Cerbereoboros by Emily Fundis*

My favorite game ever is One Night Ultimate WerewolfIn it, everyone draws a card that specifies whether they are a werewolf or a villager. Then everyone closes their eyes. The two werewolves open their eyes so they can identify each other. After that, there are some villagers who can complicate the game in other ways- for instance, the troublemaker, who can swap two other players' roles without them noticing. Then everyone opens their eyes. Everyone yells at each other for a bit about who is or is not the werewolf, and then they vote on who to kill. If a werewolf is killed, the villagers win. Otherwise, the werewolves win.

The only concrete components defined by the game are your goals and a few bits of sparse initial information, private to each player. Since everything else is open, everything becomes part of the game. There are no turns (unlike in poker), so you are always playing. Every decision to speak or not speak, every flick of your eye, every bit of wording is information that can be used against you. Unlike Werewolf, the game from which it is derived, there is no source of objective, public information whatsoever. In that version, the werewolves kill someone each round and everyone knows who got killed. In this version, nobody gets killed and there is only one round. There is no recourse to objective knowledge as defined by the game (again, unlike in poker, where some cards are shared on the table). All facts established in the course of a round are social facts- even your own role. Whether you win is based on the card that you end up with, not the card that you were originally dealt, so you can never be quite sure what strategy you should be playing.

But by far the most interesting mechanic is the tanner card. The tanner isn't on the side of the villagers or the werewolves. Tanning is an awful, horrible, smelly job- so awful that the tanner just wants to die. His win condition is to be killed. So if you are the tanner, you want the villagers to believe you're a werewolf, and you start acting as if you were a werewolf. But in that case, seeming like a werewolf is a sign that you are the tanner and not a werewolf, so werewolves in threat of being outed can act "like a werewolf" so people will think they're the tanner. Any strategy that anyone has used in a previous round is a possible strategy for someone to use to pretend they are someone they are not. After playing more than a few rounds with clever individuals, you become completely unmoored, never knowing what level you're on. Are they being sincere? Are they pretending to be sincere? Are they acting like they're pretending to be sincere?

The tanner is an element that prevents gameplay from ever becoming static. There is no solution to playing One Night Ultimate Werewolf in the way that there is a solution to tic tac toe. There is no definable strategy that will lead you to win. That by itself isn't saying much- any well-designed game should be open in this respect. Every game with a sufficiently large strategy space and sufficiently sophisticated players has a meta-game. In Magic: The Gathering, for instance, one of the most important components of tournament play is constructing your deck based on what decks you think other people will bring. But One Night Ultimate Werewolf distills this into pure form. The game is the meta-game is the meta-meta-game and so on and so on. Every level is collapsed into one.

While it is possible to get better at One Night Ultimate Werewolf, pure skill does not help you the same way it does in games like chess. You need to gain information about your opponents by playing with them- you need to develop an organic fit within a particular group. But you can't do that without simultaneously giving information about yourself. To make yourself better at beating your opponents, you necessarily make your opponents better at beating you.


Now consider the acquired immune system. Vertebrate animals have two basic systems for disease prevention: the innate and the acquired immune system. The innate immune system consists of simple barriers to entry (i.e. your skin) and a collection of molecules that can identify and respond to molecules known to be associated with pathogens. The acquired immune system is the fancier, more evolutionarily novel system- the one capable of being able to identify new pathogens and build novel weapons against them, and then store the weapons plans for future use in case of another attack. Scientists have typically taken the view that the acquired immune system is the current optimal solution to the problem of defending against invaders, citing how necessary it is for survival in vertebrates. But there's a problem with this perspective. If the acquired immune system were really a dominantly better strategy, then we would expect invertebrates to be dying of disease all the time, while vertebrates had practically no disease. This isn't the case- death rates from infectious disease are pretty comparable between vertebrates and invertebrates. Stephen Hedrick explains this in his great paper "The Acquired Immune System: A Vantage From Beneath":
"Another way of looking at this is that acquired immunity was not a final solution to the problem of parasitism. There is no final solution. As novel as the acquired immune system was, for rapidly multiplying agents, it was just another hurdle. It may have driven parasites to invent new strategies for fitness, but it did not convey invincibility or anything like it. To say the combination of innate and acquired immunity is the optimal defense is a misunderstanding of the evolutionary landscape. I don't believe there is an optimal defense. I don't believe there is a conceivable immune system that could not be obviated once the barriers to infection have been breached. For all animals and their parasites, generation upon generation, it has been evolutionary thrust and parry, until today as it was a million years ago and as it will be a million years hence, each and every species is literally plagued by parasitic microbial agents and viruses."
The term "landscape" as a metaphor for the fitness space of possible strategies is misleading- a series of gravitational pulls on objects chaotically orbiting each other is a better view of the forces involved. Everything is constantly falling down fitness gradients, but it never reaches a stable ground state because there is no ground. In the case of physical systems, gravitational forces are exerted by other massive objects that are themselves in motion. In the case of biological systems, selective forces are exerted by other organisms that are themselves evolving.

It's funny to note that the acquired immune system is the component that opens the way for autoimmune diseases. Unlike the innate immune system, which only reacts to specific markers that are only present in pathogens, the acquired immune system is capable of reacting to any marker. It's capable of going on witch hunts against innocent cells- or werewolf hunts, as it were. There's even a virus that exploits this system like the tanner card in One Night Ultimate Werewolf- HIV. It infects helper T-cells in the acquired immune system by binding to CD4 receptors: the very receptors that those helpers use to identify pathogens. In a sense, HIV wins by being caught.


The central motif here is captured by the idea of anti-induction. Inductive systems are ones where something working now means that it will continue to work in the future- they are systems admitting the possibility of static solutions. Anti-inductive systems are the opposite- if a strategy works now, you can predict that soon it will no longer work. The stock market is the canonical example of an anti-inductive system: if real estate prices have been rising, then it's probable that they will fall. If one company has a really successful business strategy, it usually means that it's bad for you to copy that strategy, as they've already filled that market niche. One Night Ultimate Werewolf and the immune system are also great examples. We see this sort of behavior in any system where there are interacting agents with comparable skill levels competing for a limited resource.

The caveat of comparable skill levels is important. If one agent in a competition is dominantly better than the others, then it will simply crush them all and the game is over. In the chaotic orbits analogy, this would correspond to one extremely massive object with other objects in uniform orbit around it. There is a sufficiently stable configuration such that talking about an energy "landscape" makes sense. But if you have a bunch of objects of approximately the same mass, their collective trajectories are extremely unpredictable.

But why should we expect to see any competitions where there are well-matched competitors in the real world? In game design, it requires a tremendous amount of skill and fine-tuning to make sure that gameplay is balanced. So why would we expect balance to exist in nature, in the absence of an agent maintaining it? The reason is that the meta-system in which games evolve- that is, the universe- is itself anti-inductive. The world does not simply stop moving when a particular competition is solved. Any time a game is won, the aftermath of that game becomes fuel for the next game. The system learns, the world learns, and it moves on to the next level of play. If one company dominates a market, the competition simply moves on to a different market. If one organism manages to drive a parasite extinct, a new one pops up to replace it. If one struggle over social norms has been resolved, the conversation shifts to the next most divisive topic.

The chaotic orbits analogy helps us here as well if we consider electromagnetic forces in addition to gravitational ones. Free protons and electrons group together to become hydrogen. Then that hydrogen collects in stars and the collective gravitational energy combines the hydrogen into heavier elements. Then chunks of those heavier elements coalesce into planets. Then some of those elements form replicators, which form cells, which combine to form multicellular organisms, which team up to form social groups, and so on up the ever-expanding chain of competition. In each case we have more complicated patterns built on top of underlying regularities. This process is so ubiquitous that one wonders whether our "fundamental" physical laws are themselves merely regularities in deeper competitive structures.

* As far as I can tell the artist did not actually give this piece a name, but I found it elsewhere under the name "Cerberoboros"- a portmanteau of "cerberus" and "ouroboros"- and decided that it was too good of a title not to use. Another example of lexarchy in action.