This is a matter of technique rather than a requirement. We provide a random number function, but we strongly advise storybuilders to refrain from using it, because it introduces problems in testing. When you use a random number, you have no idea what it is and you won't get the same result the next time. Thus, if you run a test and get odd results, then run the same test again and get normal results, you don't know whether the oddity connotes a flaw in your scripts or just a random number at an extreme value. And you can't repeat the test.
Nevertheless, there remains an occasional need for random behavior, and we recommend the use of pseudo-random factors in such cases. For example, suppose that you desire a pseudo-random number between 0 and 9. You could easily get this with the following term:
alpha <= (Gullibility[Actor] modulo 10)
This function, modulo, divides the first number by the second number and returns the remainder. Thus, (87 modulo 10) gives a result of 7; (53 modulo 10) gives a result of 3.
You can use almost anything in place of Gullibility in the above formula with little fear of unreliable results. The only danger comes when you use a variable that is normally fairly small, say, less than 10. Then you'll get a bias built into your formulas. For example, suppose that you have a custom trait that represents which fraternal club your male characters belong to: 1 means Lions Club, 2 means Optimists Club, 3 means Elks Club, and so forth. If you used this variable for pseudo-random numbers, you would have a built-in bias based on the club to which a person belongs. You should always use variables that are spread out over the full numerical range of 0 to 100.
And what modulo value should you use? I recommend 10 as the ideal value to use, for two reasons: first, you don't have to be a math whiz to calculate the results in your head when you double-check actual numbers &emdash; just grab the low digit. Second, 10 is very small relative to the 100-range of most variables used in the Erasmotron, so there's little chance of problems with bias. If you were to use a modulo value of, say, 100, then you'd have built-in bias in every case, and no pseudo-randomness, because something like (Gullibility modulo 100) will always be exactly the same as Gullibility.
But this is a technique, not a rule; if you understand the math of what you're doing, you can experiment freely. You can use these rules of thumb with confidence.