So, reading this thread about randomness and computers, the title question suddenly popped into my head. It seems like it would have profound consequences for the models of our cognitive processes if we either can or can’t create randomness, however, it also seems difficult to test for – since random processes don’t have memory, and we do, you’d probably expect people to repeat the same number twice far less often than that would happen in a random string, or other effects in that vein.
Well, not an answer, but I would throw in a little skepticism about the idea that humans can generate random numbers.
Clearly, most people, most of the time, cannot. There is a famous trick where you ask someone to think of a number between 1 and 50 where the digits are not the same and both digits are odd.
The answers you get are not uniformly distributed and the most common response is 37.
And this is just picking the first number, when asking for a sequence of numbers there is quite often a pattern of going higher, lower and sticking at the same number that is often quite predictable. The same psychologists / magicians that do the 37 trick, do other tricks where they guess what number a person will think of next in a sequence, and they claim to be using similar principles to do this trick.
But can a human in principle think of a random sequence? I guess that would depend on whether there was a random physical phenomenon that could influence our thoughts. I guess even if our brains don’t contain any randomness to any significant degree, we could still come up with random numbers via sensory information of a random phenomenon.
I doubt that this is in the spirit of the OP though.
I don’t know of any studies that have looked at this, but I am personally a sceptic. One of the things you will notice when you look at a random collection of numbers is how often the same numbers follow each other (the same with roulette in the casino). If you would ask a person to make a list they will most probably take the previous number into account too much. If you had a ten sided dice (if such a thing exists) however, sure we can
Pssh. Greatswords only do 1d10 if you’re Small. I ain’t afraid of no stinking halflings.
Why is 37 the magic number? It seems most natural to start with the first odd number (1) and then find the next (3) and slap 'em together. Am I missing something here?
Sorry, after clearing my throat, I forgot what I was going to say. Assume it was something very friendly and not at all a joke about how many humans have accidentally stepped upon and crushed members of your shorter race.
I would guess because it breaks down the question into finding two odd number from 1 to 10.
Nobody picks one, it’s just to easy. And five seems too common, and too round of a number. that leaves 3 7 and 9, and 3 and 7 are the first two of the ‘good odd numbers’
The effect is even more obvious with a four-sided die. (I know. I know. It’s just a metaphor.)
In 1985 I took a college course where the prof typically gave us multiple-choice exams. I quickly noticed that there were times when one choice, say, “C” was the correct choice for four or even five consecutive questions. You never see runs like this in multiple-choice exams.
Well, the good professor had mentioned that he used his new-fangled microcomputer to make up his tests. From my own experience with a CoCo2, I had seen this same phenomenon when a computer is asked to “randomly” pick from a limited set. So I asked him if he had had his computer pick which letter would represent the correct answer for any particular question. Bingo. He had. I wondered how many of my classmates had been freaked out by the long strings of identical answer letters, because you just don’t see that when someone picks the letters “manually”. I believe that is what they call pseudorandom - it’s sort of random, but outside considerations influence the result.
That’s not quite what “pseudorandom” means, at least it’s not what it means in the computer science world. Pseudorandom basically means something that is statistically random, but is actually generated in a way that’s deterministic (that is, totally predictable if you knew what was going on behind the scenes.)
E.g., when you ask a computer program to generate a “random number”, it’s probably actually generating a pseudorandom number. Basically, the computer has a big list of “random numbers” saved inside itself somewhere. These numbers are chosen so that if you do statistics on them they will actually appear random. That is, each number occurs the right number of times to be random, and there are repeating strings of number as often as would occur with true randomness, etc.
But really, the computer’s “random number” isn’t random, it’s just the next number on the list. So if you knew the list of numbers and knew where the computer was starting from on the list, you’d know what number it was going to pick in advance.
To add an element of randomness, programmers generally have the computer pick its starting point on the list based on the computer’s internal timer. Since you don’t know precisely when the program will be run, the number is genuinely unknown ahead of time.
I see the same thing in beading. I tend to use a lot of mixed beads, and when I want to make a “random” assortment on a string, it takes a ton of time and effort to get it to look random. Just spill 'em out of a tube and you’ll get six greens, a purple, three more greens and two blues. That’s too many greens in a row to look “random”, even though, strictly speaking, it is.
My above explanation might raise the question: If the computer is just picking from a list, how did the list makers know if you’d want a random number between 1 and 100 or between 1 and a billion?
The answer is that generally the random number function gives you a random number between 0 and 1 (e.g. 0.34231274), and you as the programmer have to multiply this number by, say, a billion if you want that to be the upper bound of your random numbers. Then you just cut off the decimal (assuming you want an integer number).
IIRC, if you have a person typing in numbers at random, it’s fairly easy to learn to predict the next number with pretty good accuracy.
All the random number generators I know of produce integers in the range between 0 and the largest positive integer that can be represented. Some languages have built-in random number functions that scale the number to be in [0, 1], but that’s not what’s being generated.
