Tuesday, March 21, 2006

Decision-making made easy

Well, sort of.

A complex decision is a whole bunch of trade-offs, profit-and-loss variables. Each variable has a probability associated with it, and they can cascade together. I use a system of "expected value" summations, and it works pretty well.

For instance, in buying a car there is the price (and the 100% likelihood that you'll have to pay it), a set of features, and a set of unknown costs (maintenance), and a set of emotional value points (prestige, convenience, dependability). Each of the costs has a probability that you'll incur it, and each of the values has a probability that you'll receive it. Some of them are related, and may need to be refactored to make the math work out for you.

You multiply each of the costs and outcomes (positive and negative) with their value to you (on some scale of your choosing) and their probability of occurring, and sum them all up. That choice gets a score.

Compare the score from all of the other choices you could make, and your decision is made.

The nice thing about this system is that by breaking down the fuzzy-factor "value" for each outcome and pairing it with a probability, you see the real cost for each while simultaneously hiding the answer from yourself. Subconciously you will tend to favor the choice you want to make, but be careful that you don't fudge the probabilities.

As a simple example, consider recreational sky-diving. The value you get from jumping -- a rush, some prestige, and maybe some sex out of it somehow -- compares with a (call it) 99% probability of landing safely and a (call it) 1% probability of landing with a splat.

For me, I assign a pretty high value to keeping my skin intact. How much would I pay someone not to flatten my skull?

stay on ground = free + 0 (death from falling) + 0 (fun)
= 0
skydiving = -$50 +
.01 (death from falling) + .99 (fun)
= -$50 - 1/100 (very big number) + .99 (small number)
= (probably something negative, and I have to pay 50 bucks).

As a side note, you can see that the resultant costs of a decision and the cost to make it happen are just two labels for the same thing. That is, whether something is a cost or benefit is just the sign on the term.

No comments: