Since propositions are represented by terms in the SNePS logic (Shapiro 1993; Shapiro 2000b), metabeliefs--which are propositions about propositions--are easily represented without leaving first-order logic.
SNePS acts may be categorized on two independent dimensions: an act may be either an external, a mental, or a control act; and an act may be either a primitive, a defined, or a composite act.
SNePS actions and, by extension, acts, may be subclassified as either external, mental, or control.
Control acts are the control structures of the SNePS acting system.
SNePS actions and acts may also be classified as either primitive, defined, or composite.
If a SNePS agent is to perform a defined act a, it deduces plans p for which it believes the proposition ActPlan(a, p), and performs a do-one of them.
In this section, we discuss five example SNePS projects that incorporate metacognition: self-awareness; lack-of-knowledge acting; consistency-maintenance and optimization; contextual vocabulary acquisition; and mathematical problem-solving.
There are two aspects to our approach to self-modeling: representation techniques in SNePS at the KL; and PML-KL interaction.
The optional default act in the SNePS control acts withall and withsome provide for acting on the basis of lack of knowledge (see Moore ).
Figure 3 demonstrates a SNePS agent solving this problem.
SNePS uses a monotonic logic, so contradiction resolution requires that at least one base (input) belief that underlies a contradiction must be retracted to eliminate that contradiction.
Belief change is managed in SNePS by SNeBR, the SNePS belief revision system (Martins and Shapiro 1988, Shapiro 2000b).