On recommendation from several LessWrongers I've been over the past year or so occasionally digging into the many long posts to be found in the archives of Unqualified Reservations (archive links best accessed from here). It is written by Mencius Moldbug, who is probably familiar to many of us here as well as to readers of Overcoming Bias. He is an erudite, controversial and most of all contrarian social critic and writer.

He sometimes repeats and often refines his key ideas. He uses his writing style as a barrier to entry (it is debatable if this does more harm than good for his quality of thought and communication, but it is an interesting way to aim for the correct contrarian cluster), thus he is an acquired taste, posters have recommended the gentle introduction series as good place to start reading him. This series is similar, while this one and The Formalist Manifesto focus more on summarizing his political thought, which may also be useful in itself.

Link to topical entry is here. Link to discussion on previous entry I read is here.

Five ways to classify belief systems

I use the word kernel to mean "belief system." Kernels, like Gaul, are divided into three parts: assertions about the real world (Hume's "is"), moral judgments about the real world (Hume's "ought"), and paranormal or other metaphysical propositions (such as David Stove's wonderful ruminations on the number 3).

Everyone, no matter how smart or stupid, has exactly one kernel. However, kernels are not assigned randomly, as if in some weird Buddhist boot process. For example, your kernel is likely to show similarities to that of your parents, friends, teachers, karate masters, favorite anchormen, etc, etc.

Let's call a kernel pattern which many people share a prototype. Methodism, environmentalism, firearms practice, snake handling and Burning Man attendance are all prototypes. While there are few Methodist environmentalists who are also snake-handling marksmen and never miss a burn, various subcombinations are not uncommon.

In general we are most interested in complete prototypes, that is, kernel patterns that are broad enough to serve as identities. It is common to describe someone as "a Methodist," or (not quite in the same way) as "an environmentalist." People who match the other prototypes above may use nouns for themselves, but they're must less likely to be described or introduced as such. An incomplete prototype simply says less about you. For example, many snake handlers are also committed peace activists who drive Range Rovers and shop at Pottery Barn.

Two common examples of a complete prototype are religions, which involve convictions about one or more anthropomorphic paranormal entities, and idealisms, which involve convictions about one or more undefined universals, or ideals.

Many people consider the distinction between religion and idealism important and/or interesting, but here at UR we don't much care for it, since only metaphysical propositions can distinguish the two. You can go from religion to idealism and back simply by adding and subtracting gods, angels, demons, saints, ghosts, etc. I personally have slain many ghosts and quite a few demons, and I once kidnapped an angel and forced her at swordpoint to lead me to the altar of Thoth, where I sacrificed her for 20,000 experience points, permanent immunity to fire, and an alignment change to chaotic evil. However, this was not in real life. And even in D&D, I've never had the misfortune to encounter a god.

On LessWrong already discussed a more extensive form of this argument in "Belief in religion considered harmful?".  

Therefore, we'll just use the word prototype to mean either religion or idealism. Of course one can study either forever. In fact, most scholars in history have spent most of their time investigating the twisty little passages, all alike, of one single prototype. However, since here at UR we are generalists, not Irish monks, Talmudic scribes or Koranic talibs, we will try and work a little more broadly.

Before you can really think about prototypes, you have to be able to name and classify them. One obvious analogy is the study of languages, which are transmitted from person to person in a vaguely similar way. Prototype transmission really has nothing in common with language transmission, but the metaproblems are the same: what does it mean to say, "X descends from Y?" Is a classification tree a tree, or a directed acyclic graph? Is variation continuous, or discrete? Etc, etc, etc.

Probably readers can add a few, but I can think of five ways to classify prototypes: nominalist, typological, morphological, cladistic, and adaptive.

As our example for each, let's use the movement generally known as the Enlightenment. There is no noun for people whose kernels match the Enlightenment prototype, but there should be, because this noun arguably applies to almost everyone on earth. Let's call these suspicious characters Luminists. Their sinister views can be described as Luminism.

A nominalist classification simply accepts the prototype's classification of itself. Luminists, for example, believe there is no such thing as Luminism. (This is very common.) Rather, they are simply people who have seen the light of reason. It just so happened that they all saw more or less the same light at more or less the same time. But since by definition there's only one such thing as reason, this explanation is not inherently implausible.

A typological classification distinguishes prototypes according to specific features. For example, when you distinguish between religions and idealisms - as between Christianity and Luminism - you are performing an act of typology. The flaws in this approach can be seen by the fact that a typological classification of languages tells us Old Saxon is a dialect of Early Apache, since they both have arbitrary word order and long, incomprehensible sentences. Meanwhile, a vampire bat is a grinning, hairy owl, IHOP and Domino's both serve round food, Congress is considering a new O visa for ostriches, Burmese tribeswomen and other long-necked bipeds, and Luminism is a kind of Confucian Sufi-Buddhism.

A morphological classification is like a typological classification with a clue. It attempts to construct a historical descent tree by looking at multiple points of similarity. Morphological classification tells us that Luminism is actually a sect of Christianity, because Luminists share a wide range of kernel features with many Christians, and there are even intermediate forms which can reasonably be described as Christian Luminists or Luminist Christians.

A cladistic classification also produces a historical descent tree, but it uses a completely different method. Cladistic classification ignores actual beliefs and looks only at patterns of conversion. It asks: if you are a Luminist and your parents were not Luminists, what were they? Since the answer is usually (if not always) "Christian," in this case cladistics produces the same result as morphology For obvious reasons, this is often so.

Besides the usual trees, both morphological and cladistic methods can also produce graph structure, that is, patterns of combination or syncretism. For example, both methods identify Hellenistic and Jewish roots for Christianity, with the cladistic method adding various Roman cults such as those of Augustus, Sol Invictus, and Mithra.

An adaptive classification is not interested at all in descent. Rather, it focuses on how and why the prototype succeeds. For example, Luminism, Christianity, Sol Invictus and Islam are all prototypes that succeeded (at one time or another) by virtue of being an official prototype, that is, by explaining the legitimacy of a government - helping to organize its supporters, strike fear into the hearts of its enemies, brainwash its dutiful taxpaying serfs, etc, etc, etc. But with the exception of the third, all the above have also done just fine in an unofficial capacity, so this official selection is not a complete explanation of their success.

Of course, I personally find the last three classification methods the most compelling, with my favorites being the morphological and adaptive methods. But words are just words, and anyone can look at these phenomena any way they like. And if you can suggest any additions to the list, the comments section is, as usual, open.

Upon introspection I generally seem to implicitly use adaptive frames for "kernels" in ancient societies I don't know very well (or which don't have a well preserved written history - say like an explanation for widespread human sacrifice in Mesoamerica) and when I just read something by Dawkins. Morphological when thinking about religion in ancient literate societies I know quite a bit about like say the Roman Empire and nominalist when deciding how I classify modern religions like Mormonism.

There are other examples, but overall Moldbug's division seems to me to capture most of my differing approaches to thinking about "kernels" and indeed they do seem to be running on different algorithms. The obvious question which I hope to discuss in the comment section is which of these approaches is most useful under different sets of circumstances and goals.

Doing some thought on it his take on the concept and his division of the categories. In itself it seems a somewhat useful framework for thinking about intellectual fashion and ideological or religious transformation.