1. Reproduction according to fitness:  Select strings from the current population (this might be the set of rules for the agent) to act as parents.  The more fit the string (the stronger the rule), the more likely it is to be chosen as a parent.  A given string of high fitness may be a parent several times over.
Under step 1, (reproduction according to fitness), above average building blocks (schemata) are used more frequently (and prosper) and below average building blocks are used less frequently (and disappear).
>Crossover can recombine schemata without greatly disturbing the desirable outcome of (Reproduction according to fitness) step 1.