Often KBs contain incomplete knowledge E.g., Airline database flight(NYC,Boston,TWA) flight(NYC,Chicago, United) Does not include all the non-existent flights ?flight(NYC,Kalamazoo,TWA) ... This can lead to the correct model as well as multiple unintended models flight(NYC,Boston,TWA) flight(NYC,Boston,TWA) flight(NYC,Boston,TWA) flight(NYC,Chicago, United) flight(NYC,Chicago, United) ? ? ? flight(NYC,Chicago, United) ?flight(NYC,Kalamazoo,TWA) flight(NYC,Kalamazoo,TWA) ?flight(NYC,Kalamazoo,TWA) ?flight(NYC,Kalamazoo,United) ?flight(NYC,Kalamazoo,United) flight(NYC,Kalamazoo,United) ?flight … ?flight … ?flight … flight … flight … Unintended models make classical logic and deduction inadequate for addressing many of the reasoning tasks we would like to perform. In many instances there are conventions for dealing with this incomplete knowledge. NMR strives to identify these conventions and to formalize them in order to define - intended semantics for KBs - procedural semantics for alternate inference systems (like NAF in Prolog)