KNOWLEDGE

AND
REASONING

Prepared By
-Anooa Joy
KNOWLEDGE-
BASED AGENTS

"An agent can represent

knowledge of its world, its
goals and the current
situation by sentences in
logic and decide what to do
by inferring that a certain
action or course of action is
appropriate to achieve its
goals."
 LOGICAL SYSTEM
• Intelligent agents need knowledge about the world to choose good actions/decisions.
• A logical system is a system that knows about its partially observable environment and can reason
about possible actions by inferring from the hidden information. Reasoning is also known
as inferencing. An agent that acts upon logical system is known as Knowledge based Agent.
• Issues in construction of a logical system:
1. Knowledge representation: how do we represent information? Knowledge representation should
be somewhat natural, expressive and efficient.
2. Knowledge reasoning: how do we use information to reach decisions and/or derive new facts?
• Knowledge in the form of a set of facts about our environment are stored in a knowledge base
(KB).
• Facts are claims about the environment which are either true or false. Facts are represented by
sentences
• A sentence is an assertion about the world. Sentences are expressed in a representation language.
 KNOWLEDGE BASED AGENT
• A knowledge-based agent(Logic Agent)comprises of 2 features:
1. Knowledge base: domain-specific content ie a list of facts that are known to the
agent.

2. Inference engine: domain-independent algorithms for inferencing new

knowledge. Current percepts to infer hidden aspects of the current state using Rules
of inference .
• Knowledge base: A set of sentences in in a formal knowledge representation
language that encodes assertions about the world.
A KNOWLEDGE BASED AGENT
• The agent must be able to:
Represent states, actions, etc.
 Incorporate new percepts
Update internal representations of the world
Deduce hidden properties of the world
Deduce appropriate actions
 A KNOWLEDGE BASED AGENT
• Declarative approach to build a knowledge based agent
• The agent operates as follows:
• Add new sentences: It TELLs the knowledge base what it perceives based on
what it wants to know.
• Query what is known: It ASKs the knowledge base what action it should
perform. The answers should follow from the KB.
• Execute Action: It performs the chosen action.

• Procedural approach to build a knowledge based agent

• Encode desired behaviors directly as program code
• Minimizing the role of explicit representation and reasoning can result in a much
more efficient system. In this approach, knowledge is stored into an empty system
in the form of program code. It designs the behavior of the system via coding
LEVELS OF A KNOWLEDGE-BASED
AGENT
• Knowledge Level: In this level, the behavior of an agent is decided by specifying the following :
• The agent’s current knowledge it has percieved.
• The goal of an agent.

• Implementation Level: This level is the physical representation of the knowledge level. Here, it
is understood that “how the knowledge-based agent actually implements its stored knowledge.”
KNOWLEDGE BASED AGENT
TECHNIQUES USED FOR KNOWLEDGE
REPRESENTATION
• Logic: It is the basic method used to represent the knowledge of a
machine. The term logic means to apply intelligence over the
stored knowledge.
Logic can be further divided as:
1. Propositional Logic: This technique is also known
as propositional calculus, statement logic, or sentential logic. It is
used for representing the knowledge about what is true and what is
false.
2. First-order Logic: It is also known as Predicate logic or First-
order predicate calculus (FOPL). This technique is used to represent
the objects in the form of predicates or quantifiers. It is different
from Propositional logic as it removes the complexity of the sentence
represented by it. In short, FOPL is an advance version of
propositional logic.
TECHNIQUES USED FOR
KNOWLEDGE REPRESENTATION
3. Rule-based System: . In the rule-based system, we impose rules over the propositional
logic and first-order logic techniques. If-then clause is used for this technique. For
example, if there are two variables A and B. Value of both A and B is True.
Consequently, the result of both should also be True and vice-versa. It is represented
as: If the value of A and B is True, then the result will be True. So, such a technique
makes the propositional as well as FOPL logics bounded in the rules.
4. Semantic Networks: The technique is based on storing the knowledge into the system
in the form of a graph. Nodes of a graph represent the objects which exist in the real
world, and the arrow represents the relationship between these objects. Such techniques
show the connectivity of one object with another object. For example, Consider the
given knowledge stored in a machine:
• Ram has a cycle.
• Ram is a boy.
• Cycle has a bell.
• Ram is 12 years old.
• Cycle has two paddles.
TECHNIQUES USED FOR
KNOWLEDGE REPRESENTATION
5. Frames: In this technique, the knowledge is stored via slots and fillers. Slots are the entities
and Fillers are its attributes similar to database. They are together stored in a frame. So, whenever
there is a requirement, the machine infers the necessary information to take the decision. For
example, Tomy is a dog having one tail. It can be framed as:
Tomy((Species (Value = Dog))
(Feature (Value = Tail)))
6. Script: It is an advanced technique over the Frames. Here, the information is stored in the form
of a script. The script is stored in the system containing all the required information. The system
infers the information from that script and solves the problem
LOGIC AS A KR LANGUAGE
WUMPUS WORLD
WUMPUS WORLD
GAME DESCRIPTION
The Wumpus World is a cave consisting of rooms
connected by passageways. Lurking somewhere in the
cave is the Wumpus, a beast that eats any agent that
enters its room. The Wumpus can be shot by an agent,
but agent has only one arrow. Some rooms contain
bottomless pits that trap any agent that wanders into
the room. Occasionally, there is a heap of gold in a
room. The goal is to collect the gold and exit the world
without being eaten.
WUMPUS WORLD ENVIRONMENT
• The agent always starts in the field [1,1].
• The task of the agent is to find the gold, return to the field
[1,1] and climb out of the cave.
• Squares adjacent to Wumpus are smelly and squares
adjacent to pit are breezy (not diagonal)
• Glitter iff gold is in the same square
• Shooting kills Wumpus if you are facing it
• Wumpus emits a horrible scream when it is killed that can be
heard anywhere
• Shooting uses up the only arrow
• Grabbing picks up gold if in same square
• Releasing drops the gold in same square
 PEAS DESCRIPTION
• Performance measure
• gold: +1000, death: -1000
• -1 per step , -10 for using the arrow
• Environment
• Squares adjacent to Wumpus are smelly
• Squares adjacent to pit are breezy
• Glitter iff gold is in the same square
• Gold is picked up by reflex, can’t be dropped
• Shooting kills Wumpus if you are facing it. It screams
• Shooting uses up the only arrow
• Grabbing picks up gold if in same square
• Releasing drops the gold in same square
• You bump if you walk into a wall
• Actuators: Face , Move, Grab, Release, Shoot
• Sensors: Stench, Breeze, Glitter, Bump, , Scream
WUMPUS WORLD CHARACTERIZATION
1. Deterministic Yes – outcomes exactly specified
2. Static Yes – Wumpus and Pits do not move
3. Discrete Yes
4. Single-agent Yes – Wumpus is essentially a natural feature
5. Fully Observable No – only local perception
6. Episodic No—What was observed before (breezes, pits, etc) is very
useful.
EXPLORING THE WUMPUS WORLD

1. The KB initially contains the rules of the environment.

2. Location: [1,1]
Percept: [¬Stench, ¬Breeze, ¬Glitter, ¬Bump]=[None, None, None, None, None]
Action: Move to safe cell e.g. 2,1
3. Location: [2,1]
Percept: [¬Stench, Breeze, ¬Glitter,¬ Bump]
INFER: Breeze indicates that there is a pit in [2,2] or [3,1]
Action: Return to [1,1] to try next safe cell
EXPLORING THE WUMPUS WORLD

4. Location: [1,2] (after going through [1,1])

Percept: [Stench, ¬Breeze, ¬Glitter, ¬Bump]
INFER: Wumpusis in [1,1] or [2,2] or [1,3]
INFER… stench not detected in [2,1], thus not in [2,2]
REMEMBER….Wumpus not in [1,1]
THUS… Wumpus is in [1,3]
THEREFORE[2,2] is safe because of lack of breeze in [1,2]
Action: Move to [2,2]
REMEMBER: Pit in [2,2] or [3,1]
THEREFORE: Pit in [3,1]!
LOGIC
LOGIC
• The objective of knowledge representation is to express knowledge in a computer-tractable form, so
that agents can perform well.
• Logics are formal languages for representing information such that conclusions can be drawn.
• A formal knowledge representation language is defined by:
• its syntax, which defines all possible sequences of symbols that can be put together to constitute sentences
of the language.
• its semantics, which determines the facts in the world to which the sentences refer. It define the "meaning"
of sentences.

• Each sentence makes a claim about the world. An agent is said to believe a sentence about the world.
• E.g., the language of arithmetic
• x+2 ≥ y is a sentence; x2y +> {} is not a sentence
• x+2 ≥ y is true iff the number x+2 is no less than the number y
• x+2 ≥ y is true in a world where x = 7, y = 1
• x+2 ≥ y is false in a world where x = 0, y = 6
INFERENCING WITH KNOWLEDGE
AND ENTAILMENT
• Inferencing is how we derive:
• Conclusions from existing knowledge;
• New information from existing information. Inferencing might be used in both ASK and TELL
operations.

• Entailment is the generation or discovery that a new sentence is TRUE given existing
sentences. Entailment means that one thing follows logically from another. Entailment is a
relationship between sentences (i.e., syntax) that is based on semantics. Knowledge base KB
entails sentence α if and only if α is true in all worlds where KB is true ie, KB ╞ α
• E.g.
1. KB containing “the Phillies won” and “the Reds won” entails “Either the Phillies won or the
Reds won”
2. x+y = 4 entails 4 = x+y
MODELS
• Logicians typically think in terms of models, which are
formally structured worlds with respect to which truth can be
evaluated.
• We say m is a model of a sentence α if α is true in m.
• M(α) is the set of all models of α, then KB ╞ α iff
M(KB) ╞ M(α)
• E.g.
1. KB = Phillies won and Yankees won
2. α = Phillies won
THE CONNECTION BETWEEN
SENTENCES AND FACTS

• Semantics maps sentences in logic to facts in the world.

• The property of one fact following from another is mirrored by the property of one sentence being entailed
by another.
• If KB is true in the real world, then any sentence α derived from KB by a sound inference procedure is
also true in the real world
ENTAILMENT IN THE WUMPUS WORLD
• Situation after detecting nothing in [1,1], moving right, breeze in [2,1]
• Consider possible models for KB assuming only pits
• 3 Boolean choices⇒ 8 possible models
WUMPUS MODELS

•KB = wumpus-world rules + observations α1 = “[1,2] is safe”

WUMPUS MODELS

•KB = wumpus-world rules + observations

α2 = “there is no pit in [2,2]", KB ╞ α2
SOUNDNESS AND COMPLETENESS
• A sound inference method derives only entailed sentences. Ie, KB ├i α = sentence α can be derived
from KB by inference procedure I
• Soundness: i is sound if whenever KB ├i α, it is also true that KB╞ α
• Analogous to the property of completeness in search, a complete inference method can derive any
sentence that is entailed.
• Completeness: i is complete if whenever KB╞ α, it is also true that KB ├i α
• Preview: we will define a logic (first-order logic) which is expressive enough to say almost
anything of interest, and for which there exists a sound and complete inference procedure. That
is, the procedure will answer any question whose answer follows from what is known by the KB.
PROPOSITIONAL LOGIC
• Propositional Logic also known as simply “Boolean logic”is a method to achieve knowledge representation and
logical inferencing.
• Propositional logic consists of Syntax and Semantics

SYNTAX
• The symbols and the connectives together define the syntax of the language. Again, syntax is like grammar.
• TRUTH SYMBOLS: T (true) and F (false) are provided by the language. Either T or F.
• PROPOSITIONAL SYMBOLS: P, Q, R, etc. mean something in the environment. Proposition symbols are sentences.
• E.g: P means “It is hot”, Q means “It is humid”, R means “It is raining”, “If it is hot and humid, then it is raining”
P^Q=>R
• Syntax can have:
• ATOMIC SENTENCE: Truth and propositional symbols are considered ATOMIC SENTENCES. Atomic sentences must have
truth assigned (i.e., be assigned T or F).
• COMPLEX SENTENCES: More complex sentences are formed using connectives. Sentences formed in this way can be
called Well-Formed Formula (WFF). The evaluation of complex sentences is done using truth tables for the connectives.

SEMANTICS
• Need to be able to evaluate sentences to true or false. The truth tables define the semantics of the language.
LOGICAL CONNECTIVES

• ¬ or NOT or NEGATION: If S1 is a sentence, then ¬S1 is a sentence

• ∧ or AND or CONJUNCTION: If S1, S2 are sentences, then S1 ∧ S2 is a sentence
• ∨ or OR or DISJUNCTION: If S1, S2 are sentences, then S1 ∨ S2 is a sentence
• ⇒ or IFTHEN or IMPLICATION: If S1, S2 are sentences, then S1 ⇒ S2 is a sentence
• ⇔ or IFF or BICONDITIONAL: If S1, S2 are sentences, then S1 ⇔ S2 is a sentence
• Parentheses can be used to indicate precedence.
• KB is conjunction (AND) of all facts.
PROPOSITIONAL LOGIC TRUTH TABLE

P Q ¬P P∧Q P ∨Q P⇒ Q P⇔ Q

False False True False False True True

False True True False True True False

True False False False True False False

True True False True True True True

PRECEDENCE OF OPERATORS
• Just like arithmetic operators, there is an operator precedence when evaluating logical operators as
follows:
1. Expressions in parentheses are processed (inside to outside)
2. Negation
3. AND
4. OR
5. Implication
6. Biconditional
7. Left to right
• Use parentheses whenever you have any doubt!
PROPOSITIONAL LOGIC EXAMPLES
• Example 1: If it is humid, then it is raining.
• P=It is humid. And Q=It is raining.
• It is represented as (P→Q).

• Example 2: It is noon and Ram is sleeping.

• Solution: A= It is noon. And B= Ram is sleeping.
• It is represented as (A Ʌ B).

• Example 3: If it is raining, then it is not sunny.

• Solution: P= It is raining. And Q= It is sunny.
• It is represented as P → ( ~Q)

• Example 4: Ram is a man or a boy.

• Solution: X= Ram is a man. And Y= Ram is a boy.
• It is represented as (X V Y).

• Example 5: I will go to Delhi if and only if it is not humid.

• Solution: A= I will go to Delhi. And B= It is humid.
• It is represented as (A ⇔ ~ B).
HOW CAN WE REPRESENT THE WUMPUS
WORLD?
• We can represent the Wumpus world (things we know and things we discover) in terms of logic
as follows:
• Consider the propositional symbols (partial formulation):
• P(i,j) is T if there is a pit in (I,J), otherwise F.
• B(i,j) is T if there is a breeze in (I,J), otherwise F.
• We can update as we explore:
• ¬B(1,1) – no breeze in square (1,1).
• B(2,1) – breeze in square (2,1).
• ¬P(1,1)- no pit in starting square.
• "Pits cause breezes in adjacent squares"
• B1,1 ⇔ (P1,2 ∨P2,1)
• B2,1 ⇔ (P1,1 ∨P2,2 ∨P3,1)
LOGICAL EQUIVALENCE
• Two sentences are logically equivalent, denoted by α ≡ β iff they are true in the same models,
i.e., iff: α╞ β and β ╞ α.
• If the value of P and Q is true in the same set of models, then they are said to be logically
equivalence.
• It can be used as inference rules in both directions.
Example
• (A⇒ B) ≡ (¬B ⇒ ¬A) (contraposition)
INFERENCE RULES WITH LOGICAL EQUIVALENCES
Rule Name Rule
(A Ʌ A) ≡ A
Idempotency Law
(AV A) ≡ A
(A Ʌ B) ≡ (B Ʌ A)
Commutative Law
(AV B) ≡ (B V A)
~(A Ʌ B) ≡ ~A V ~B)
De morgan’s Law
~(A V B) ≡ (~A Ʌ ~B)
AV(B V C) ≡ (A V B) V C
Associative Law
A Ʌ(B Ʌ C) ≡ (A ɅB) Ʌ C
A Ʌ(B V C) ≡ (A Ʌ B) V (A Ʌ C)
Distributive Law
A V (B Ʌ C) ≡ (A V B) Ʌ (A V C)
Contrapositive Law A → B ≡ ~B → ~A
Implication Removal A → B = ~A V B
Biconditional Removal A ⇔ B = (A → B) Ʌ (B → A)
A Ʌ (A V B) ≡ A
Absorption Law
AV (A Ʌ B) ≡A
Double-negation elimination ~(~A)=A
INFERENCE RULES IN PROPOSITION
LOGIC
• Inference rules are those rules which are used to describe certain conclusions. The inferred
conclusions lead to the desired goal state.
• In propositional logic, there are various inference rules which can be applied to prove the given
statements and conclude them.
 COMMON RULES
1.

2.

3.

4.

Hypothetical Syllogism can be represented as: If (P→Q) Ʌ (Q→R)= (P→R)​

A
5. Introduction: B
AɅ B

AɅ B
6. And Elimination:
A
 VALIDITY AND SATISFIABILITY
• Validity: If a sentence is valid in all set of models, then it is a valid sentence. Validity is also
known as tautology, where it is necessary to have true value for each set of model.
Eg: A V ¬A, A ⇒ A,
• Satisfiability: If a sentence is true atleast for some set of values, it is a satisfiable sentence.
• It can be done by truthtable enumeration.
• (P V Q) → (P Ʌ Q)
P Q PVQ PɅQ (P V Q) → (P Ʌ Q)
False False False False True
False True True False False
True False True False False
True True True True True

• from the above truth table, it is clear that the given expression is satisfiable but not valid.
EXAMPLE 2:

• ((A → B) Ʌ A) → B

A B A→B (A → B) Ʌ A ((A → B) Ʌ A) → B

False False True False True

False True True False True
True False False False True
True True True True True

• the given expression is valid as well as satisfiable.

LOGICAL INFERENCE PROBLEM

• Given a knowledge base KB (a set of sentences) and a sentence (called a theorem). Does a
KB semantically entail ? In other words in all interpretations in which sentences in the KB
are true, is also  true? Ie, KB |= α ?
• Three approaches:
• Truth-table approach
• Deduction using Inference rules
• Proof by Contradiction or Resolution-refutation
DEDUCTION THEOREM & PROOF BY
CONTRADICTION
Deduction Theorem (connects inference and validity)
• KB ╞ α if and only if KB⇒α is valid
Proof By Contradiction or Refutation or reductio ad absurdum
• KB ╞ a is valid if and only if the sentence KBɅ¬α is a contradiction.
• Monotonic
• If we have a proof, adding information to the DB will not invalidate the proof ie set of entailed
sentences can only increase information to KB.
DEDUCTION EXAMPLE

• P: “It is hot”, Q : “It is humid” and R : “It is raining”. (SYMBOLS).

• Given KB as:
1. “If it is hot and humid, then it is raining”: P^Q=>R
2. "If it is humid, then it is hot": Q=>P
3. “It is humid”: Q

• Question: Is it raining? (i.e., is R entailed by KB?)

SOLUTION
CHALLENGE

• Given KB.
• PɅ Q
• P→ R
• QɅ R →S

• Can you conclude S

SOLUTION
PROOF BY CONTRADICTION
• Assume our conclusion is false, and look for a contradiction. If found, the opposite of our
assumption must be true.
1. Arrive at conclusion R
1 PvQ
2 P→R
3 Q→R
SOLUTION

Step Formula Derivation

1 PvQ Given
2 P→R Given
3 Q→R Given
4 ¬R Negated Conclusion
5 QvR 1,2
6 ¬P 2,4
7 ¬Q 3,4
8 R 5,7
9 F 4,8
FORMALIZING THE WW IN PL
• The Wumpus World knowledge base:
• There is no pit in [1, 1] (agent percept): R1 : P11
• A square is breezy if and only if there is a pit in a neighboring square. (Rule of the WW).
We state this for the square B11 only: R2 : B1,1  (P1,2 P2,1)
• There is no breeze in square [1; 1]. (agent percept) R3 :  B11
• The agent can now use the PL inference rules and logical equivalences to prove the
following: There is no pit in squares [1,2] or [2, 1]
• Theorem:  P12 Ʌ  P21
FORMALIZING THE WW IN PL
• Apply biconditional elimination to R2:

• R4 : (B11 (P12  P21)) Ʌ ((P12  P21)  B11)

• Apply And-elimination to R4:

• R5 : (P12  P21)  B11

• Apply logical equivalence for contrapositives to R5:

• R6 :  B11  (P12  P21)

• Apply modus ponens to R6 and R3:

• R7 :  (P12  P21)
• Apply de Morgan's rule to R7:

• R8 :  P12 Ʌ  P21
KB IN RESTRICTED FORMS
• If the sentences in the KB are restricted to some special forms some of the sound inference
rules may become complete
• Example:
• Horn form (Horn normal form)
• CNF (Conjunctive Normal Forms)
 PROPOSITIONAL THEOREM PROVING
• Search for proofs is a more efficient way than enumerating models (We can ignore irrelevant
information). Truth tables have an exponential number of models.
• The idea of inference is to repeat applying inference rules to the KB.
• Inference can be applied whenever suitable premises are found in the KB
• Theorem proving means to apply rules of inference directly to the sentences.
• Two ways to ensure completeness:
1. Proof by resolution: use sequence of powerful inference rules (resolution rule) and construction
of / search for a proof. Resolution works best when the formula is of the special form CNF.
Properties
• Typically requires translation of sentences into a normal form.
2. Forward or Backward chaining: use of modus ponens on a restricted form of propositions (Horn
clauses)
NORMAL FORMS
• Literal: A literal is an atomic sentence (propositional symbol), or the negation of an atomic
sentence. Eg:- p (positive literal), ¬p (negative literal)
• Clause: A disjunction of literals. Eg:- ¬p ∨ q
• Conjunctive Normal Form (CNF): A conjunction of disjunctions of literals, i.e., a conjunction
of clauses Eg:- (AV¬B) ^ (BV¬ CV ¬ D)
• DNF( Disjunctive Normal Form): This is a reverse approach of CNF which is disjunction of
conjunction of literals. Eg:- (A1 Ʌ B1) V (A2 Ʌ B2) V…V (An Ʌ Bn)
• In DNF, it is OR of AND’s, a sum of products, or a cluster concept, whereas, in CNF, it is
ANDs of OR’s a product of sums.
CNF TRANSFORMATION
• In propositional logic, the resolution method is applied only to those clauses which are
disjunction of literals. There are following steps used to convert into CNF:
1) Eliminate bi-conditional implication by replacing A ⇔ B with (A → B) Ʌ (B →A)
2)Eliminate implication by replacing A → B with ¬A V B.
3) In CNF, negation(¬) appears only in literals, therefore we move negation inwards as:
¬ ( ¬A) ≡ A (double-negation elimination
¬ (A Ʌ B) ≡ ( ¬A V ¬B) (De Morgan)
¬(A V B) ≡ ( ¬A Ʌ ¬B) (De Morgan)

4) Finally, using distributive law on the sentences, and form the CNF as:
(A1 V B1) Ʌ (A2 V B2) Ʌ …. Ʌ (An V Bn).

• Note: CNF can also be described as AND of ORS

• Transform to CNF: B1,1  (P1,2  P2,1)
CNF TRANSFORMATION EXAMPLE
B1,1  (P1,2  P2,1)
1. Eliminate , replacing α  β with (α  β)(β  α).
(B1,1  (P1,2  P2,1))  ((P1,2  P2,1)  B1,1)

2. Eliminate , replacing α  β with α β.

(B1,1  P1,2  P2,1)  ((P1,2  P2,1)  B1,1)

3. Move  inwards using de Morgan's rules and double-negation:

(B1,1  P1,2  P2,1)  ((P1,2  P2,1)  B1,1)

4. Apply distributivity law ( over ) and flatten:

(B1,1  P1,2  P2,1)  (P1,2  B1,1)  (P2,1  B1,1)
 METHOD 1: RESOLUTION METHOD IN FOL
• In propositional logic, resolution method is by application of inference rule gives a new clause
when two or more clauses and are coupled together to prove theorem.
• Using propositional resolution, it becomes easy to make a theorem prover sound and complete for
all. The process followed to convert the propositional logic into resolution method is known
as Resolution refutation contains the below steps:
1. Convert the given axiom(all sentences) into clausal form, CNF.
2. Negate the desired conclusion (converted to CNF)
3. Apply resolution rule until either – Derive false (a contradiction) – Can’t apply any more
4. If we derive a contradiction, then the conclusion follows from the axioms
5. If we can’t apply any more, then the conclusion cannot be proved from the axioms.
• This is known as resolution Algorithm.
• Resolution refutation is sound and complete.
EXAMPLE

• Prove R from:

1 (P → Q) → Q
2 (P → P)→ R
3 (R → S) → ¬ (S → Q)
SOLUTION
• Convert to CNF
1 PvQ
1. (P → Q) → Q  ¬(¬ P v Q) v Q
2 PvR
(P ¬ Q) v Q
3 ¬PvR
(P v Q) Ʌ(¬ Q v Q)
(P v Q) Ʌ T 4 RvS
5 Rv¬Q
2. (P → P)→ R  ¬(¬ P v P) v R
6 ¬Sv¬Q
 (P ¬ P) v R
7 ¬R Neg
 (P v R)Ʌ (¬ P v R)
8 S 4,7
3. (R → S) → ¬ (S → Q)  ¬(¬ R v S) v ¬ (¬ S v Q)
9 ¬Q 6,8
 (R Ʌ ¬ S) v (S Ʌ ¬ Q)
10 P 1,9
(R v S) Ʌ (¬ S v S) Ʌ (R v ¬ Q) Ʌ (¬ S v ¬ Q)
11 R 3,10
 (R v S) ɅT Ʌ (R v ¬ Q) Ʌ(¬ S v ¬ Q)
12 F 7,11
PROPOSITIONAL RESOLUTION EXAMPLE
• Consider the following Knowledge Base:
1. The humidity is high or the sky is cloudy.
2. If the sky is cloudy, then it will rain.
3. If the humidity is high, then it is hot.
4. It is not hot.

• Goal: It will rain.

• Use propositional logic and apply resolution method to prove that the goal is derivable from the
given knowledge base.
 SOLUTION
• Solution: Let’s construct propositions of the given sentences one by one:
Let, P: Humidity is high. Q: Sky is cloudy. R: It will rain S: It is hot.
1. It will be represented as P V Q.
2. It will be represented as Q → R.
3. It will be represented as P → S.
4. It will be represented as ¬S.
CHALLENGES

1. Given KB = (B1,1  (P1,2  P2,1)) Ʌ  B1,1 Prove :  P1,2

SOLUTION

Given KB
• R1: (¬B1,1  P1,2  P2,1)  (¬P1,2  B1,1)  (¬P2,1  B1,1)
• R2:  B1,1
• R3: Negation of theorem=  ( P1,2)= P1,2
• Given R1 can be split up as R4: : (¬B1,1  P1,2  P2,1) R5: (¬P1,2  B1,1) R6: (¬P2,1  B1,1)
• Consider R5 and R2 apply Modus Ponens R6: ¬P1,2
• Consider R6 and R3 which leads to a negation
 HORN CLAUSES AND DEFINITE CLAUSES
• DEFINITE CLAUSE: A disjunction of literals of which exactly one is positive.
• (L1,1   breeze  B1,1 ) Yes
• ( B1,1  P1,2  P2,1 ) No
• HORN CLAUSE: A disjunction of literals of which atmost one is positive, ie is a CNF clause with
exactly one positive literal. The positive literal is called the head. The negative literals are called the
body. All definite clauses are Horn Clauses.
• (L1,1  breeze  B1,1 ) Yes
• ( B1,1  P1,2  P2,1 ) Yes
• ( B1,1  P1,2  P2,1 ) No

• Horn clauses are closed under resolution, ie if 2 Horn closes are resolved we get back a horn clause.
• Not all sentences in propositional logic can be converted into the Horn form
• GOAL CLAUSE: A clause with no positive literal.
• (L1,1  breeze  B1,1 ) No
• ( B1,1  P1,2  P2,1 ) Yes
HORN CLAUSES
• Horn clauses can be re-written as implications ie, logic proposition of the form: p1 ^
….. ^ pn → q .
• Eg:  C   B  A can be written as C B → A
• KB = conjunction of Horn clauses.
• Modus Ponens (for Horn Form

• Inference with Horn Clauses can be done using forward and backward chaining
algorithms.
• The Prolog language is based on Horn Clauses.
• Deciding entailment with Horn Clauses is linear in the size of the knowledge base.
FORWARD AND BACKWARD CHAINING
• These algorithms are very natural and run in linear time
FORWARD CHAINING:
• Based on rule of modus ponens. If know P1, …,Pn& know (P1 Ʌ... ɅPn)→Q. Then can conclude Q.
Whenever the premises of a rule are satisfied, infer the conclusion. Continue with rules that became
satisfied.
• Forward chaining is also known as a forward deduction or forward reasoning method when using an
inference engine. Forward chaining is a form of reasoning which start with atomic sentences in the
knowledge base and applies inference rules (Modus Ponens) in the forward direction to extract more
data until a goal is reached.
BACKWARD CHAINING:
• In Backward chaining, we will start with our goal predicate and then infer further rules.
• Search start from the query and go backwards.
FORWARD CHAINING
• IDEA: It begins from facts(positive literals) in knowledge base and determines if the query can be entailed by
knowledge base of definite clauses. If all premises of an implication are known its conclusion is added to set of
known facts. Eg: Given L1,1 and Breeze and (L1,1 Ʌ Breeze) → B1,1 is in knowledge base then B1,1 can be added.
p1,……, pn p1 ^ ….. ^ pn → q
• Every inference is an application of modus ponens ie Can be used with forward
q
chaining.
FORWARD CHAINING STEPS
1. Start with given proposition symbols (atomic sentence).
2. Iteratively try to infer truth of additional proposition symbols
3. Continue until
– no more inference can be carried out, or
– goal is reached
 FORWARD CHAINING
• Fire any rule whose premises are satisfied in the KB, add its
conclusion to the KB, until query is found
• AND-OR graph: Multiple links joined by an arc indicates
a conjunction where every link has to be proved, while
multiple links without an arc indicates disjunction, where
any link has to be proved.

AND-OR
GRAPH
FORWARD CHAINING

• Process agenda item B

• Process agenda item A • Decrease count for horn clauses
• Decrease count for horn in which B is premise
clauses in which A is premise • A ^ B → L has now fulfilled
premise
• Add L to agenda
 FORWARD CHAINING

• Process agenda item L • Process agenda item M • Process agenda item P

• Decrease count for horn clauses in • Decrease count for horn clauses in • Decrease count for horn clauses in
which L is premise which M is premise which P is premise
• B ^ L → M has now fulfilled • L ^ M→P has now fulfilled • P → Q has now fulfilled premise
premise premise • Add Q to agenda
• Add M to agenda • Add P to agenda • A ^ P → L has now fulfilled
premise
FORWARD CHAINING

• Process agenda item P • Process agenda item Q

• Decrease count for horn clauses in which P is premise • Q is inferred
• P → Q has now fulfilled premise • Done
• Add Q to agenda
• A ^ P → L has now fulfilled premise
• But L is already inferred
FORWARD CHAINING CHALLENGE
SOLUTION
BACKWARD CHAINING
• Idea: Works backwards from the query q
• to prove q by Backward Chaining:
• Check if q is known already, or
• Prove by Backward Chaining all premises of some rule concluding q
• Avoid loops: check if new subgoal is already on the goal stack
• Avoid repeated work: check if new subgoal
• has already been proved true, or has already failed
 BACKWARD CHAINING

• Current goal: Q • Current goal: P

• A and B are known to be true
• Q can be inferred by P → Q • P can be inferred by L ^ M → P
• Q needs to be proven
• P needs to be proven • L and M need to be proven
BACKWARD CHAINING

• Current goal: L • Current goal: L

• L can be inferred by A ^ P → L • Current goal: L
• A is already true • L can be inferred by A^ B → L
• P is already a goal • Both are true
• repeated sub-goal
BACKWARD CHAINING

• Current goal: L
• L can be inferred by A ^ B → L • Current goal: M • Current goal: M
• Both are true • M can be inferred by B^L→ M • M can be inferred by B ^ L→ M
• L is true • Both are true
• Current goal: M • M is true
BACKWARD CHAINING

• Current goal: P • Current goal: Q

• P can be inferred by L ^ M → P • Q can be inferred by P → Q
• Both are true • P is true
• P is true • Q is true
FORWARD VS BACKWARD
Forward chaining:
• Data-driven, automatic, unconscious processing.
• May do lots of work that is irrelevant to the goal
Backward chaining:
• Goal-driven, appropriate for problem-solving.
• Complexity of BC can be much less than linear in size of KB
DISADVANTAGES OF PL
• Consider now the following WW rule: If a square has no smell, then neither the square nor any of its adjacent squares can
house a Wumpus. How can we formalize this rule in PL?
• We have to write one rule for every relevant square! For example:  S11   W11 ^  W12 ^W21
• For an example having large environment say we have a vacuum cleaner (Roomba) to clean a 1010 squares in the
classroom. Use PL to express information about the squares.
• This is a very disappointing feature of PL. There is no way in PL to make a statement referring to all objects of some kind
(e.g., to all squares).
LIMITATION
1. PL is not expressive enough to describe all the world around us. It can't express information about different object and
the relation between objects.
2. PL is not compact. It can't express a fact for a set of objects without enumerating all of them which is sometimes
impossible.
3. Propositional logic is declarative: pieces of syntax correspond to facts
• Not to worry: this can be done in First order logic!
anooja@somaiya.edu

THANK YOU