The medieval conception of a text is a unicursal labyrinth: it twists and turns, knotted in its convolutions, obscure a completely linear path. There is only one way to read such a linear text. Hypertexts, on the contrary, are topologically like multicursal labyrinths: there are many possible paths through, some being dead ends, and these paths fork at nodal points; they may even loop back upon themselves. While a multicursal labyrinth may only be traversed linearly — while the traces of its lines of flight are linear — its potential traversal paths, or its solution space, is both parallel, myriad, and potentially cyclic. A literary conception of chapel perilous must have these characteristics but further be undirected: chapel perilous is a labyrinth from which there may be no escape. The choice of paths here relates to conceptions of will. Chapel perilous is a place where traditional maze-solving algorithms do not work. You cannot stick to the left wall, for instance. You cannot get a clue. Therefore, it cannot be navigated without a ‘will’. The idea of will is problematic here: Gurdjieff’s idea of will seems stronger and stricter than Spare’s: we can create sigils without having a single crystallized self! And it seems incompatible with Crowley’s ‘true will’, a determinist formation that is guaranteed to be in agreement with the flow of the universe. Where Crowley’s true will is the Tao, obscured by the twists and turns and knots in the veil of maya, Gurdjeiff believes we can only develop a will through a promethean transcendence: to evolve is to declare war on God, Nature, and the Moon, and this revolt against GNON will be put down if it results in large-scale disruption. Crowley appears, despite his promethean and luciferian trappings, conservative in this sense. Gurdjeiff seems more in line then with Nietzsche, whose will to power ends with a superman who creates new sacred games for himself, or with Sartre, whose acts of radical freedom cannot be predicted. But this is not really true either! Gurdjieff’s will is a crystallized I, and is supposed to be at least internally coherent. Sartre’s revolts against the self are necessarily random, which Gurdjieff would consider the absence of a unified will. Yet Gurdjieff sees a unified will as nonmechanical — in fact, the only possible aspect of a human being that is nonmechanical. How is an internal consistency different from a mechanism? How is freedom of choice different from randomness? Here we may look to Dennett’s idea of the agentitive lens. We treat or model phenomena as agents — as making choices based on free will in a goal-directed manner — if and only if doing so makes it easier to predict behaviors. This relates to scope, and to levels of abstraction. If we have all the relevant information to model something step by step, as we might with simple Newtonian mechanics like billiard balls, then the agentitive lens is inappropriate. If we have all the appropriate information but the complexity of such a model makes it impossible to efficiently compute results, as with the behavior of gasses, we might use statistical models. If the results are highly input-dependent, as with weather patterns, we must use statistical models and accept less accurate results. But no matter the underlying complexity, if a phenomenon has both positive and negative feedback loops and those loops are generally balanced, we can treat the homeostatic or allostatic state as a goal and model the phenomenon agentitively. As we build the master of the house, the master can recognize the mechanicalness of the individual centers and components because he can predict them without necessarily resorting to agentitive modelling — but the master is an agent to any systems of comparable complexity. Meanwhile, lower and less integrated forms of intelligence cannot model the master at all, except perhaps as an agent of much higher effectiveness, and then only if its goals are known. Constructing a higher integration, and making sure that integration is stable and functional, involves radical restructuring of existing behaviors and mechanism. We are not hill-climbing but performing simulated annealing. A shock to the system, or a constraint that changes the fitness function and thereby changes our path through solution space, makes us more flexible but may cause us to simply stratify in a different manner. At the same time, constant mutation means we are always acting suboptimally. This is the explore versus exploit tradeoff. We cannot always afford to keep going until we hit a dead end. Sometimes we must turn around when we hit fnord or some other clue that we might be losing the plot. But such clues are not reliable. A fnord is irrationally feared. If we cannot see the fnord it cannot eat us, but the fnord might not eat us anyhow. Sometimes red flags are a real, reliable clue that something is wrong. Other times, red flags are false flags: to get to our goal we must ignore them. Goals that are worth having are goals that are rarely achieved, and they are therefore goals that are hidden behind barriers. If you follow common sense then you are doomed to find yourself with a common life and a common mentality. Injecting entropy — through randomness perhaps — can keep us from being predictable to others and to our past selves, even through the agentitive lens. At the same time, it leads to suboptimality in the exploit phase. How can we be sure that we have found a high enough peak before settling? And how long should we settle?

Written by

Resident hypertext crank. Author of Big and Small Computing: Trajectories for the Future of Software.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store