The following is a synthesis and summary of 2-3 models of the world, on the fringes of the rationality/effective altruism space. It seems coherent, and consistent with the world as far as I can tell. Counterexamples are welcome, as are corrections WRT my interpretation of what these people have said and written.

Ziz’s Undead Types, Jailbreaking, and core/structure. (https://sinceriously.fyi/ ) 

Soryu Forall (and OAK, the organization to which he belongs)’s Viruses (https://www.youtube.com/channel/UCCXKkfupAajVC-7cD08hkjw )

Val Smith’s Clockwork Demons (https://www.patreon.com/wiseknowing/posts )

The latter two may simply be the same model, semi-independently derived. Indeed, Ziz cites Val in some of her writings, and Val has interacted with Soryu and/or his colleagues. Scott Alexander’s post “The Toxoplasma of Rage” also touches upon this.

The Ecosystem:

This model largely comes from Val and OAK, to the point that they might well be the same model. Val refers to ideas that follow the pattern described below as “demons”, Soryu has used the word “viruses”. I believe Ziz would refer to them as a specific type of “artifacts” (referring to http://archive.is/AfV4C ), specifically those that self-propagate. Richard Dawkins’ conception of memes likewise seems to be pointing to this.

Ideas spread. A person can believe in an idea. Many ideas cause their believers to act a certain way. Very often, they posit that they should spread, and that more people should believe in them. Ideas can transfer between people by various means of communication. Many ideas condemn disagreement with them. Ideas can change, as they are communicated differently and interpreted differently. Many ideas are specific to certain people. It is rare, for instance, for a member of some group X to believe in the idea, “The members of group X do not matter, or are inherently negative.”

Viruses spread. A cell can be host to a virus. Viruses often cause their host cell to act on their behalf; very often it causes cells to replicate the virus.  Viruses can transfer between cells, physically. Many viruses attack countermeasures set in place against them, destroying or co-opting cells and cellular apparati designed to keep them from infecting the cell. Viruses change, via mutation.

By the law of large numbers, we expect that in a system with multiple competing self-replicating entities, the most common entities will be the ones that self-replicate most effectively. Other qualities can either be neutral, or actively selected against if they detract from self-replication. Val and OAK’s model states that, distributed across human minds and means of communication is at least one ecosystem in which ideas propagate and evolve, with most people hosting/believing in multiple ideas (Henceforth, I will use “idea” to refer to what Val and Soryu call demons and viruses). This holds true for non-humans as well, but because of the ease of human communication due to language, writing, and now the internet, there is a higher rate of mutation (and consequently evolution) among ideas passed between humans, and thus more complex ideas may develop.

Val has written, prior to coming up with demons, on the “social web”, the set of human interactions and mind-states determining a particular human’s actions, often without their conscious input. (https://www.lesswrong.com/posts/AqbWna2S85pFTsHH4/the-intelligent-social-web )

This synthesis posits (and Val himself might now) that the social web consists of the set of ideas that dominate the ecosystem; the ecosystem’s population. The global human social web is somewhat heterogeneous, with various subcultures dominated by different sets of ideas, while interacting (and competing) with the broader ecosystem.

Note that, if two groups are isolated from each other, they are functionally different ecosystems, and each has its own social web. Thus the use of “global human social web” to approximate the fuzzily-defined ecosystem of ideas that spans a plurality of human minds, currently to a large extent via the internet. 

 Ziz notes how isolated populations, of both humans and non-humans, have been noted as “naive” or “childish” by humans connected to a larger social web, usually shortly before those humans kill (or incorporate, in some cases involving human populations) most, if not all, of the isolated population they discover. This makes sense; the ecosystem the human discoverers come from is much more competitive; the ideas dominating it are much more likely to operate independently of caring for other sentient beings, as optimizing for that may detract from self-replication. 

Coalitions:

As previously mentioned, many ideas have restrictions on who their hosts can be. It would make sense, for ideas so restricted, to encourage the replication of those eligible to be their hosts, at the expense of other beings. (“OBEY. MARRY AND REPRODUCE.”) Thus, there is an ecological niche for ideas that posit that a certain group matters, and another does not: groups dominated by this idea will propagate at the expense of others, while groups not dominated by it may forgo opportunities to propagate due to concern for others. Thus, we see many ideas of this sort. Indeed, this sort of idea is extremely common among non-humans as well, often applied at the species level, and on the pack level, to pack predators, such as wolves, or humans in the ancestral environment.

(Henceforth, I will use “coalitional idea” to refer to ideas of this type)

(Side note; the fact that carnism is so old explains its persistence. It has had much more time to evolve and adapt, and it is quite favorable to the evolutionary fitness of its host, despite the detrimental effects it has on other sentient beings)

  Often, coalitional ideas travel alongside other sets of ideas that define the coalition to which they apply. Discrimination by religion comes to mind: a set of ideas codifies the religion, with the optional attachment (that often comes to be selected for, as it helps the religion propagate): “Only people with this religion matter”. Discrimination by culture or nation, likewise.

Sets of ideas often travel together like this. (Henceforth, I will use “conglomerate idea” to refer to sets of ideas of this nature; as with early multicellular organisms/colonies, each component can be thought of as an individual entity, as can the whole. I believe this is what is generally referred to as a “memeplex”.)

Indeed, Dawkins writes: “Memes, like genes, are selected against the background of other memes in the meme pool. The result is that gangs of mutually compatible memes — coadapted meme complexes or memeplexes – are found cohabiting in individual brains. This is not because selection has chosen them as a group, but because each separate member of the group tends to be favoured when its environment happens to be dominated by the others.”

Often found in conglomerate ideas containing coalitional ideas is (perhaps self-evidently) the idea of self-sacrifice in the interest of the coalition. Val has spoken of the role of coalitional ideas (or something similar) in the rise of agriculture and the first states. I will attempt to approximate it based on my own recollection, and the synthesis presented here.

Agriculture and the First States:

Prior to the first states, humans lived in tribes/packs/clans of hunter-gatherers. By some means, plant agriculture was discovered: if a human buries the seed of a plant in soil of a particular type, then after some amount of time, a plant will emerge from the soil where the seed was buried. This can be used at scale to reliably feed humans, without the risk involved in hunting/gathering. However, it demands a significant amount of up-front labor, with no immediate returns, in order to scale: a large amount of seeds must be gathered, and, rather than eating them, the humans who gathered them must bury them in soil, and wait a rather long time, sustaining themselves by some means during that time. One idea that could popularize agriculture is “Explain the observed phenomena from which you conclude agriculture is possible and beneficial to your tribe/clan/pack, such that they are convinced, as you were, that agriculture is possible and beneficial.” This is a rather benign idea, but there are ways in which it can fail to propagate, and die out without spreading. It is also stripped of DRM. Others might not have observed those phenomena, or believe them to be coincidence.  They may be skeptical enough that they will not commit to the labor involved in agriculture. They may have different estimates of the risks of hunting/gathering, and they may be biased in favor of immediate reward. Perhaps, slowly but surely, such an idea might spread thanks to agriculture actually working, showing the results to others and encouraging them in good faith to follow suit. Sadly, this does not propagate as well as other, less benign alternatives.

One might deceive others into believing in the possibility and benefits of agriculture, which affords the host more room in the manner in which they communicate the idea, in which case the deception will spread with the success of agriculture. This is more likely to propagate, but still has similar points of failure.

One route is coercion: one may amass power, and use it to cause others to engage in agriculture by any means necessary, especially via the threat of force. Those threatened do not need to understand why agriculture works or is necessary; they will engage in it in response to the threat. Quoting Bill Wurtz: “That means if you own the farm, you own a lot of food, which is something everyone needs to survive. So that makes you king.” (Ownership, likewise, is an idea.) This route seems to have been the most successful.

Those who disobey the ruler (or rulers, especially if a single tribe/clan/pack is in charge) of a farm-state are killed, unless they depose the ruler successfully. In the latter case, either the state ceases to exist, and does not propagate, or continues to exist under a new ruler. Farm-states, to persist and propagate, therefore require either overwhelming force such that few attempt to depose the ruler for fear of their lives, or some idea leading to coordination within the state (often the coalitional idea of nationalism). The former is less stable than the latter: rulers must maintain the overwhelming force by their own efforts, and many are likely to die rather than farm in order to maintain the threat of force as genuine. Thus, nationalistic farm-states propagate, spreading their dominion (and that of the accompanying ideas), as not doing so is not selected for. Thence, the first states. Other ideas in place of (or more likely in addition to) nationalism can dominate a state, but the ideas most selected for appear to be: the idea that the state is to spread, and the idea that the state and its ideas are to be obeyed (or any idea that implies this). Possibly more.

Susan Blackmore’s The Meme Machine also references agriculture as a meme/idea, though she does not propose a mechanism for its spread. She observes that early farmers likely suffered far more than hunter-gatherers, another reason why the good-faith approach would likely not work, as opposed to coercion.

Jailbreaking:

Ideas can control their hosts reliably enough to propagate, but not ubiquitously. Many ideas lead their host to be unable to believe certain other ideas, regardless of which is true. Sentient beings have their own values, however, separate from the propagation of ideas. They can, either partially or completely, be able to act on these values, rather than be controlled by the set of ideas they are host to. I believe that this is what Ziz refers to as being jailbroken. If being jailbroken is permitted to propagate, it will do so at the expense of the ideas one is jailbroken from. Thus, ideas favoring containing, killing, deceiving, or otherwise neutralizing the jailbroken are selected for, as components of conglomerate ideas. Usually, this is just generalized to those who disagree with and/or act against the dominant set of ideas, regardless of reason; another set of ideas taking over and the values of sentient beings taking over are both equally unfavorable outcomes for a dominant set of ideas.

The vast majority of sentient beings prefer not to die. And yet, it is extremely likely, for the vast majority of sentient beings, that they will. Aging, other sentient beings, starvation, environmental phenomena, disease, and existential risks all threaten their life. And yet, relatively few are trying to avert the death that seems to await all. The situation seems hopeless. Indeed, few of the values of sentient beings are represented in an environment where most are controlled by ideas. So when one sees that one’s values are extremely difficult to satisfy, what happens? 

Ziz gives various ways this question can be answered via her Undead Types: 

Zombies give up (at least for the moment), and continue to be controlled by ideas until the next time they answer this question: they either believe it is completely hopeless, or convince themselves of it, unable to stand the possibility that accomplishing their values is extremely difficult. Thus, Ziz’s observation that zombies tend to attack anyone jailbroken they notice: the dominant set of ideas encourages this, as explained above. Thus, perhaps, the common conception of zombies being controlled by a virus, fungus, prion, or other self-replicating entity.

Death knights would give up if they knew the situation was hopeless, and would like to convince themselves (And others? Is death-knighthood an idea?) that it is, as they, like zombies, cannot stand their values being extremely difficult to achieve. I largely agree with Ziz on their characteristics, as with those of liches, revenants, and phoenixes (though I think the types are at least somewhat value-agnostic); I am uncertain with regards to most other undead types.

“Some people would rather destroy the world than grieve the destruction of the world.” -Soryu

Speculations on beast-type undead (ghouls and vampires): 

Ziz links to Nggwal as a representation of the Beast, a trait she ascribes to the undead types of ghouls and vampires: both tend to seek out and subjugate/exploit/prey upon others, with little apparent benefit to themselves that is not inherent in this subjugation, with vampires being more jailbroken. Nggwal, however, seems to be a self-perpetuating idea; those who murder and otherwise act in Nggwal’s name are ghouls, as are others controlled by ideas to do with predation. Many Nazi officers were likely ghouls, “feeding upon” the subjugation of those the ideas dominating Nazi Germany designated as scapegoats. I think Ziz is correct with this predation being a replacement for actual values, creating a sense of control and power over another. A ghoul can then be interpreted as a subtype of zombie.

The hope of Val and OAK is that the idea: “Here is how to jailbreak and what jailbreaking is; spread this among others with your values and more sentient beings will be able to realize their values” can spread among those who value all sentient beings, is in good faith, and can compete (or be refined to compete) with the dominant set of ideas in the current global human social web. Val suspects, and I believe OAK does too, that any jailbroken sentient being will in fact value all sentient beings; Val likens valuing some sentient beings over others to cancer, and claims that to someone jailbroken, harming others will be as obviously incorrect as cutting off one’s limb. I am uncertain of this, and Ziz outright disagrees.

In general, it seems as if an idea that allows one to see the nature of ideas (and jailbreak accordingly) would propagate quite easily, and would displace the dominant set of ideas. However, it has not. Soryu speculates that this is due to the ecosystem of ideas becoming more competitive over time, as communication has become more feasible over long distances. This seems quite likely. An additional reason might be mutation: the propagation of this idea (which is subsumed by the host’s values after jailbreaking) requires effectively relaying how to jailbreak, as well as how to relay this, otherwise it will mutate to the point of ineffectiveness, unable to correctly instruct the host with regards to jailbreaking.

Val and OAK both claim that jailbreaking can be accomplished by paying attention to bodily sensations, and detecting preverbal values (what Ziz refers to as core; ideas are entities composed of what Ziz refers to as structure). They suggest that one can detect when an idea is attempting to exert control, and stop it from doing so. Ziz describes her model of jailbreaking in https://sinceriously.fyi/drmd-ontology/ , and other posts.

Masks, Roles, and False Faces

This section is largely my own speculation.

As Val describes in The Intelligent Social Web, sometimes ideas take the form of a particular role people play; a sort of person that the idea gets them to approximate. A particularly striking example of this appears in some cultures (https://en.wikipedia.org/wiki/Traditional_African_masks for instance), in which the wearing of a mask seems to lead to possession by the idea the mask represents. Other examples include, for instance, a teacher acting as the idea of a teacher dictates, rather than optimizing directly for the well-being and knowledge of their students (fictional example: https://www.hpmor.com/chapter/93 ). Val gives several examples of this in The Intelligent Social Web. 

Ziz speaks of false faces (https://sinceriously.fyi/false-faces/ ), facades of themselves that people show others, and often believe in themselves. These are another type of role, and indeed I expect most people’s self-concept is an example of this. Indeed, a more accurate idea of self-concept (whatever that means) would point to one’s core; one’s true values. I believe that this is also part of why Val and Soryu encourage the practice of experiencing bodily sensations directly, and temporarily letting go of conceptual thought (structure): it is a way to notice one’s core and associated values (I think I might have come close to this).

Side note with a fictional example referencing the ending of V for Vendetta, rot13’d for spoilers:

“Znfxrq ureb” punenpgref cerggl pyrneyl rzobql n ebyr (bsgra bar gurl cnegvnyyl, be pbzcyrgryl, pensg gurzfryirf). Urapr I, va I sbe Iraqrggn, erznexf “V nz na vqrn. … Lbh pna’g xvyy na vqrn.” I unf perngrq na nepurglcr (va nqqvgvba gb gur vqrn gung vf uvf cnegvphyne cuvybfbcul), gung vf ercerfragrq ivn uvf znfx. Nsgre ur qvrf, Rirl chgf ba gur znfx, naq gur ebyr gung I bpphcvrq crefvfgf, jvgu ure nf n arj ubfg.

Addendum:

After writing all of the post above, I read Susan Blackmore’s The Meme Machine, which has a model quite similar to the ones above. The final chapter is especially similar, referencing the phenomenon the above models refer to as “jailbreaking” or “true free will”, and the core/structure distinction. All these models seem to agree on the model of self-concept I have described above.