Consider a binary classification problem. In the Pittsburg approach each member of the population represents a set of rules. Each rule encodes information regarding the data features (that is, the conditions of the if rule) and information about the class (that is, the conclusion or then part in the if rule).
Then in principle the same member can learn a couple or more rules that are in contradiction for the same example from the data. For instance, for a certain member that has learnt Rule 1 and Rule 2 the following may happen: Rule 1 would classify an example as belonging to class A, and rule 2 would classify the same example as belonging the class B.
How does a member reach its final decision (class A or B) when it can have rules reaching different conclusions?