Knowledge Representation Questions and Answers

Answer: A

To say that John believes the earth is flat, the proposition must be an object that belief relates John to, since a bare sentence cannot appear as an argument. The same device turns events, times and categories into objects. Instantiation, verification and database conversion are different operations. Promoting something from the level of statement to the level of object, so that it can be quantified over and described, is the standard move whenever a representation needs to talk about itself.

Answer: B

Believing relates an agent to the content of the belief, and that content must be an object in the language before it can be an argument, since a sentence cannot be substituted into an argument position. The belief may also be false, which a fact cannot be. Uncertainty, time and ownership are related complications rather than the reason. That a representation must be able to refer to its own sentences, if it is to describe mental states, is a genuine structural requirement.

Answer: C

Lois believes Superman can fly and does not believe Clark Kent can, though the two are the same person, so substitution of identicals fails inside a belief context. That breaks a rule of ordinary logic and requires special treatment. Inspection, inconsistency and identity are different issues. A context in which a basic logical principle fails must be marked as such in the representation, or the reasoner will draw conclusions the believer would reject.

Answer: D

Under the closed world assumption the absence of a fact licenses its negation, which is exactly what a database query does when it reports no matching record. It suits domains where the information is genuinely complete. The domain closure assumption, that only named objects exist, is a separate assumption often adopted alongside it. Knowing which assumption a system makes determines what its silence means, which is not always obvious from its output.

Answer: A

On the web, or in a medical record, the absence of a statement means nothing has been recorded rather than that it is false, so no negative conclusion may be drawn. Description logics adopt this assumption for that reason. Infinite domains, speed and volatility are separate considerations. Choosing the assumption that matches how the knowledge was actually collected, rather than the one that makes reasoning convenient, is essential if the conclusions are to be trusted.

Answer: B

Under the open world assumption silence is not denial, so a patient with no recorded allergy is not thereby known to have none, which is exactly the right behaviour in that domain. Distinctness also requires explicit statement unless a unique names assumption is made. Asserted facts are of course accepted. Refusing to draw a conclusion that a closed world system would draw automatically is the safety property that makes the assumption appropriate where completeness cannot be guaranteed.

Answer: C

Under this assumption two constants that differ syntactically are taken to refer to different things, which databases assume and description logics generally do not. Without it, discovering that two names refer to the same individual is a legitimate inference. Naming conventions and uniqueness of identifiers are different matters. Whether identity must be asserted or may be assumed changes what a reasoner can conclude about counts and about distinctness, which matters more often than it appears.

Answer: D

Describing an action requires saying not only what it changes but that everything else stays the same, and enumerating those frame axioms for every action and every fluent is impossible. Successor state axioms and non-monotonic solutions address it. Frame selection, domain boundaries and slot design are different problems sharing the word. A difficulty that seems trivial to a person and proves intractable to formalise is characteristic of commonsense reasoning.

Answer: A

Writing one axiom per fluent, saying exactly when it becomes or stays true, replaces the quadratic number of frame axioms with a linear number of compact ones. The second option is the naive approach the technique replaces, the third is the non-monotonic alternative, and the fourth changes the problem. Reorganising a description so that its size grows linearly rather than quadratically, without losing content, is a representational improvement of the most practical kind.

Answer: B

Fluents are the facts that change, written with a situation argument, so that on the table holds in one situation and not in another. Actions map situations to situations and eternal facts need no situation argument. Continuity, actions and time points are different notions. Distinguishing what changes from what does not, and marking the difference in the syntax, is what makes reasoning about action possible at all.