who genuinely like, respect, and trust you. The mind will clamp down if there is any threat of further injury: of being judged as stupid or bad, or of someone else seizing the space for themself, or being cut short by the too-rapid giving of advice or consolation before the real feeling is even able to come out. Unpacking your feelings is selfish in this way, and it asks of others that they consciously let your selfishness take the room for a while, "holding space", which takes skill and effort and is what called "emotional labor" (to put couple more terms on the map). Holding space such that the sharer is safe in turn requires that the listener has worked through enough of their own feelings such that they don't feel too threatened by anything they hear and overreact. And this implies, rather tragically, that the failure to be listened-to is contagious—everybody needs space to be heard and nobody can give it.
And if you don't have anyone in your own life you can talk to, there's always therapy.
By now the scope of this "theory of everything" has started to show itself. To me, this framework offers a map, on which can be located a great many things which, while familiar, are hard to put into relation to each other. "Learning" links them all, and at the same time gives a clear purpose and direction to things like "art", "spirituality", and "emotional processing", which otherwise tend to get muddled by too many layers of moralization.
At this point I feel like I’ve said a lot, and really ought to draw things to a close. But a few more connections suggest themselves…
X. Learning in Societies
We’ve just seen how "conversation" is the primary setting for the processing of emotions. But at the same time, to talk about one’s feelings also serves to circulate them into the larger community. Here they can inform others' worldviews: every intermediary, hearing similar stories from different people, will start to recognize the patterns, and will be start to think about what can be done to solve the underlying problems. This is "backpropagation" playing out in a society-level learning process: we compare notes, share stories, and build consensus, ultimately working together towards a collective model-synthesis which aims to solve the problems which are threatening many people at once. And if something needs to be done urgently, this gossip will become emotionally and morally-charged. The effect is to submit the task to the collective to be taken up by, well, whoever is up to the task.6
And is this not the nature of almost all of politics? Clearly the degree of fear felt by individuals over political issues can often be wildly out of proportion to their personal experience; often the feelings are almost entirely caused by whatever news environment they partake of. (Think of environmentalism on the left, or anti-immigration sentiments on the right). A "movement" then can be seen as the sum of innumerable experiences of pain and fear all backpropagating together, taking aim at whatever social norms or institutions are causing the pain. (Here I think of feminism and anti-capitalism in particular.) At that point something must be changed in order to transmute the fear into action, "acceptance" is unacceptable, and if nothing is done for long enough then violence will be the predictable outcome.7
What usually happens is not a complete victory for the reformers but a model-synthesis of the old and new—it is usually unimaginable to actually do away with the old regime entirely. (The general public is not capable of contemplating an end to capitalism and cannot seriously engage with the idea.) The synthesis that actually happens is never enough to solve the problem, though, and the two paradigms wind up existing side-by-side. The society can then be said to be in a state of anxiety itself, exactly like that of an individual, except that in a society one has to learn to respect others' differences; the ideal end state is not to be "of one mind" but one of mutual respect.
Most of the institutions of the modern world are the result of someone “backpropagating” on a mountain of pain and hardship: liberalism—the Constitution—public education—worker's comp—the Fed—vaccines—civil rights—women's rights—etc.; this is the story of progress. It is an empirical fact that people expect to be able to hold institutions accountable for the things that hurt individual; in the long term any institution that doesn’t honor this duty is going to forced to change its ways.
Science and technology are the product of the same process—learning—but with “material scarcity” as the problem to be ultimately solved. (Though for practitioners, curiosity is usually the more immediate aim.)
Culture too: the ideals I inherited, for example, where liberal-elite ones esteeming literature, education, and music. To a kid the values of your parents feel objectively good and important, and it is rather confusing to grow up and find that they are only important to a narrow portion of society, and that most of their instantiations in the world seem to miss the point entirely. Seen through the lens of learning, these “ideals” instead take on a character of “heuristics”: they are ideas of what is important for a good life which were arrived at by certain people in a certain era, and which, like any heuristic, can be expected both to fail to generalize to other eras, and to fail to capture the essence of the underlying good of the thing. To see ones’ youthful ideals this way is to step free of them, that is, to become more free—free to borrow from each what is good and leave behind what is bad, and free to turn my attention to what is good and important in my own era.
And it is perhaps the most important conclusion of this view is that even the ethics and principles we take to be “good” and “right” are to be seen as only approximations to the ground truth of "good" itself. We take for granted, for example, that "natural rights" are real and important. But it is not the case that natural rights are a metaphysical fact of the universe; rather, this concept was arrived at by people, who were working backwards from an enormous accumulation of anger and grief to arrive at a solution to the problems of their own era. Since then, the concept has held up to scrutiny—it's a good one. Does this make natural rights "just a social construct" and therefore something we can do away with? Not at all. If we were to dismantle the concept and retread all that ground, we’d almost certainly come to those same conclusions; not so much has changed in human nature. But we might be able to do better. Seeing such “sacred and immutable” ideas instead as "the best-effort product of human wisdom" serves to demystify these heuristics—which, I think, is far more palatable, and actually makes them easier to respect, while allowing us to imagine how things could be otherwise in the face of new evidence.
"Natural rights" is an easy one, but the reader is encouraged to apply this view to the environment, immigration, capitalism, gender politics, YIMBYism, vaccines, the Constitution, civil rights, public education… I'll take on some of these in the future, but this is not the time open any more cans of worms.
XI. Learning As An Ethic
The theory I've laid out here is the result of my own years of backpropagating (okay I have to admit this word is corny) on all of the "errors" in the world: on the inadequacy of my own education, on left/liberal concerns like environmentalism and anti-capitalism, on the culture wars and political polarization of the 2010s, and on my own mental health and on the epidemic of mental illness in my generation, and the contagious ideas that cause these things. The result of this synthesis—which will take many more essays to convey in full—is a model in which almost all of these things can themselves be seen as learning processes which are failing in one way or another.
Each of us, as I observed above, operates by our nature according to an ethic of "safety first", and will naturally undertake the work of learning from our own mistakes (and others) if not prevented from doing so. If our attempts to solve the problem are inadequate—too low-order—then we’ll soon find out, and will have more learning to do. One does right by oneself, then, by learning well. Everything becomes a “skill issue”.
The same principle applies in a community, no matter the size: a relationship, a family, a town, a nation. And by the same token we are called to learn and work not only for our own safety, but for others’, both because their pain and fear becomes our own by way of empathy, and because the principles we work out to solve our own problems will demand that we solve others’ as well.
The underlying principle then is to keep learning, and to this end, all of human suffering is information—it tells us what we need to do. This asks nothing of us which is not already in our nature. The difficulty is in doing it well.
All suffering, in this view, is information. Take that, Buddhists.
“Trauma” can be seen as the same sort of low-order heuristic, but instantiated in the body or in more primal parts of the nervous system rather than in the beliefs of the mind. PTSD responses are the (often counterproductive) reactions of those “stuck” models. My understanding is that effective "trauma therapy" usually requires a physical or somatic intervention to release the stuck fight/flight/freeze response, after which what will remain is a mere belief which can then be learned-from.
This is a very general feature of human nature, roughly equivalent to “flow state”, wu wei, Alexander technique, and Yoda’s line “do or do not do, there is no try” from Star Wars. If you want to do a thing you can’t be thinking about you—you have to be thinking about the thing!
If you need a standard from which to judge art as Good or not, the learning picture has one to offer: does it aid in clarifying and untangling fears—does it transmute fear? Or does it merely spread it?
I have a relatively low opinion of "therapy" in part because it gives up on the essential aim of "circulating" one's feelings into the world. To merely move through feelings without changing anything requires accepting a tragic degree of impotence! The patient is not fooled, nor are they fooled that the person "holding space" for them is doing it because they are paid.
Think of the UHC CEO shooting—Luigi's own healthcare fiasco is hardly enough to explain the assassination on its own. But it makes perfect sense when seen as a concentrated expression of the collective's anger against the health insurance industry. This is how human minds actually work: much of what we do is really the collective recruiting us to its own ends. To imagine otherwise is to be rather out of touch with reality.