|
|
|
|
|
|
|
Fig. 16.
The bucket brigade algorithm |
|
|
|
|
|
|
|
|
Its strength increases because its income exceeds its payoutit makes a "profit." Subsequently, the suppliers of the suppliers begin to benefit, and so on, back to the early stage-setting rules. |
|
|
|
|
|
|
|
|
Things can go wrong. A supplier might, through a message to the effectors, convert an environmental state to one that diverts its consumer rule from a payoff-directed path. That is, it might fail in its stage-setting role. In that case, the consumer suffers because the diversion will prevent it from receiving payments from its consumers; however, the diverting supplier rule generally suffers even more, because it is at an earlier stage in its "getting rich" effort. Or it may be that the consumer has a condition that attends to the state of the enviroment and does not even bid when the diverting state occurs. In either case, the diverting supplier soon loses enough strength so that it no longer wins the competition. It then ceases to influence subsequent activity. |
|
|
|
|
|
|
|
|
The whole process, of course, takes repeated plays of the game. But it only requires that a rule interact with its immediate suppliers and consumers. It requires no overt memory of the long and complicated sequences leading to payoff. Avoiding extensive overt memories is almost a sine qua non for large, parallel systems acting |
|
|
|
|
|
