The silver rule : (there are other versions that are incorrect) is:
"Do unto others as you would have do unto you, but repay in kind"
In other words, be nice, but if someone smacks you on the cheek, wack 'em back good and solid.
The premise of the model was ape or monkey grooming behavior, but it would apply to any altruistic social behavior that bestows an adaptive advantage. Basically, there were two types of apes:
1) Those that always groomed those who asked (golden rule apes)
2) Those that never groomed other apes even if they asked (evil apes)
The computer model showed that given this situation, the golden rule apes became extinct and all you had were evil apes. You can see how this works, if P_gr(t) = P_gr(0)* exp(k_gr*t) and P_ea(t) = P_ea(0) * exp(k_ea*t), the growth constant of the evil apes, k_ea would always be greater than k_gr, as the evil apes can spend more time finding food, mating and procreating than the golden rule apes. Once the GR apes.
They introduced a new model of ape, the silver rule ape:
3) Silver rule apes groom anyone ONCE, but after they groomed an individual, they would never groom back until that individual groomed them back.
Adding just a few "silver rule" apes would result in the evil apes becoming extinct. The evil apes lost the adaptive advantage as they were no longer being groomed as often.