Propositional and First Order Logic Questions and Answers

Answer: A

The separation of tell from ask means consequences are derived on demand rather than eagerly, which keeps the base small at the cost of work at query time. Some systems do forward chain eagerly instead, which is the opposite trade. Normal form conversion and consistency checking are optional preprocessing. Choosing between computing consequences when they are learned and when they are needed is the same eager against lazy trade found throughout computing.

Answer: B

Forward chaining is data driven, firing every rule whose conditions are met and adding the conclusions, which suits a system reacting to a stream of observations. Backward chaining is goal driven and does less work when only one question matters. Rule composition and textual order are different notions. Matching the direction of inference to whether data or questions arrive first is a decision that can change the work done by orders of magnitude.

Answer: C

Working back from the query touches only the rules relevant to it, whereas forward chaining would derive an enormous number of true but irrelevant facts. Streaming data favours forward chaining. Variables and eventual need are separate considerations, the last favouring forward chaining. Directing computation by what is actually wanted, rather than computing everything and selecting afterwards, is a principle far wider than inference.

Answer: D

An agent reasoning about action needs to say not only what each action changes but what it leaves untouched, and stating that concisely is the frame problem. Perceptual boundaries, belief validity and vocabulary are different notions. That the difficulty lies in describing what does not happen, rather than what does, is the counter-intuitive core of reasoning about action and it is worth stating in exactly those terms.

Answer: A

Adding a sentence extends what the agent knows without any code being written, which is the declarative promise that motivated the whole approach. Speed is generally poor, every logic makes an ontological commitment, and first order questions are not answerable in bounded time. The ability to extend a system by stating facts rather than by programming is the property that justified the cost, and it is worth naming precisely.

Answer: B

Almost everything an agent believes is probable rather than certain and admits exceptions, which classical logic handles badly and probability theory handles naturally. That objection drove the field's shift towards probabilistic methods. Storage, soundness and relations are not the difficulty. An approach whose limitation lies in the character of the knowledge rather than in the machinery is one that must be supplemented rather than optimised.

Answer: C

Bayesian networks and the non-monotonic logics were both developed to handle knowledge that is uncertain or defeasible, and the probabilistic branch has proved by far the more influential. Restriction to certain domains would exclude almost everything interesting, and adding rules to a monotonic logic cannot make it non-monotonic. Two distinct responses to the same limitation, one of which prospered, is a piece of the field's history worth knowing when reading older literature.

Answer: D

Where a conclusion must be justified, audited or guaranteed, an explicit derivation from stated premises is exactly what is required, and no learned model supplies it. Speed, learning and coverage are where the alternatives are stronger. Judging a tradition by the properties it uniquely supplies, rather than by whether it won, is what allows it to be combined sensibly with the approaches that displaced it.

Answer: A

In propositional logic each symbol stands for a whole statement with no internal structure, so nothing can be said about the objects the statement concerns. Object denotation, connectives and bindable variables belong to first order syntax. That the smallest unit is a whole fact, rather than an object or a property, is exactly the limitation that first order logic removes, and stating it precisely explains why the step upward was needed.

Answer: B

The five connectives build compound sentences from simpler ones, and each has a truth table fixing exactly when the compound is true. Quantifiers belong to first order logic, and the other lists are arithmetic and programming constructs. That every connective is truth functional, so the truth of a compound depends only on the truth of its parts, is what makes truth tables a complete account of the semantics.