Skip to content
CampusPrep
AD-304 · Artificial Intelligence/Quick Revision Short Notes

Artificial Intelligence (AD-304) - Unit 5 Short Notes

How unit 5 is examined

This unit covers what an expert system is, its architecture, forward and backward chaining, limitations and development; the marks sit in characteristics, components, chaining and limitations.

Expert systems (ES) and its Characteristics

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. <mark>An expert system is a knowledge-based computer program that emulates the decision-making ability of a human expert in a narrow domain, by applying a knowledge base of facts and rules through an inference engine to give advice and explain its conclusions.</mark>

Diagram.

<figure class="ds-fig" style="margin:1.4rem 0;overflow-x:auto"><svg xmlns="http://www.w3.org/2000/svg" id="dsfig-u5-01" viewBox="0 0 596 303.6" width="596" height="303.6" role="img" aria-label="ES architecture. U = user, UI = user interface, IE = inference engine, KB = knowledge base, WM = working memory, EF = explanation facility, KA = knowledge acquisition subsystem, KE = knowledge engineer and domain expert."><style>#dsfig-u5-01 .e{stroke:#454C5A;stroke-width:1.4;fill:none}#dsfig-u5-01 .e.hi{stroke:#2340B8;stroke-width:2.6}#dsfig-u5-01 .n{fill:#FFFFFF;stroke:#16181D;stroke-width:1.4}#dsfig-u5-01 .n.hi{fill:#E3E9FC;stroke:#2340B8;stroke-width:2.2}#dsfig-u5-01 .n.rb-b{fill:#16181D;stroke:#16181D}#dsfig-u5-01 .n.rb-r{fill:#BD3227;stroke:#BD3227}#dsfig-u5-01 text{font-family:"JetBrains Mono",ui-monospace,Menlo,Consolas,monospace;font-size:13px}#dsfig-u5-01 .t{fill:#16181D;font-weight:500}#dsfig-u5-01 .t.inv{fill:#FFFFFF;font-weight:700}#dsfig-u5-01 .kd{stroke:#16181D;stroke-width:1.2}#dsfig-u5-01 .dot{fill:#16181D}#dsfig-u5-01 .ann{fill:#2340B8;font-size:11px;font-weight:700}#dsfig-u5-01 .lbl{fill:#6F7787;font-family:system-ui,-apple-system,sans-serif;font-size:12px;font-weight:700}#dsfig-u5-01 .ptr{fill:#2340B8;font-size:12px;font-weight:700}#dsfig-u5-01 .ah{fill:#454C5A}#dsfig-u5-01 .ah.hi{fill:#2340B8}#dsfig-u5-01 .wl rect{fill:#FFFFFF;stroke:#DCE0E7}#dsfig-u5-01 .wl .t{font-size:12px;font-weight:700}#dsfig-u5-01 .wl.hi rect{fill:#2340B8;stroke:#2340B8}#dsfig-u5-01 .wl.hi .t{fill:#FFFFFF}html.dark #dsfig-u5-01 .e{stroke:#B1B7C3}html.dark #dsfig-u5-01 .e.hi{stroke:#8FA3FF}html.dark #dsfig-u5-01 .n{fill:#161920;stroke:#E6E8ED}html.dark #dsfig-u5-01 .n.hi{fill:#1E2748;stroke:#8FA3FF}html.dark #dsfig-u5-01 .n.rb-b{fill:#E6E8ED;stroke:#E6E8ED}html.dark #dsfig-u5-01 .n.rb-r{fill:#FF7E71;stroke:#FF7E71}html.dark #dsfig-u5-01 .t{fill:#E6E8ED}html.dark #dsfig-u5-01 .t.inv{fill:#0F1115}html.dark #dsfig-u5-01 .kd{stroke:#E6E8ED}html.dark #dsfig-u5-01 .dot{fill:#E6E8ED}html.dark #dsfig-u5-01 .ann{fill:#8FA3FF}html.dark #dsfig-u5-01 .lbl{fill:#858D9C}html.dark #dsfig-u5-01 .ptr{fill:#8FA3FF}html.dark #dsfig-u5-01 .ah{fill:#B1B7C3}html.dark #dsfig-u5-01 .ah.hi{fill:#8FA3FF}html.dark #dsfig-u5-01 .wl rect{fill:#161920;stroke:#2A2E37}html.dark #dsfig-u5-01 .wl.hi rect{fill:#8FA3FF;stroke:#8FA3FF}html.dark #dsfig-u5-01 .wl.hi .t{fill:#0F1115}</style><defs><marker id="ah15" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah" d="M0,1 L9,5 L0,9 z"/></marker><marker id="ahh15" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah hi" d="M0,1 L9,5 L0,9 z"/></marker></defs><path class="e" d="M61,126 L122.2,126" marker-end="url(#ah15)" marker-start="url(#ah15)"/><path class="e" d="M164.2,126 L242.6,126" marker-end="url(#ah15)" marker-start="url(#ah15)"/><path class="e" d="M280.7,113.8 L366.9,52.2" marker-end="url(#ah15)" marker-start="url(#ah15)"/><path class="e" d="M280.7,138.2 L366.9,199.8" marker-end="url(#ah15)" marker-start="url(#ah15)"/><path class="e" d="M263.6,145 L263.6,242.6" marker-end="url(#ah15)"/><path class="e" d="M251.1,249.3 L157,141.8" marker-end="url(#ah15)"/><path class="e" d="M485.4,40 L405,40" marker-end="url(#ah15)"/><path class="e" d="M548.5,142.9 L512.7,59.3" marker-end="url(#ah15)"/><circle class="n" cx="40" cy="126" r="18"/><text class="t" x="40" y="126" dy=".35em" text-anchor="middle">U</text><circle class="n" cx="143.2" cy="126" r="18"/><text class="t" x="143.2" y="126" dy=".35em" text-anchor="middle">UI</text><circle class="n" cx="263.6" cy="126" r="18"/><text class="t" x="263.6" y="126" dy=".35em" text-anchor="middle">IE</text><circle class="n" cx="384" cy="40" r="18"/><text class="t" x="384" y="40" dy=".35em" text-anchor="middle">KB</text><circle class="n" cx="384" cy="212" r="18"/><text class="t" x="384" y="212" dy=".35em" text-anchor="middle">WM</text><circle class="n" cx="263.6" cy="263.6" r="18"/><text class="t" x="263.6" y="263.6" dy=".35em" text-anchor="middle">EF</text><circle class="n" cx="504.4" cy="40" r="18"/><text class="t" x="504.4" y="40" dy=".35em" text-anchor="middle">KA</text><circle class="n" cx="556" cy="160.4" r="18"/><text class="t" x="556" y="160.4" dy=".35em" text-anchor="middle">KE</text></svg><figcaption style="font-size:.82em;opacity:.72;margin-top:.45rem">ES architecture. U = user, UI = user interface, IE = inference engine, KB = knowledge base, WM = working memory, EF = explanation facility, KA = knowledge acquisition subsystem, KE = knowledge engineer and domain expert.</figcaption></figure>

Key points.

  1. An ES has high performance: within its narrow domain it reaches conclusions of the quality and speed of a human expert.
  2. It is understandable: the explanation facility answers "why?" and "how?", so users trust the advice.
  3. It is reliable and consistent: the same input gives the same answer, never tired or biased.
  4. It is responsive, often answering in real time.
  5. The knowledge base is kept separate from the inference engine, so rules can be changed without rewriting the program.
  6. It reasons symbolically with heuristic rules and copes with incomplete or uncertain data using certainty factors.
  7. It is restricted to a narrow domain and does not have common sense.

Difference from conventional programs.

Basis Conventional program Expert system
Knowledge Mixed into code Separate knowledge base
Method Algorithm, fixed steps Heuristic inference over rules
Output Definite answer Advice with explanation, may be uncertain
Data Needs complete data Copes with incomplete data

Examples. MYCIN (diagnosing blood infections), DENDRAL (identifying molecular structure), PROSPECTOR (mineral exploration), XCON (configuring computers).

Role in AI and problem-solving.

  1. An ES emulates a human expert's decision-making, so a non-expert gets expert-quality advice in a narrow domain.
  2. It solves problems with no exact algorithm by non-algorithmic, heuristic reasoning with rules of thumb instead of fixed steps.
  3. Early AI built general problem solvers such as GPS (Newell and Simon), which failed on real problems; Feigenbaum's "knowledge is power" shifted AI to knowledge-based systems, whose power comes from domain knowledge, not general search.
  4. Expert systems were the first commercial success of AI (XCON at DEC) and gave AI two lasting ideas: knowledge separate from inference, and explainable reasoning.

Benefits by domain.

  • Commercial, loan approval and insurance: every application meets the same rules, so decisions are consistent and fast, at lower cost than scarce senior underwriters.
  • Commercial, configuration: XCON configured VAX computer orders at DEC and saved about 25 million dollars a year.
  • Technical, fault diagnosis: the system is available at any hour on the shop floor and keeps the expertise of engineers after they retire or leave.
  • Technical, geology: PROSPECTOR takes a geologist's expertise to every site and helped locate a molybdenum deposit.

Answer frame. Open with the definition; draw the architecture block diagram; develop characteristics 1-7, then the conventional-program table; close with MYCIN and DENDRAL as examples. For "role in AI", develop role points 1-4, then the benefits by domain, and close with "ES was the first commercial success of AI". For "advantages and limitations", use five advantages from Benefits and four limitations from Expert Systems Limitation.

Pitfall: Characteristics alone lose marks; add the conventional-program difference and named examples.

Asked: [7 marks] (Nov 2022, Jun 2024, Dec 2025) Explain expert systems and their characteristics. Asked: [7 marks] (Nov 2022, Jun 2024) What is Expert System? Write about its characteristics. List and describe the key characteristics that distinguish expert systems from other computer programs. Asked: [7 marks] (Dec 2023) What are the key advantage and limitations of expert systems in the field of Artificial Intelligence? Asked: [7 marks] (Dec 2024) Define what Expert Systems (ES) are and explain their role in artificial intelligence and problem-solving.

Requirements of ES

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Low weight</span>

Definition. The requirements of an ES are the factors that must be right for it to be built and work well.

Key points.

  1. Domain suitability: the problem must be well defined, bounded and need expert judgment, and a real expert must be available.
  2. Knowledge acquisition and representation: knowledge must be extracted from experts efficiently and stored as production rules, frames or semantic networks.
  3. Inference mechanism: choose forward or backward chaining and a conflict-resolution strategy that suits the problem.
  4. Uncertainty handling: a certainty factor (CF) from $-1$ (definitely false) to $+1$ (definitely true) is attached to each rule, and two positive CFs for the same conclusion combine as $CF = CF_1 + CF_2(1 - CF_1)$, so 0.6 and 0.5 give 0.8.
  5. Fuzzy logic gives a membership degree from 0 to 1, and Bayes' rule $P(H|E) = P(E|H)\,P(H)/P(E)$ updates the probability of hypothesis $H$ given evidence $E$.
  6. Interface and explanation: the system must answer "why do you ask this?" and "how did you conclude this?", because users accept advice only when they can see its reasoning, and the same rule trace lets developers debug wrong conclusions.
  7. Scalability: a modular, partitioned knowledge base (rule groups per sub-task) and efficient pattern matching such as the Rete algorithm, which remembers partial matches instead of re-testing every rule each cycle, keep a large rule base fast and maintainable.
  8. Validation and testing: run known-answer test cases, have experts review conclusions, and check the rule base for conflicting, redundant, circular and unreachable rules.

Asked: [7 marks] (Dec 2023) What are the critical factors to consider when designing and developing an effective expert system?

Components and capability of expert systems

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. <mark>The components of an expert system are the knowledge base, inference engine, working memory, user interface, explanation facility and knowledge acquisition subsystem, which together let it advise, explain and solve problems like an expert.</mark>

Diagram. Draw the architecture block diagram given under Expert systems (ES) and its Characteristics, with arrows as shown.

Key points.

  1. The knowledge base holds domain facts and heuristic IF-THEN rules supplied by experts; its quality decides the system's power.
  2. The inference engine is the reasoning part: it matches rules against facts and fires them by forward or backward chaining, and it resolves conflicts when several rules apply.
  3. Working memory holds the facts of the current case, the user's answers and the conclusions derived so far.
  4. The user interface lets the non-expert enter facts and read advice through menus, questions or natural language.
  5. The explanation facility shows how a conclusion was reached (the rule trace) and why a question is asked, which builds user trust.
  6. The knowledge acquisition subsystem lets the knowledge engineer add, edit and test rules gathered from the domain expert, without reprogramming.

Working cycle. The interface passes the user's query to the inference engine, which matches facts against the knowledge-base rules and fires those that apply, and the solution returns with its explanation.

Capabilities. It can advise and recommend decisions, instruct and train novices, explain its reasoning (why and how), and solve complex domain problems at expert level, even with uncertain data.

Answer frame. Open with the definition; draw the block diagram; describe points 1-6 in data-flow order; give the working cycle and capabilities; for "discuss", close with limitations and an application.

Asked: [7 marks] (Nov 2022, Jun 2023, Jun 2025) Explain about the components and capability of Expert system. List and explain the components and capabilities of Expert System. Describe the architecture and working of an expert system. Asked: [7 marks] (Dec 2025) Discuss components, inference engine and limitations of expert systems.

Inference Engine: forward and backward chaining

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. <mark>The inference engine is the ES module that applies the rules of the knowledge base to the known facts to derive conclusions; forward chaining is data-driven (facts to goal) and backward chaining is goal-driven (goal to facts).</mark>

Key points.

  1. Forward chaining starts from known facts: every rule whose IF part matches enters the conflict set, conflict resolution selects one rule, it fires and adds its new fact to working memory, and the match-resolve-act cycle repeats until the goal is reached or no rule matches.
  2. It is therefore bottom-up and data-driven.
  3. Backward chaining starts from a hypothesis (goal), finds rules whose THEN part concludes it, and treats their IF conditions as sub-goals recursively, until known facts or the user's answers prove them.
  4. It is top-down and goal-driven, and it explores only rules relevant to the goal, so its search space is small.
  5. Forward suits planning, monitoring and control, where data is plentiful and goals many; backward suits diagnosis and verifying one hypothesis, as in MYCIN.

Example. Rules: R1 IF A and B THEN C; R2 IF C THEN D. Facts: A, B. Forward: R1 fires, C is added, R2 fires, D is proved. Backward, goal D: R2 needs C; R1 needs A and B; both are facts, so D is proved.

Role of the inference engine. In a rule-based ES the inference engine is the control program: it matches the facts in working memory against the rules of the knowledge base, chooses one rule by conflict resolution (priority, specificity, recency), fires it, and repeats. Forward and backward chaining are its two control strategies, and the explanation facility reads its rule trace.

Basis Forward chaining Backward chaining
Direction Known facts to goal Hypothesis (goal) to facts
Approach Data-driven, bottom-up Goal-driven, top-down
Search space Wide: all matching rules enter the conflict set Narrow: only rules that conclude the goal
Speed and efficiency May fire many irrelevant rules, so it can be slow Focused, so it is fast for one goal
Suited to Planning, monitoring, control Diagnosis, debugging
Example system XCON MYCIN

Answer frame. For "explain", define both, give both traces, then applications. For "compare", define both, draw the table, and end with when to prefer each. For "usage of backward chaining", define it, give the steps (goal, matching rules, sub-goals, ask the user), MYCIN, and the advantage of asking only relevant questions.

Asked: [7 marks] (Nov 2022, Jun 2025) Explain about forward and backward chaining. Show how forward and backward reasoning are used in expert systems. Asked: [7 marks] (Jun 2023) Discuss the usage of backward chaining in Inference Engine? Asked: [7 marks] (Dec 2023, Dec 2024) What are the main differences between forward chaining and backward chaining in expert systems, and when would you prefer to use each approach? Compare and contrast the inference engines used in expert systems with examples of scenarios where each is more suitable.

Expert Systems Limitation

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">High weight</span>

Definition. <mark>Limitations of expert systems are the weaknesses that restrict them to narrow domains and stop them learning or reasoning like a human.</mark>

Key points.

  1. Knowledge acquisition bottleneck: extracting expert knowledge is slow and costly, since experts cannot easily state what they know.
  2. Knowledge representation is hard: IF-THEN rules poorly capture uncertain, temporal or intuitive knowledge, such as "probably flu, and the fever was higher last week but is falling".
  3. Brittleness: outside its narrow domain or at the edge of its rules, it fails sharply instead of degrading gracefully.
  4. No common sense or general reasoning, and no creativity.
  5. It cannot learn from experience, so every change needs a manual rule update.
  6. Large rule bases become slow and hard to maintain, since a new rule can contradict others.

Classical versus heuristic search (part b). Classical (uninformed, blind) search uses only the problem definition; heuristic (informed) search uses a heuristic function $h(n)$, the estimated cost from $n$ to the goal, to expand the most promising node first.

  1. Greedy best-first expands the node with least $h(n)$ and is not optimal; A* expands the node with least $f(n) = g(n) + h(n)$, where $g(n)$ is the path cost from the start to $n$.
  2. $h$ is admissible when it never overestimates, $h(n) \le h^*(n)$ (the true cost to the goal); with an admissible $h$ (consistent $h$ for graph search), A* is complete and optimal.
  3. In the table, $b$ = branching factor, $d$ = solution depth, $m$ = maximum depth.

Algorithms.

BFS: put start in a FIFO queue; remove the front node, stop if goal, else add its unvisited children at the back; repeat.
DFS: as BFS, but with a LIFO stack, so the deepest node is expanded first.
A*: put start in OPEN with f = g + h; remove the least-f node, stop if goal; else expand it with g(child) = g(node) + step cost and insert children by f; repeat.
Greedy best-first: as A*, but order OPEN by h(n) alone.
Basis Classical (uninformed) Heuristic (informed)
Examples BFS, DFS, uniform-cost (UCS) A*, greedy best-first, hill climbing
Domain knowledge None Uses heuristic $h(n)$
Speed Slow, explores blindly Faster, guided to goal
Time and space BFS $O(b^d)$ both; DFS $O(b^m)$ time, $O(bm)$ space Far less with a good $h$; A* worst case still exponential
Completeness BFS, UCS complete; DFS not (infinite paths) A* complete; greedy and hill climbing not
Optimality BFS optimal only when all step costs are equal; UCS optimal in general; DFS not optimal A* optimal if $h$ admissible; greedy and hill climbing not optimal

<figure class="ds-fig" style="margin:1.4rem 0;overflow-x:auto"><svg xmlns="http://www.w3.org/2000/svg" id="dsfig-u5-02" viewBox="0 0 467 252" width="467" height="252" role="img" aria-label="Heuristic h = S4 A5 B3 C6 D6 E2 G0 (admissible). Highlighted path S-B-E-G costs 5."><style>#dsfig-u5-02 .e{stroke:#454C5A;stroke-width:1.4;fill:none}#dsfig-u5-02 .e.hi{stroke:#2340B8;stroke-width:2.6}#dsfig-u5-02 .n{fill:#FFFFFF;stroke:#16181D;stroke-width:1.4}#dsfig-u5-02 .n.hi{fill:#E3E9FC;stroke:#2340B8;stroke-width:2.2}#dsfig-u5-02 .n.rb-b{fill:#16181D;stroke:#16181D}#dsfig-u5-02 .n.rb-r{fill:#BD3227;stroke:#BD3227}#dsfig-u5-02 text{font-family:"JetBrains Mono",ui-monospace,Menlo,Consolas,monospace;font-size:13px}#dsfig-u5-02 .t{fill:#16181D;font-weight:500}#dsfig-u5-02 .t.inv{fill:#FFFFFF;font-weight:700}#dsfig-u5-02 .kd{stroke:#16181D;stroke-width:1.2}#dsfig-u5-02 .dot{fill:#16181D}#dsfig-u5-02 .ann{fill:#2340B8;font-size:11px;font-weight:700}#dsfig-u5-02 .lbl{fill:#6F7787;font-family:system-ui,-apple-system,sans-serif;font-size:12px;font-weight:700}#dsfig-u5-02 .ptr{fill:#2340B8;font-size:12px;font-weight:700}#dsfig-u5-02 .ah{fill:#454C5A}#dsfig-u5-02 .ah.hi{fill:#2340B8}#dsfig-u5-02 .wl rect{fill:#FFFFFF;stroke:#DCE0E7}#dsfig-u5-02 .wl .t{font-size:12px;font-weight:700}#dsfig-u5-02 .wl.hi rect{fill:#2340B8;stroke:#2340B8}#dsfig-u5-02 .wl.hi .t{fill:#FFFFFF}html.dark #dsfig-u5-02 .e{stroke:#B1B7C3}html.dark #dsfig-u5-02 .e.hi{stroke:#8FA3FF}html.dark #dsfig-u5-02 .n{fill:#161920;stroke:#E6E8ED}html.dark #dsfig-u5-02 .n.hi{fill:#1E2748;stroke:#8FA3FF}html.dark #dsfig-u5-02 .n.rb-b{fill:#E6E8ED;stroke:#E6E8ED}html.dark #dsfig-u5-02 .n.rb-r{fill:#FF7E71;stroke:#FF7E71}html.dark #dsfig-u5-02 .t{fill:#E6E8ED}html.dark #dsfig-u5-02 .t.inv{fill:#0F1115}html.dark #dsfig-u5-02 .kd{stroke:#E6E8ED}html.dark #dsfig-u5-02 .dot{fill:#E6E8ED}html.dark #dsfig-u5-02 .ann{fill:#8FA3FF}html.dark #dsfig-u5-02 .lbl{fill:#858D9C}html.dark #dsfig-u5-02 .ptr{fill:#8FA3FF}html.dark #dsfig-u5-02 .ah{fill:#B1B7C3}html.dark #dsfig-u5-02 .ah.hi{fill:#8FA3FF}html.dark #dsfig-u5-02 .wl rect{fill:#161920;stroke:#2A2E37}html.dark #dsfig-u5-02 .wl.hi rect{fill:#8FA3FF;stroke:#8FA3FF}html.dark #dsfig-u5-02 .wl.hi .t{fill:#0F1115}</style><defs><marker id="ah16" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah" d="M0,1 L9,5 L0,9 z"/></marker><marker id="ahh16" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="ah hi" d="M0,1 L9,5 L0,9 z"/></marker></defs><path class="e" d="M55.8,115.5 L153.2,50.5"/><path class="e hi" d="M55.8,136.5 L153.2,201.5"/><path class="e" d="M188,40 L279,40"/><path class="e" d="M184.8,50.5 L282.2,115.5"/><path class="e hi" d="M188,212 L279,212"/><path class="e hi" d="M317,212 L408,212"/><g class="wl"><rect x="94.9" y="74" width="19.2" height="18" rx="9"/><text class="t" x="104.5" y="83" dy=".35em" text-anchor="middle">1</text></g><g class="wl hi"><rect x="94.9" y="160" width="19.2" height="18" rx="9"/><text class="t" x="104.5" y="169" dy=".35em" text-anchor="middle">2</text></g><g class="wl"><rect x="223.9" y="31" width="19.2" height="18" rx="9"/><text class="t" x="233.5" y="40" dy=".35em" text-anchor="middle">3</text></g><g class="wl"><rect x="223.9" y="74" width="19.2" height="18" rx="9"/><text class="t" x="233.5" y="83" dy=".35em" text-anchor="middle">3</text></g><g class="wl hi"><rect x="223.9" y="203" width="19.2" height="18" rx="9"/><text class="t" x="233.5" y="212" dy=".35em" text-anchor="middle">1</text></g><g class="wl hi"><rect x="352.9" y="203" width="19.2" height="18" rx="9"/><text class="t" x="362.5" y="212" dy=".35em" text-anchor="middle">2</text></g><circle class="n" cx="40" cy="126" r="18"/><text class="t" x="40" y="126" dy=".35em" text-anchor="middle">S</text><circle class="n" cx="169" cy="40" r="18"/><text class="t" x="169" y="40" dy=".35em" text-anchor="middle">A</text><circle class="n" cx="169" cy="212" r="18"/><text class="t" x="169" y="212" dy=".35em" text-anchor="middle">B</text><circle class="n" cx="298" cy="40" r="18"/><text class="t" x="298" y="40" dy=".35em" text-anchor="middle">C</text><circle class="n" cx="298" cy="126" r="18"/><text class="t" x="298" y="126" dy=".35em" text-anchor="middle">D</text><circle class="n" cx="298" cy="212" r="18"/><text class="t" x="298" y="212" dy=".35em" text-anchor="middle">E</text><circle class="n" cx="427" cy="212" r="18"/><text class="t" x="427" y="212" dy=".35em" text-anchor="middle">G</text></svg><figcaption style="font-size:.82em;opacity:.72;margin-top:.45rem">Heuristic h = S4 A5 B3 C6 D6 E2 G0 (admissible). Highlighted path S-B-E-G costs 5.</figcaption></figure>

Worked expansion. BFS expands S, A, B, C, D, E, G (7 nodes). A* expands S ($f=4$), B ($f=2+3=5$, beating A at $1+5=6$), E ($f=3+2=5$), G ($f=5+0=5$): A expands 4 nodes and returns the optimal path S-B-E-G of cost 5.*

Classical versus modern AI (part c). Classical (symbolic) AI, such as MYCIN and XCON, encodes explicit rules and explains every step but breaks outside them; modern (data-driven) AI, such as deep-learning vision, large language models and AlphaGo, learns from large data and reaches superhuman accuracy but is hard to explain.

Basis Classical (symbolic) AI Modern (data-driven) AI
Method Hand-coded rules and logic Learns from data (machine and deep learning)
Nature Deterministic, rigid Probabilistic
Performance Expert-level only in a narrow domain; poor on noisy data State of the art in vision, language, games
Adaptability Rules edited by hand Retrained on new data
Explanation Clear rule trace Often a black box
Data need Little data, much expert effort Very large data and compute

NLP (part d). <mark>Natural language processing (NLP) is the branch of AI that lets computers understand, interpret and generate human language, in text or speech.</mark>

  1. Natural language understanding (NLU) maps sentences to meaning and resolves ambiguity; natural language generation (NLG) turns meaning or data into fluent sentences.
  2. Lexical analysis splits text into tokens and finds word roots.
  3. Syntactic analysis (parsing) checks grammar and builds a parse tree.
  4. Semantic analysis finds the meaning of words and sentences, rejecting meaningless ones like "hot ice cream".
  5. Discourse integration uses earlier sentences (what "it" means); pragmatic analysis finds intent ("Can you pass the salt?" is a request).
  6. Applications include machine translation, chatbots, sentiment analysis, speech assistants and spam filtering.
  7. NLP matters to AI because language is the natural human interface, so people use machines without programming and the knowledge in text becomes usable.

Answer frame. (a) points 1-6, bottleneck first; (b) define $h(n)$ and $f(n)$, algorithms, table, graph and expansions, closing "heuristic search wins when a good admissible $h$ exists"; (c) one paragraph, the table and hybrid neuro-symbolic systems; (d) definition, NLU and NLG, stages, applications, importance.

Pitfall: Writing "BFS is always optimal" loses marks; BFS is optimal only when all step costs are equal.

Asked: [14 marks] (Jun 2024) Attempt any two: a) Analyze the limitations of expert systems, including knowledge acquisition and representation. b) Differentiate between classical search algorithms and heuristic search algorithms. c) Compare and contrast classical AI systems and modern AI systems in performance and adaptability. d) Define natural language processing (NLP) and its importance in AI.

Expert System Development Environment

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Medium weight</span>

Definition. <mark>An ES development environment is the set of tools, shells and languages used to build, test and maintain an expert system, from knowledge acquisition to deployment.</mark>

Key points.

  1. Choose a suitable domain, such as legal advice, financial planning or disease diagnosis, and define the problem and users.
  2. Programming languages (LISP, Prolog) give full flexibility, while an ES shell (CLIPS, JESS, EMYCIN) supplies a ready inference engine and interface, so only the knowledge base is built.
  3. The knowledge engineer acquires knowledge from experts and regulations by interviews and case studies and encodes it as IF-THEN rules.
  4. The architecture (knowledge base, inference engine, working memory, interface, explanation module) is built and tested by incremental prototyping.

Example (legal advice, tenant-law adviser). Facts: rent_unpaid_months, notice_given, lease_registered. Rules: R1 IF rent_unpaid_months >= 2 AND notice_given = yes THEN landlord may seek eviction (CF 0.8); R2 IF notice_given = no THEN eviction is invalid, serve a notice first (CF 1.0); R3 IF rent_unpaid_months >= 2 AND tenant paid arrears within the notice period THEN eviction is barred (CF 0.9). Case: 3 months unpaid, notice given, arrears not paid: R1 matches; R2 and R3 do not, so the advice is "landlord may seek eviction" with CF 0.8, explained by R1. Choice: forward chaining, as the user supplies facts and the system derives consequences; if several rules match, the most specific (R3 over R1) wins.

Example (financial planning). Rules: F1 IF risk_profile = low AND age > 55 THEN allocate 30% equity, 70% debt (CF 0.8); F2 IF risk_profile = high AND age < 35 THEN allocate 80% equity, 20% debt (CF 0.85); F3 IF emergency_fund < 6 months of expenses THEN build the emergency fund before investing (CF 0.95). Case: age 28, high risk, 8 months' fund: F2 fires, giving 80:20 with CF 0.85.

User interface. It asks one question at a time (age, income, goal and horizon, then a risk questionnaire), skipping questions no rule needs, and shows the allocation with its confidence; a "why" button shows the rules fired and the facts used.

Example (disease diagnosis). Rules: R1 IF fever AND cough AND fatigue THEN flu; R2 IF fever AND chills AND sweating THEN malaria; R3 IF sneezing AND itching AND NOT fever THEN allergy; R4 IF flu AND high fever AND breathlessness THEN see a doctor urgently; R5 IF malaria THEN advise a blood smear test. Method: backward chaining, testing one disease at a time by asking only its symptoms; conflicts go by specificity. Case: fever, cough, fatigue, no chills: R1 fires and R2 fails at chills, so the outcome is flu.

Keeping knowledge current. Live feeds: a new court ruling, tax rate or health guideline flags every rule citing it. Automated refinement: the system compares its advice with recorded outcomes and lowers the certainty factor of a rule that is often wrong (R1 at CF 0.8 advised eviction but courts often refused, so its CF drops), proposing a fix for review. Dynamic rule updating: the rule base is versioned, so rules are added, edited or retired with an effective date through the knowledge-acquisition editor, without reprogramming the inference engine. Expert validation: an expert approves each change on test cases before release.

Answer frame. Open with the domain and users (legal or financial example plus user interface for Jun 2024, disease for Jun 2025); draw the architecture; develop acquisition, rules, inference method and the case; close with the four updating mechanisms.

Asked: [7 marks] (Jun 2024, Jun 2025) Develop a comprehensive expert system for a complex domain, such as legal advice or financial planning, and explain how the knowledge is kept up to date. Design a rule-based expert system for diagnosing diseases.

Technology (ES tools)

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. ES technology means the tools used to build expert systems: languages, shells and knowledge-engineering aids.

Key points.

  1. Prolog and LISP are languages for symbolic reasoning that give full control but need programming effort.
  2. Shells such as CLIPS, EMYCIN and JESS supply the inference engine and interface, so only the rules are added.

Benefits of Expert Systems

<span style="display:inline-block;padding:.16em .6em;border:1.5px solid currentColor;border-radius:999px;font-size:.68em;font-weight:700;letter-spacing:.06em;text-transform:uppercase;opacity:.75">Not asked since 2022</span>

Definition. Benefits of ES are the advantages that make organisations use them beside human experts.

Key points.

  1. High availability: it works around the clock and spreads scarce expertise everywhere.
  2. Consistency: it treats every case by the same rules, free of fatigue and emotion.
  3. Lower cost and retained expertise: it saves hiring and training experts and keeps expertise when they leave (see Benefits by domain).
  4. Speed and explanation: it solves domain problems faster than a human and justifies its conclusions.

Last-minute revision

  • Characteristics: high performance, understandability, reliability, responsiveness, knowledge separate from inference.
  • Forward chaining is data-driven, wide and suits planning; backward is goal-driven, focused and suits diagnosis.
  • Limitations: acquisition bottleneck, brittleness, no common sense, no learning.
  • Knowledge stays current by live feeds, automated refinement, versioned rule updates and expert validation.

Memory hooks

  • KIWUE: Knowledge base, Inference engine, Working memory, User interface, Explanation.
  • Forward = Facts first; Backward = Blueprint (goal) first.
  • Limitations spell BNC: Bottleneck, Narrow (brittle), Common sense missing.
  • MYCIN is Medicine and backward; DENDRAL is Chemistry.

Coverage checklist

  • Expert systems (ES) and its Characteristics: Nov 2022, Jun 2024, Dec 2025 explain; Dec 2023 advantages and limitations; Dec 2024 define and role.
  • requirements of ES: Dec 2023 critical factors.
  • components and capability of expert systems: Nov 2022, Jun 2023, Jun 2025 components; Dec 2025 components, inference and limitations.
  • Inference Engine Forward & backward Chaining: Nov 2022, Jun 2025 explain; Jun 2023 backward usage; Dec 2023, Dec 2024 compare.
  • Expert Systems Limitation: Jun 2024 attempt any two (a-d).
  • Expert System Development Environment: Jun 2024, Jun 2025 design.
  • technology: not asked recently.
  • Benefits of Expert Systems: not asked recently.
Go to where you left off?

Quick Add to Notes

Save questions, your own notes and screenshots into notes filed by unit. It takes a free account.

Create free account

Have an account? Log in