Subgoal
From the SL4 Lexicon:
subgoal:
Servant goal created for a supergoal, based on a probabilistic confidence estimate that the creation of the subgoal will lead to greater fulfillment of the supergoal. Many subgoals will take the form of heuristics, rules of thumb for achieving goals in the real world. Some subgoals might not appear like subtasks of the supergoal at first glance, rather contributing something to the entire network of subgoals, serving the supergoal in that way. Some subgoals may be convergent, present across many or all classes of theoretically possible minds, such as "obtain a power source" or "gather perceptual information". "Protect myself" may be a Convergent Subgoal, but by no means is it universal. (We can imagine classes of AI systems whose behavior is basically selfless.) See also Convergent Subgoal.
An event or world-state whose desirability is contingent on its predicted outcome. An action taken in order to achieve some other goal. A thing not desirable in itself, but rather desirable for some other end.
If the "ParentGoal" is a wet sponge, then placing a dry sponge in water is a "child goal" of getting a wet sponge. The parent goal, in turn, is probably a child goal of something else. "Parent goal" and "child goal" describes the relation between two individual goals, while "SuperGoal" and "SubGoal" describes the general distinction between things intrinsically desirable, and things which are desirable only because they are predicted to lead to other things.
It is very important to distinguish between SuperGoals and SubGoals in Friendly AI, since SubGoals are directly and entirely dependent on the current world-state.
(quoted from CFAI's glossary (see CFAI) )
