Knowledge Representation and
Reasoning
(Introduction)
Content
• Historical background – What is AI & representation?
• Reasoning as manipulation of representations.
• Knowledge Representation Hypothesis
• Reflection Hypothesis
• The Need for a Good Representation
• Key Concepts: Knowledge, Representation, and
Reasoning
• Knowledge-Based Systems
• Why Knowledge? and Why Reasoning?
• The Role of Logic
What is Artificial Intelligence ?
u Artificial intelligence is the study of systems that act in a
way that to any observer would appear to be intelligent.
u Artificial Intelligence involves using methods based on the
intelligent behavior of humans and other animals to solve
complex problems.
u AI is concerned with real-world problems (difficult tasks),
which require complex and sophisticated reasoning
processes and knowledge.
3
Artificial Intelligence
John McCarthy coined the term
“Artificial Intelligence” in 1956.
• Pioneered the automation of
common-sense reasoning
• Invented the Lisp programming
Language
Professor Emeritus of
Computer Science at
• Invented time-sharing
Stanford University
Provisional definition of intelligence
Notice that according to this view,
given some goal, it should be
achieved by some manipulations (i.e.
computations ?) applied to some
representation.
Prof. Aaron Sloman,
University of Birmingham The idea of representation has
a mental component —
“Intelligence consists of the
principled manipulation of “X represents Y for Z”.
representations, in order to
achieve some goal.”
Where are the representations?
In natural intelligence, we get a picture as follows:
Knowledge Representation Hypothesis
A successful representation of some knowledge must:
1. be in a form that is “understandable” by humans (i.e. a
“linguistic” form);
2. cause the system using the knowledge to “behave” as if it
has such knowledge.
This gives the following picture.
Reflection Hypothesis
Again, from Brian Smith
“Inasmuch as a computational process can be constructed to
reason about an external world in virtue of comprising an
ingredient process (interpreter) formally manipulating
representations of the world.
So, too a computational process could be made to reason about itself in
virtue of comprising an ingredient process (interpreter) formally
manipulating representations of its own operations and structures.”
A doctor can diagnose his own disease using diagnosis knowledge that s/he
owns.
Reflection Hypothesis
In other words:
If we can build a system to reason about the world by manipulating
representations, then we can build a system to reason about itself, using a
representation of itself.
Here, we can think of an agent that can reason about some properties of itself.
E.g. “if I believe that plan (action) P achieves G, and if G is a goal of mine, then I
shall execute P .”
So, procedural control knowledge can become declarative in the reflective
system.
Reflection Hypothesis
This gives:
Self
The Need for a Good Representation
• Language to represent Knowledge.
• Inference mechanism to use Knowledge 12
What is knowledge?
Easier question: how do we talk about it?
We say “John knows that ...” and fill the blank with a proposition
– can be true / false, right / wrong
Contrast: “John fears that ...”
– same content, different attitude
Other forms of knowledge:
• know how, who, what, when, ...
• sensorimotor: typing, riding a bicycle
• affective: deep understanding
Belief: not necessarily true and/or held for appropriate reasons
and weaker yet: “John suspects that ...”
4
Why knowledge?
For sufficiently complex systems, it is sometimes useful to describe systems
in terms of beliefs, goals, fears, intentions
e.g. in a game-playing program
“because it believed its queen was in danger, but wanted to still control the center
of the board.”
more useful than description about actual techniques used for deciding how
to move
“because evaluation procedure P using minimax returned a value of +7 for this
position
= taking an intentional stance (Dan Dennett)
Is KR just a convenient way of talking about complex systems?
• sometimes anthropomorphizing is inappropriate 6
e.g. thermostats
• can also be very misleading!
fooling users into thinking a system knows more than it does
Why representation?
Note: intentional stance says nothing about what is or is not
represented symbolically
e.g. in game playing, perhaps the board position is represented, but the goal of getting
a knight out early is not
KR Hypothesis: (Brian Smith)
“Any mechanically embodied intelligent process will be comprised of structural
ingredients that a) we as external observers naturally take to represent a propositional
account of the knowledge that the overall process exhibits, and b) independent of such
external semantic attribution, play a formal but causal and essential role in engendering
the behaviour that manifests that knowledge.”
Two issues: existence of structures that
• we can interpret propositionally
• determine how the system behaves
Knowledge-based system: one designed this way! 7
Query: What is the color of grass?
KR and AI
Much of AI involves building systems that are knowledge-based
ability derives in part from reasoning over explicitly represented knowledge
– language understanding,
– planning,
– diagnosis,
– “expert systems”, etc.
Some, to a certain extent
game-playing, vision, etc.
Some, to a much lesser extent
speech, motor control, etc.
Current research question:
how much of intelligent behaviour is knowledge-based?
9
Challenges: connectionism, others
Why bother?
Why not “compile out” knowledge into specialized procedures?
• distribute KB to procedures that need it
(as in Example 1)
• almost always achieves better performance
No need to think. Just do it!
– riding a bike
– driving a car
– playing chess?
– doing math?
– staying alive??
Skills (Hubert Dreyfus)
• novices think; experts react
10
• compare to current “expert systems”:
knowledge-based !
Advantage
Knowledge-based system most suitable for open-ended tasks
can structurally isolate reasons for particular behaviour
Good for
• explanation and justification
– “Because grass is a form of vegetation.”
• informability: debugging the KB
– “No the sky is not yellow. It's blue.”
• extensibility: new relations
– “Canaries are yellow.”
• extensibility: new applications
– returning a list of all the white things 11
– painting pictures
Cognitive penetrability
Hallmark of knowledge-based system:
the ability to be told facts about the world and adjust our behaviour
correspondingly
for example: read a book about canaries or rare coins
Cognitive penetrability (Zenon Pylyshyn)
actions that are conditioned by what is currently believed
an example:
we normally leave the room if we hear a fire alarm we do not leave
the room on hearing a fire alarm
if we believe that the alarm is being tested / tampered
can come to this belief in very many ways
so this action is cognitively penetrable a non-example:
blinking reflex 12
P1: "All cats are mammals."
P2: "All mammals are animals."
P3: "Whiskers is a cat."
Sorts of knowledge
1. Knowing that (declarative knowledge):
I know that Abu Dhabi is the capital of the UAE
2. Knowing how (procedural knowledge):
I know how to drive to Abu Dhabi!
How to look-up capital cities in an Atlas.
Procedural knowledge is harder to formalise than declarative
knowledge.
Using logic
No universal language / semantics
• Why not English?
• Different tasks / worlds
• Different ways to carve up the world
No universal reasoning scheme
• Geared to language
• Sometimes want “extralogical” reasoning
Start with first-order predicate calculus (FOL)
• invented by philosopher Frege for the formalization of mathematics
• but will consider subsets / supersets and very different looking
15
representation languages
Knowledge level
Allen Newell's analysis:
• Knowledge level: deals with language, entailment
• Symbol level: deals with representation, inference
Picking a logic has issues at each level
• Knowledge level:
expressive adequacy, theoretical complexity, ...
• Symbol level:
architectures, data structures,
algorithmic complexity, ...
Next: we begin with FOL at the knowledge level
Thank You,,,,,,,