CZ3005 Artificial Intelligence

Week 8a – Logical Agent

Yu Han

[Link]@[Link]

Associate Professor
College of Computing and Data Science (CCDS)
Nanyang Technological University (NTU)
Singapore
Second Half Lecturer

Associate Professor Yu Han

• Email: [Link]@[Link]

• Office: S3-B2c-94

• Website: [Link]

2
 Course Overview (2nd Half)
Week Date Venue Topic Tutorial Lab

8 10/03/2026 LT1A Logical Agent (a)

8 13/03/2026 LT1 Logical Agent (b)

9 17/03/2026 LT1A Propositional Logic (a)

9 20/03/2026 LT1 Propositional Logic (b)

10 24/03/2026 LT1A First-Order Logic (a)

Tutorial 4
10 27/03/2026 LT1 First-Order Logic (b) One Lab session in Week 10/11
(Lab Attendance is
11 31/03/2026 LT1A Default Logic (a) NOT Compulsory)

11 03/04/2026 LT1 Default Logic (b)

12 07/04/2026 LT1A Fuzzy Logic (a)

Tutorial 5
12 10/04/2026 LT1 Fuzzy Logic (b)

13 No Class

3
 Why Logic?
• The earliest way people imagined how to build
AI system

• Still a relevant tool to guide today’s AI with

human knowledge

• Good interpretability (which deep learning-

based AI lacks)
4
An Example Logical Agent
[Link]

5
Why Logic?

6
 Why Logic?
• New life for an old technology:
– Deep Learning and Logic Programming are starting
to merge today for building Explainable AI (XAI)
systems.
– Logic Explained Networks:

7
 From Search to Logical Reasoning
• Representation of Knowledge and the
Reasoning Processes are central to the entire
field of AI

• A knowledge-based system (KBS) is a

computer program that reasons and uses a
knowledge base to solve complex problems.

8
Knowledge and Intelligence

Sensing
Knowledge
Decision
Maker

Action

9
 The Knowledge-Based Approach
• Agents that know
– Achieve competence by being told new knowledge
or by learning
– Achieve adaptability by updating their knowledge
– > Knowledge representation
• State of the world, properties and evolution of the world;
goals of the agent, actions and their effect

• Agents that reason Logic

– Use knowledge to deduce course of actions
– > Knowledge inference

10
 Knowledge-Based Agents
• Knowledge base (KB)
– Set of sentences i.e., representations of facts (DB)
– Knowledge representation language
• Adding and querying knowledge
– Tell: add a sentence to the KB
– Ask: retrieve knowledge from the KB
• Answers must follow from what has been Tell’ed (told)

• Inference mechanism
– Determine what follows from the KB

11
 Problem Formulation of KBS
• Knowledge Based System
– States: Instances of the KB (sets of sentences)
–> Use Tell to build the KB
e.g. Tell(KB, “Smoke  Fire”)
Tell(KB, “Fire  Call_999”)
…
Tell(KB, “Smoke”)

– Operators: Add / Infer a new sentence

– Goal: Answer a query
–> Use Ask to query the KB
e.g. Ask(KB, “? Call_999”)

12
A Generic Knowledge-Based Agent
function KB-Agent (percept) returns action
static KB, // a knowledge base
t // a time counter, initially 0

Tell (KB, Make-Percept-Sentence (percept, t))

action  Ask (KB, Make-Action-Query (percept, t))
Tell (KB, Make-Action-Sentence (action, t))
tt+1
return action

– > 3 steps: interpretation, inference, execution

– > KB: background knowledge (observed )
+ acquired information (deduced)
13
 Levels of Knowledge
• Epistemological level Tell Ask
– Declarative description of knowledge
• e.g. facts: “there is smoke in the kitchen”, “it is not warm
enough”
rules: “if there is smoke then there must be a fire”
• Logical level
– Logical encoding of knowledge (into sentences)
• e.g. facts: Smoke; rules: Implies(Smoke, Fire)
• Implementation level
– Physical representation of knowledge (sentences)
• e.g. - the string “Implies(Smoke, Fire)”, or
- a “1” entry in a 2-dimensional array: Implies[X,Y]

14
Example: the Wumpus World

15
 Example: the Wumpus World
• Problem description (PAGE)
– Environment
• Grid of squares, walls; stench breeze
PIT
• Agent, gold, pits, wumpus.
stench
– Goal PIT
breeze
• Find the gold, return to S at [1,1]. breeze
– Percepts stench breeze

• A list of 5 symbols, e.g. [Stench,

Breeze, Glitter, Bump, Scream]; breeze breeze
PIT
• Agent’s location not perceived. S
– Actions
• Go-Forward, Turn-Left, Turn-Right,
Grab, Shoot (1 arrow only), Climb.
16
 The Wumpus World
• Problem description (cont’d)
– Initial state
• Agent at [1,1]; gold, pits and wumpus in random squares.
– Path-cost
• Climbing out with the gold: +1000 (without: 0) • Each action: –1
• Getting killed (pit or wumpus): –10000
– Knowledge
• “In all squares adjacent to the one where the wumpus is,
the agent will perceive a stench.”
• “In all squares adjacent to a pit, the agent will perceive a breeze.”
• In the square where the gold is, the agent will perceive a glitter.”
• When walking into a wall, the agent will perceive a bump.”
• When the wumpus is killed, the agent will perceive a scream.”
17
 Acting and Reasoning in
the Wumpus World
(0) Initial state (1) after {F}
[nil, nil, nil, nil, nil] [nil, Breeze, nil, nil, nil]

4 4

3 3

2 OK 2 P?
OK
A A
V
1 1 B
OK OK P?
OK OK
1 2 3 4 1 2 3 4
A = Agent G = Glitter, Gold P = Pit V = Visited
B = Breeze OK = Safe square S = Stench W = Wumpus 18
 Acting and Reasoning in
the Wumpus World
(1) after {F} (6) after {F, L, L, F, R, F}
[nil, Breeze, nil, nil, nil] [Stench , nil, nil, nil, nil]

4 4

W?
W!
3 3
A
2 P? 2 S W?
OK OK OK
A B P!
V V
1 B 1 V
P?
OK OK OK OK
1 2 3 4 1 2 3 4
A = Agent G = Glitter, Gold P = Pit V = Visited
B = Breeze OK = Safe square S = Stench W = Wumpus 19
 Acting and Reasoning in
the Wumpus World
(6) after {F, L, L, F, R, F} (10) after {F, L, L, F, R, F ,R, F, L, F}
[Stench , nil, nil, nil, nil] [Stench, Breeze, Glitter, nil, nil]

P?
4 4
A P?
W! W!
3 OK 3 SG
B
A S
V
2 S 2 V
OK OK OK OK
OK OK
B B
V P! V P!
1 V 1 V
OK OK
OK OK
1 2 3 4 1 2 3 4
A = Agent G = Glitter, Gold P = Pit V = Visited
B = Breeze OK = Safe square S = Stench W = Wumpus 20
 Summary
• Intelligent agents need …
– Knowledge about the world to make good
decisions.
• Knowledge can be …
– Defined using a knowledge representation
language.
– Stored in a knowledge base in the form of
sentences.
– Inferred, using an inference mechanism and rules.
21
Thank you!

22