Normally, in the process of making a decision, a decision-maker will know at least everything known for earlier decisions, as well as what those earlier decisions were.
That can be represented in the decision net by including links to a decision node from all earlier decision nodes, and from all their parent nodes. These links are called no-forgetting links, since they indicate that all the information the decision maker knew earlier, he still knows when making later decisions. Keep in mind that links entering a decision node are informational links, and indicate what the decision maker will know when he makes the decision.
Having more information to make decisions always leads to a policy having greater expected utility, if that information is relevant. Many of the no-forgetting links may not be relevant; they provide information about the past that does not help in the new decision. If these links are included, the optimal decision functions will simply ignore them, and the expected value of the policy will be the same, but they may result in very large relation tables for the later decisions.
When constructing a decision net, you do not need to concern yourself with no-forgetting links, as long as you put enough to create a directed path running through all the decision nodes (to indicate their time sequence). When Netica solves the decision net it will remove irrelevant no-forgetting links, and add any relevant ones that are not present, based on the structure of the net.
If you wish to
see all the no-forgetting links implied by a decision net, make sure there
is a directed
path running through all the decision nodes and choose Modify →
Add No-Forgetting Links.
This is useful when solving the decision net using the alternate
method of node
absorption (instead of compiling).