Brief notes on skeptic thought

Skepticism and faith

Exact model

This is for problem domains that are well-understood and small enough that an exact modelling is possible. Small state machines, regular expressions, (most) program configuration, networks with up to 20-30 routers and not too many route-maps. This means there is no guessing, there is no hand-waving and there's no defocussing from details that might, possibly, be essential.

Exact modelling has its limits, it quickly grows to a point where exact modelling and precise reasoning isn't possible over the whole of the model.

Statistical model

This is an aggregation and generaliseation of multiple exact models, giving one the option to dive down to exact modelling in the spots needed. Typically useful for HA-clusters, clustered mail platforms and other situations where complex processes interact. It's allegedly also useful for deriving the gas laws and other feats of statistical mechanics.

Psychological model

This is a useful imprecise modelling of something that's not well- understood enough to be modelable exactly or statistically. It's the method of ascribing a psyche to a process and model wants and wishes. It's not always useful, but often can manage decent predictions. It can open a path for a more exact understanding, but in and of itself won't give you the power to model the process exactly.

It's a modelling level that is useful for understanding people or model an economy (though some economy modelling fall in the "magical" stratum).

Magical model

This is a useful model for something that isn't well-understood enough to even have a psychological model. A typical problem space where this occurs frequently is in computer hardware maintenance. There's a couple of things you do by rote. No one (NB: hyperbole) can explain why these things are done, but they work and so get handed from practicioner to practicioner. Typical cases would be the ritual taking-apart-and-put- together that resolves quite a few issues.

This is more an aggregation of isolated facts than an actual model. It has (almost) no predictive power but can function as a placeholder for facts until there's a large-enough fact corpus to elevate the model to a psychological model.

So, faith, where does it land on this scale? It would, I claim, be in the "magical model", occasionally straining to find itself in the psychological model statum. I'd also say that "Blind Faith" is the struggle to keep one's modelling to stay in the magical model stratum, for whatever reason (possibly because elevating it to a more precise modelling means risking losing one's faith, this may well be a traumatic experience).

Using these definitions, it's possible to see that a self-confessed skeptic can have (some) faith while being skeptical and wanting to model other things as exact as is humanly possible. Only someone who strives to move all one's modelling from the "exact" end of the spectrum to the "magical" end is a definite non-skeptic.

This is one of Ingvar's essays