My rough guess as to where you’re going with this is something like “scenario 1 is a waste of words since scenario 2 achieves the same results more efficiently (namely, the misunderstanding is cleared up either way).”

Basically, yes.

The problem, really, is—what? Not misunderstanding per se; that is solvable. The problem is the double illusion of transparency; when I think I’ve understood you (that is, I think that my interpretation of your words, call it X, matches your intent, which I assume is also X), and you think I’ve understood you (that is, you think that my interpretation of your words is Y, which matches what you know to be your intent, i.e. also Y); but actually your intent was Y and my interpretation is X, and neither of us is aware of this composite fact.

How to avoid this? Well, actually this might be one of two questions: first, how to guarantee that you avoid it? second, how to mostly guarantee that you avoid it? (It is easy to see that relaxing the requirement potentially yields gains in efficiency, which is why we are interested in the latter question also.)

Scenario 1—essentially, verifying your interpretation explicitly, every time any new ideas are exchanged—is one way of guaranteeing (to within some epsilon) the avoidance of double illusion of transparency. Unfortunately, it’s extremely inefficient. It gets tedious very quickly; frustration ensues. This approach cannot be maintained. It is not a solution, inasmuch as part of what makes a solution workable is that it must be actually practical to apply it.

By the way—just why is scenario 1 so very, very inefficient? Is it only because of the overhead of verification messages (a la the SYN-ACK of TCP)? That is a big part of the problem, but not the only problem. Consider this extended version:

Scenario 1a:

Alice: [makes some statement]

Bob: What do you mean by that? Surely not [straightforward reading], right? Because that would be obviously wrong. So what do you mean instead?

Alice: Wait, what? Why would that be obviously wrong?

Bob: Well, because [reasons], of course.

So now we’ve devolved into scenario 2, but having wasted two messages. And gained… what? Nothing.

Scenario 2—essentially, never explicitly verifying anything, responding to your interpretation of your interlocutors’s comments, and trusting that any misinterpretation will be inferred from your response and corrected—is one way of mostly guaranteeing the avoidance of double illusion of transparency. It is not foolproof, of course, but it is very efficient.

Scenarios 1 and 2 aren’t our only options. There is also…

Scenario 3:

Alice: [makes some statement]

Bob: Assuming you meant [straightforward reading], that is obviously wrong, because [reasons].

Note that we are now guaranteed (and not just mostly guaranteed) to avoid the double illusion of transparency. If Bob misinterpreted Alice, she can correct him. If Bob interpreted correctly, Alice can immediately respond to Bob’s criticism.

There is still overhead; Bob has to spend effort on explaining his interpretation of Alice. But it is considerably less overhead than scenario 1, and it is the minimum amount of overhead that still guarantees avoidance of the double illusion of transparency.

Personally, I favor the scenario 3 approach in cases of only moderate confidence that I’ve correctly understood my interlocutor, and the scenario 2 approach in cases of high confidence that I’ve correctly understood. (In cases of unusually low confidence, one simply asks for clarification, without necessarily putting forth a hypothesized interpretation.)

Scenarios 2 and 3 are undermined, however—their effectiveness and efficiency dramatically lowered—if people take offense at being misinterpreted, and demand that their critics achieve certainty of having correctly understood them, before writing any criticism. If people take any mis-aimed criticism as a personal attack, or lack of “interpretive labor” (in the form of the verification step as a prerequisite to criticism) as a sign of disrespect, then, obviously, scenarios 2 and 3 cannot work.

This constitutes a massive sacrifice of efficiency of communication, and thereby (because the burden of that inefficiency is borne by critics) disincentivizes lively debate, correction of flaws, and the exchange of ideas. What is gained, for that hefty price, is nothing.

Zetetic explanation

by Benquo 5 min read27th Aug 2018138 comments


There is a kind of explanation that I think ought to be a cornerstone of good pedagogy, and I don't have a good word for it. My first impulse is to call it a historical explanation, after the original, investigative sense of the term "history." But in the interests of avoiding nomenclature collision, I'm inclined to call it "zetetic explanation," after the Greek word for seeking, an explanation that embeds in itself an inquiry into the thing.

Often in "explaining" a thing, we simply tell people what words they ought to say about it, or how they ought to interface with it right now, or give them technical language for it without any connection to the ordinary means by which they navigate their lives. We can call these sorts of explanations nominal, functional, and formal.

In my high school chemistry courses, for instance, there was lots of "add X to Y and get Z" plus some formulas, and I learned how to manipulate the symbols in the formulas, but this bore no relation whatsoever to the sorts of skills used in time-travel or Robinson Crusoe stories. Overall I got the sense that chemicals were a sort of magical thing produced by a mysterious Scientific-Industrial priesthood in special temples called laboratories or factories, not things one might find outdoors.

It's only in the last year that I properly learned how one might get something as simple as copper or iron, reading David W. Anthony's The Horse, the Wheel, and Language and Vaclav Smil's Still the Iron Age, both of which contain clear and concrete summaries of the process. Richard Feynman's explanation of triboluminescence is a short example of a zetetic explanation in chemistry, and Paul Lockhart's A Mathematician's Lament bears strong similarities in the field of pure mathematics.

I'm going to work through a different example here, and then discuss this class of explanation more generally.

What is yeast? A worked example

Recently my mother noted that when, in science class, her teacher had explained how bread was made, it had been a revelation to her. I pointed out that while this explanation removed bread from the category of a pure product, to be purchased and consumed, it still placed it in the category of an industrial product requiring specialized, standardized inputs such as yeast. My mother observed that she didn't really know what yeast was, and I found myself explaining.

Seeds, energy storage, and coevolution

Many plants store energy in chemicals such as proteins and carbohydrates around their seeds, to help them start growing once they're in wet ground. Some animals seek out the seeds with the most extra energy, and poop the occasional seed elsewhere. Sometimes this helps the plant reproduce more than it otherwise would have; in such cases, the plant may coevolve with the animals that eat it, often investing much larger amounts of energy in or around the seed, since the most calorific seeds get eaten most eagerly.

Humans coevolved with a sort of grass. If you've seen wild grass, you may have observed stalks with seed pods on them, that look sort of like tiny heads of wheat. Grain is basically massively a grass that coevolved with us to produce plump, overnourished seeds.

Energy extraction

Of course, there's only so much we can do to select for digestibility. Often even plants that store a lot of surplus energy need further treatment before they're easy to digest. Some species evolved to specialize in digesting a certain sort of plant matter efficiently; for instance, ruminants such as cattle and sheep have multiple stomachs to break down the free energy in plant matter. Humans, with unspecialized omnivorous guts, learned other ways to extract energy from plants.

One such way is cooking. If you heat up the starches inside a kernel of wheat, they'll often transform into something easier to digest. But bread made this way can still be difficult to digest, as many eaters of matzah or hardtack have learned. Soaking or sprouting seeds also helps. And a third way to make grains more digestible is fermentation.

Cultured food

Where there's dense storage of energy, there's often leakage. Sometimes a seed gets split open for some reason, and there's a bit of digestible carbohydrate exposed on the surface. Where there's free energy like this, microbes evolve to eat it.

Some of these microbes, especially fungal ones, produce byproducts that are toxic to us. But others, such as some bacteria and yeasts, break down hard-to-digest parts of wheat into substances that are easier for us to digest. Presumably at some point, people noticed that if they wet some flour and left it out for a day or two before cooking it, the resulting porridge or cracker was both tastier and more digestible. (Other fermented products such as sauerkraut may have been discovered in a similar way.)

Of course, while grain-eating microbes will often tend to be found on grain, allowing for such accidental discoveries, there is no guarantee that they'll be the kind we like. Since they mostly just eat accidental discharges of energy, there also just aren't very many of them, compared to the amount of energy available to them once the flour is ground up and mixed with water. It takes a while for them to eat and reproduce enough to process the whole batch.

Eventually, people realized that if they took part of a good batch of dough or porridge and didn't cook it, but instead added it to the next batch, this would yield an edible product both more reliably (because the microbes in the starter would have a head start relative to any potentially harmful microbes) and more quickly (again, because they'd be starting with more microbes relative to the amount of grain they needed to process). This is what we call a sourdough "culture" or "starter".

(You can make a sourdough starter at home by mixing some flour, preferably wholemeal, with water, covering it, and adding some more flour and water each day until it gets bubbly. Supposedly, a regularly fed starter can stay active for generations.)

Breads are particularly convenient foods for a few reasons. First, grains have a very high maximum caloric yield per acre, allowing for high population density. Second, dry grains or flour can be stored for a long time without going bad; as a result, stockpiles can tide people over in lean seasons or years, and be traded over large distances. Third, a loaf of bread itself has some amount of more local portability and durability, relative to a porridge.

Yeast-specific products

One of the microbes found in a sourdough culture, yeast, has a particularly simple metabolism with two main byproducts. It pisses alcohol, and farts carbon dioxide. Carbon dioxide is a gas that can leaven or puff up dough, which makes it nicer to eat. Alcohol is a psychoactive drug, and some people likes how it makes them feel. Many food cultures ended up paying special attention to grain products that used one or the other of these traits: beer and leavened bread.

In the 19th century CE, people figured out how to isolate the yeast from the rest of the sourdough culture, which allowed for industrial, standardized production of beer and bread. If you know exactly how much yeast you're adding to the dough, you can standardize dough rising times and temperatures, allowing for mass production on a schedule, reducing potentially costly surprises.

The price of this innovation is twofold. First, when using standardized yeast to bake bread, we forgo the digestive and taste benefits of the other microbes you would find in a sourdough starter. Second, we become alienated from a crucial part of the production of bread, to the point where many people only relate to it as a recipe composed of products you can buy at a store, rather than something made of components you might find out in the wild or grow self-sufficiently.

Additional thoughts on explanation

I'm having some difficulty articulating exactly what seems distinct about this sort of explanation, but here's a preliminary attempt.

Zetetic explanations will tend to be interdisciplinary, as they will often cover a mixture of social and natural factors leading up to the isolation of the thing being explained. This naturally makes it harder to be an expert in everything one is talking about, and requires some minimal amount of courage on the part of the explainer, who may have to risk being wrong. But they're not merely interdisciplinary. You could separately talk about the use of yeast as a literary motif, the chemistry of the yeast cell, and the industrial use in bread, and still come nowhere close to giving people any real sense of why yeast came into the world or how we found it.

Zetetic explanations are empowering. First, the integration of concrete and model-based thinking is checkable on multiple levels - you can look up confirming or disconfirming facts, and you can also validate it against your personal experience or sense of plausibility, and validate the coherence and simplicity of the models used. Second, they affirm the basic competence of humans to explore our world. By centering the process of discovery rather than a finished product, such explanations invite the audience to participate in this process, and perhaps to surprise us with new discoveries.

Of course, it can be hard to know where to stop in such explanations, and it can also be hard to know where to start. This post could easily have been twice as long. Ideally, an explainer would attend to the reactions of their audience, and try to touch base with points of shared understanding. Such explanations also require patience on both sides. Another difficulty this approach raises is that plain-language explanations rooted in everyday concepts may not match the way things are referred to in technical or scientific literature, although this problem should not be hard to solve.

In some cases, one might want to forwards-chain from an interesting puzzle or other thing to play with, rather than backwards-chaining from a product. Lockhart seems to favor exploration over explanation for mathematics, and of course there's no particular reason why one can't use both. In particular, the explanation paradigm seems useful for deciding which explorations to propose.

Related: The Steampunk Aesthetic, Truly Part Of You