I've decided to make my first real post on this blog on the subject of rational thought. On SL4 there has been a recent flurry of activity on the subject of rational thought, what it means to be rational, and whether there exist arational/irrational sources of evidence.

The nature of this discussion leads to very high tempers, so I'm going try a different tack. I'm going to use the microuniverse of MineSweeper. Now, MineSweeper is a popular computer game(though not nearly as popular as Solitaire, unfortunately) largely because it comes bundled with lots of Operating Systems. Every Window's package I've ever had the privilige to attempt to use came with one; The Red Hat/KDE package I'm using now came with an implementation; OSX has it's own, IIRC; if you're really without minesweeper, you poor baby, get it here. this is a java game, but it works pretty well.

Minesweeper is an entirely regular and understandable game with completely understandable underlying rules. What concerns me in this discussion, is the way people play and understand it. The information provided to you in a standard minesweeper game is sufficient to solve it, given that you survive long enough to gather some lower bound of information. In fact, most minesweeper games appear to contain the most risk in the first 5 clicks or so. However, the important losses occur after the first 5 clicks, when the game has progressed, yet the player still loses, in spite of, or perhaps because of the information s/he has.

In order to solve this mystery, play a couple of games of minesweeper. Now, when you were playing, after a while, were you absolutely sure you were right, when you clicked on a mine and ended your game?(this paragraph assumes you lost at least once, if you did not, you are either a rational reasoner, or you've already mentally solved minesweeper; go away) . Now, this uncertainty is because you aren't playing Minesweeper as an entirely rational process. I'll explain.

Minesweeper's simple rules lead to interactions, which lead to the numbers and blanks you reveal by clicking. Analysing those numbers and blanks, within that context, gives you information. A rational process is one that takes information about the world(in this case, minesweeper) and assigns it strengths and priorities in keeping with actual facts. ((ex: a 1 implies there is only one mine within 1 block, that's a certainty; you've never seen a mine be in the upper left hand corner, that's fantastically unlikely to be right, and doesn't include a reason why there wouldnt' be, so there is no chance that it's right for the right reasons; ergo, you can rely on the former, but the latter is a fools bargain)). Rational processes are not wrong. They are in line with reality, and thus the world(minesweeper). If you recieve a wrong result, either your process is not rational, you executed your process incorrectly, or you applied your process inappropriately. Irrational Processes, like praying for a blank block, selecting the four corners, blind guessing, or always selecting to the left may in fact succeed occasionally, and may out of statistical anomaly even make more than 50% success. But you are not making the right block choices for real reasons. Irrational actions don't take information, make justified conclusions from that, and generalize, as rational processes do, they either act on information that is invented, percieved(but not extant) patterns, or simple whim.

Only rational processes can allow one to win a game of minesweeper reliably, because rational processes are in line 100% with the larger universe. Minesweeper is very simple, and it's easy to see what processes are rational or not, because irrational ones lead to ended games.

Generalize to the Real World. Suddenly, the nature of the argument changes. What constitutes a justifiable conclusion from evidence? What's really evidence? What about people who seem to have powers or knowledges from elsewhere?(supernaturalism) It seems like the world is too much to deal with 100% certainty.

The real answer is that it's not actually any more complicated to make individual decisions rationally. Lets look at what carries over from Minesweeper.

1. Only one real world. There is only one minesweeper board, and wishing for another arrangement won't change where the mines are. The only way you get anywhere is by abiding by that arrangement, and reacting to it. The world is the same. There is only one physical world you live in. The people next door live in the same world. Soliphists(people who believe the universe to be imaginary) still get killed by trucks. And no amount of wishing or hoping, or thinking seems to change anything about the world.

2. The world is consistent. The relationships within minesweeper, the way numbers translate into mines and blocks, never changes. Next game, the mines don't give different results, and patterns within one game that are applicable, still apply within the next. The world is the same. Gravity remains constant, fields interact the same way, day to day. There is no such thing as a physics of epochs. Time periods always were and always will be subject to the same basic rules. That said, somethings are very variable. Social structures, human relationships, these are high-level dependent structures that change through time, just as the individual positions of mines between games. But there is an underlying consistency that continues throughout.

3. Truth is self-supporting. If something works within minesweeper, it works for a reason. There are no supernatural elements. The real world has phenomena, and all such phenomena, and all objects within the real world, follow rules. If something is true, the surrounding environment reflects this. True facts is the only supportable objects, because all other facts support them.

Rational thought is not a force, it's not an object, and it's not a person. It is a special kind of relationship, between evidence, behavior, and conclusion. A formal exegesis of this relationship exists, it's called the Bayesian Probability Theorem. Basically, this theorem describes how you relate different kinds of evidence into a cohesive conclusion, that remains inline with the universe. (if you are not familiar with the theorem, start here). My friend Eliezer Yudkowsky is what you would formally call a Bayesian Rationalist.(he considers the Bayesian Evidencial process superior to the more traditional Popperian approach). But I would call him a BPT enthusiast. He invokes the BPT at almost every opportunity, citing it as the basis of rationality. As a detail-oriented person, I must disagree. The Bayesian Probability Theorem is not the well from which rationality springs, nor is it the basis of all rational thought. It is a formalization of the relationship that Evidence has to Conclusion within said rational thought.

Formalism aside, Rational conclusions are the only conclusions worth making, because they're the conclusions that work the most. If you aren't using your best judgement, you may as well give up. Otherwise your making the wrong decisions for the wrong reasons, even if you occasionally get lucky. And as my friend Eliezer says. "That's worse than nothing."