“Predicate” circumscription will not result in any new positive instances of P …or for that matter of any other predicates. ? sometimes predicate circumscription doesn’t give us what we wantE.g., KB= (block(x) ? ? ab(x) ? ontable(x)) ? ?ontable(B1) ? block(B1) ? block(B2) ? B1 ? B2Task: Want to infer ontable(B2), but currently, KB ?? ontable(B2) and KB ?? ? ontable(B2)Let’s see what happens if we compute CIRC(KB; ab)KB solitary in ab ? circumscription is equivalent to predicate completion Rewrite block(x) ? ? ab(x) ? ontable(x) as block(x) ? ? ontable(x) ? ab(x) CIRC(KB;ab) ? (COMP;P) ? block(x) ? ? ontable(x) ? ab(x) ? ?ontable(B1) ? block(B1) ? block(B2) ? B1 ? B2Failure: (CIRC;ab) ?? ontable(B2) and (CIRC;ab) ?? ? ontable(B2) (CIRC;ab) cannot force ontable(B2) -- it’s new positive information.
Limitation of “Predicate” Circumscription