Dependent Fluents (1995)
Enrico Giunchiglia and Vladimir Lifschitz
We discuss the persistence of the indirect effects of an action---the question when such effects are subject to the commonsense law of inertia, and how to describe their evolution in the cases when inertia does not apply. Our model of nonpersistent effects involves the assumption that the value of the fluent in question is determined by the values of other fluents, although the dependency may be partially or completely unknown. This view leads us to a new high-level action language ARD (for Actions, Ramifications and Dependencies) that is capable of describing both persistent and nonpersistent effects. Unlike the action languages introduced in the past, ARD is "non-Markovian," in the sense that the evolution of the fluents described in this language may depend on their history, and not only on their current values.
In Proceedings of International Joint Conference on Artificial Intelligence (IJCAI), pp. 1964-1969 1995.

Vladimir Lifschitz Faculty vl [at] cs utexas edu