Expert Systems Questions and Answers

Answer: A

A rule base encodes surface associations without the underlying model, so a case just outside its coverage produces a confident wrong answer rather than an admission of ignorance. Human experts degrade gracefully instead. Editing conflicts, firing order and speed are engineering issues. Brittleness motivated both model-based reasoning and the very large common-sense knowledge bases attempted afterwards. Checking the units of each quantity catches a wrong formula before the arithmetic begins.

Answer: B

Model-based diagnosis simulates the correct behaviour of a device from its structure and component models, then localises faults by comparing prediction with observation. It handles novel faults that no rule anticipated, at the cost of needing a model. Rule-based, case-based and connectionist approaches make different trades. The contrast between shallow associations and deep models is the central distinction here. Naming the cost of a choice alongside its benefit is what turns a fact into a design decision.

Answer: C

Case-based reasoning retrieves the most similar stored case, adapts its solution to the new situation, tests the result and retains it as a new case. It suits domains where experience is plentiful but general rules are hard to state, such as law and helpdesk support. Model-based reasoning, chaining and meta-reasoning are different. Its retain step gives it a learning capability rule bases lack. A single measurement chosen to respond to the one asymmetry beats several that respond to shared behaviour.

Answer: D

The four Rs are retrieve a similar case, reuse its solution, revise it against the outcome, and retain the revised case for future use. Retention is what makes the system improve with experience. The acquisition cycle, the production system cycle and plan-generate-test belong elsewhere in this topic. The mnemonic is standard and is asked for verbatim often enough to be worth learning as such. A limit that survives making every external component ideal is an intrinsic one, and worth knowing for that reason.

Answer: A

A truth maintenance system records the justification for each belief so that withdrawing a premise automatically withdraws whatever depended on it. That makes non-monotonic reasoning practical and avoids rebuilding every conclusion from scratch. Evidence combination, conflict resolution and question selection are separate mechanisms. The dependency record also supports explanation, since it is exactly the derivation of a belief. A limit that survives making every external component ideal is an intrinsic one, and worth knowing for that reason.

Answer: B

A JTMS links each belief to the beliefs justifying it, so retraction propagates along the network. The alternative ATMS instead labels each belief with the assumption sets supporting it, which allows several contexts to be explored at once at greater cost. Probabilities and work queues are not what either records. The contrast between the two designs is the usual examination question here. A limit that survives making every external component ideal is an intrinsic one, and worth knowing for that reason.

Answer: C

An ATMS labels each datum with the sets of assumptions under which it holds, so switching between candidate worlds costs nothing and comparison across them is immediate. The price is the size of the labels, which can grow combinatorially. It uses more memory, not less, and it records derivations rather than dispensing with them. Choose it when many contexts matter, and a JTMS otherwise.

Answer: D

Meta-level knowledge says which rules to prefer, when to give up on a line of enquiry and how to organise the search, as distinct from knowledge about the domain itself. TEIRESIAS and NEOMYCIN both made strategy explicit in this way, which improved both explanation and tutoring. Object-level, deep and hypothetical reasoning name different things. Making control knowledge declarative is the general move. Framing the question as which quantity the change acts upon makes the answer follow without calculation.

Answer: A

When strategy is stated as knowledge rather than buried in the engine, it can be examined, explained, taught and changed like any other knowledge. NEOMYCIN exploited exactly this to tutor diagnostic method rather than merely diagnostic conclusions. Speed, rule count and automatic verification are not consequences. The pattern recurs whenever a system is asked to justify how it went about a problem. Working one numerical case through both cases side by side fixes the distinction better than any rule of thumb.

Answer: B

An association between symptom and cause carries no account of why the association holds, so an unanticipated fault produces either silence or a confident error. A causal model can reason about the novel case from first principles. Storage and speed favour the shallow representation, and authorship is unrelated. This is brittleness stated at the level of representation rather than behaviour. The shared properties explain why these belong to one family, and the differences explain why more than one exists.