> (2) The essence of credit assignment is to provide the system with hypotheses that anticipate future consequences, strengthening rules that set the stage for later, overtly rewarding activities. 
>For cas this process leads to a question we have not really explored so far.  Just what is it that should be considered rewarding?  
>Generally, evolution has built in certain internal detectors that record the status of “reservoirs” of food, water, sex, and the like.  The organism’s behavior is largely directed at keeping these detectors from “empty”.
>For more sophisticated organisms, much stage setting and anticipation goes into this task.  It is a kind of never-ending game with intermittent payoffs.  The value of any behavior depends on the current position in the game and the status of the reservoir.
>Said another way, figures of merit for cas are usually implicitly defined.  Competition, with local payments, is one of the few techniques we have for handling such problems in distributed systems.
>Competition is the basis of the credit assignment technique used to describe this aspect of adaptive agents.