When being a human random
by phil on Sunday Jun 15, 2003 10:22 AM
When being a human random number generator, do you employ a little bit of bias to get more authenticity?
I know I do. If somebody told me that I had to generate numbers randomly between 1 and a 100, I'd probably want to include some of the really high numbers and really low numbers, like, I'd probably throw a 2 and 5 in there along with a 91 or a 92. If you throw in a 99, then it looks like you're fishing. The lower numbers don't seem to get the same mal-treatment, if you said 3, that sounds like a reasonable number for a random number generator to create. 1 and 100 are both offlimits, though, as it's too difficult, or rather, too conveninet that a 1 and/or a 100 were generated. Imagine, you start off your random sequence with "1, 100, 50..." people would think something was up. You have to start with a number that has like both an even number and an odd number in the mid-range. Like, 34, or 67. If you start with like, 11, someone will think, "well, is this generator biased toward early numbers." Now, don't start at 50 either, that's too evenly split. As a responsible random number generator, you have to give expected unexpectdeness. Like, start with like 29. Thirties are kind of like, well, this guy is trying to get close to 50 and away from the low numbers. 29 is safe because it's kind of unstable, kind of out there. Oh yeah, back to 1 and a 100. You have to throw in a 1 or a 100 in there eventually (not both). Otherwise, it looks like you're not covering the whole range of random numbers. You also have to put some expected anomolies, like have a 33 followed by another 33 every now and then. Ironically, another thing that could be done is a random number generator that generated like 3 of the same numbers in a row, like "7, 7, 7" because that would be like, "wow, that's as bitchy as a random number generator comes." If you do "7, 7, 7, 7" then they'll think something is broken. They'll probably give it a 5th chance, and if on the 5th try you produce another number, then you'll be safe from the "it's broken principle." Actually, I think there might be a really high number of times someone will retry a random number generator if it kept producing the same number. I think it depends on the context. If you were a CS professor showing off your program and by the 4th try it was the same number, you'd quickly shelve it. If you were a frustrated programmer, you would probably shelve it after the 3rd. But if you found a claimed-to-be legitamite random number generator, you'd probably keept on trying it, at least laughing at maybe having discovered a problem in it. And then, when after the 10th pull it gives you a new number, you'd start cheering for the wit of the generator.
This is a special section of random number generation theory that I think could be mined and tapped by movie directors in the realm of creating suspense from characters who are supposed to have "free will" and therefore the appearance of a random number generator.