15. Why Essentialism Never Works

For our purposes, we can define essentialism as “explaining a phenomenon in terms of a special essence that conveys the phenomenon.” For example, many people believe that the explanation for human consciousness and/or intelligence is the immaterial soul. Mechanistic explanations, by contrast, model phenomena as systems of moving parts. The dynamics of these parts are described, usually mathematically, and the behavior of the system as a whole is predicted using that description.

There is a recurring theme in the history of science in which an essentialist explanation for a phenomenon is overturned by a mechanistic explanation. People thought heat was a fluid called phlogiston, but the kinetic theory of heat won. It was supposed in most cultures and for most of history that living beings must be distinguished from dead and inanimate ones by the presence of some spirit, or animus, or elan vital. But it is the mechanistic theory of life that has transformed our world. Now, as AI becomes more powerful and more salient, some people insist that it will permanently plateau below the level of full general intelligence because it lacks some vital mental essence which humans possess. We’ll see how that prediction pans out.

The reverse has, to my knowledge, never happened. Is this merely an accident of history? Is it equally likely, in a given situation, for a mechanistic or essentialist explanation to be superior, and we’ve just happened to flip 20 heads in a row? Or is there a reason essentialism is always in retreat?

Julian Huxley once joked that saying life is driven by a “vital force” is like explaining trains by hypothesizing a “locomotive force”. The thing about essentialist explanations -vital forces, spirits, souls, etc- is that they don’t do anything. Okay great, life is driven by a vital spirit. How does that spirit work? Can it be isolated in a lab? Is it discrete or continuous? Are there different kinds of it, and if so how could that be measured? None of these questions are ever answered. And because of this, no new technology can be developed using essentialism. And I am using technology in the broadest sense of the term: the application of knowledge to solve problems. At best, we get negative meta-predictions of the form “you will fail to find a mechanistic explanation for this phenomenon”.

It is important to note that not all essentialist explanations are completely vacuous, as described above. Some are straightforwardly false instead. “My cows are dying because my neighbor is cursing me. If I kill him, my cows will stop dying” is, in its structure, a perfectly valid hypothesis. It has an explanation (my neighbor is cursing me) and a prediction (if he dies, the effect will end). Eminently falsifiable. And, in fact, false.

What would it be like for essentialist explanations to work? Imagine we’re in the Van Helsing universe. Werewolves and vampires and demons abound. The best known ways to fight them come from the Church: holy water, crosses, relics, etc. One day, grizzled monster hunter Van Helsing rushes back to the Vatican with news of a new kind of creature: a great mass of green and purple ooze slowly swallowing up villages. He tells his priest, and the priest consults the arch demonologist. What happens at this point? There are a few possibilities:

1. The demonologist and his stacks of books are essentially a lookup table, and green and purple ooze is not on the list. This would mildly downgrade our confidence in the perfection of the Church’s world model. Perhaps their imperfect lists were just arrived at through trial and error over centuries, and their explanations about Jesus are just flimflam added on top.

2. The demonologist and his stacks of books are essentially a lookup table, and green and purple ooze is on the list. He listens to a thorough description of the ooze. “Purple and green eh? With or without iridescent swirls? Without? That’s a hargbeest from the seventh circle of Hell. The Secret Scriptures say to kill them with aluminum blessed by a clergyman ranked bishop or above. There are no records of anyone encountering one of these. Some people even speculated the entries on them were forgeries. I guess they’re not.” This is already getting a bit technical sounding - specific metals, specific ranks of priest - but it is not a technical explanation. There’s no theory as to why aluminum kills hargbeests. While this scenario would make me curious about how the Church came to possess such a good lookup table, and would at least mildly update me towards their theological position, it is still probably more likely that the knowledge came either from trial and error or from someone who had a mechanistic explanation that has been forgotten. This scenario does not really involve essentialist explanations. It involves no explanations at all.

3. The demonologist hears the description of the monster, and takes out paper and ink. “You’ve heard of it then?” asks Van Helsing. “No,” says the demonologist, who we can call Gregory, “never. Nothing in the bestiaries about something like this. How big did you say it was? And was the iridescence caused by the surface shifting, or something moving under the surface?” He steadily takes notes. After he’s done asking questions, he continues writing for a bit. When he’s finished, he says “whatever this thing is, it’s very likely made primarily of abyssal ectoplasm, with some infernal-ectoplasmic admixture. Given the size, you’ll need 2 quarts of holy water infused with rosemary. You’ll need to inject it past the skin. Come, the cool engineer priest who makes your fully automatic crossbows has been working on a giant battle syringe.” And, indeed, this plan works, though Van Helsing wishes Gregory had warned him that the ooze was going to explode.

Only in the third scenario is the explanation able to generalize out of distribution, and that’s precisely because it is mechanistic! While Gregory did not know about this specific creature, he was able to identify the likely composition of its skin and its interior, and thus devise a plan to destroy it. He understood how the components behave in isolation, then deduced how the whole would respond to different stimuli.

Would this event be enough to vindicate all Catholic beliefs? It would be evidence in favor of the theory used by the demonologist. Or rather, it would be evidence in favor of the minimal equivalent theory: the simplest theory that makes all the same predictions. Any concepts used in the minimal equivalent theory are thus supported in their reality by the theory’s success. If the theory just doesn’t work without these creatures appearing from some other parallel world, then the success of the holy water injection is evidence for the existence of such a parallel world. If it does work without that part, it’s not. Perhaps one day Van Helsing will encounter a Buddhist monk who gives him the exact same monster hunting tips about holy water and silver bullets but with an entirely different mythological structure justifying it.

And that’s the key thing: what it means for explanations to work is exactly that their parts all fit together and connect to something measurable in a useful way. An explanation is useful if it provides more information than a mere list of all previous observations. “Everyone I’ve seen eat the red mushrooms has died” and “everyone I’ve seen eat the red mushrooms has died, because spirits got into them” make the same predictions. The bit about spirits isn’t helping unless it allows you to predict in advance what other mushrooms are poisonous.

The only way for intangible things like gravity or souls or poisonous compounds in mushrooms to be supported by evidence is for theories that involve them as essential concepts to make falsifiable predictions that come true. And a key component of being falsifiable is being precise. You need your intangible concept to produce precise, testable predictions about what is going to happen. For this to work, it has to be sensitive to inputs, otherwise it would predict the same thing in all situations. So it needs to map inputs to outputs in a relatively precise way. That’s just mechanism again. To the precise extent an explanation becomes useful, it becomes mechanistic. Explanations are mechanisms.

Essentialism, then, is not an alternative scientific paradigm but rather an unrelated language game that has nothing to do with understanding the universe. Essentialism is not explanation-shaped.