Uncertain Knowledge and Reasoning
In artificial intelligence (AI), uncertain knowledge and reasoning are essential because, in real-
world scenarios, information is often incomplete or imprecise. There are various methods and
frameworks to handle uncertainty, allowing machines to reason and make decisions in such
environments.
1. Basic Probability
Probability theory is a mathematical framework used to quantify uncertainty. In the context of
AI, probability provides a way to model the likelihood of different outcomes based on available
evidence.
Probability Distribution: A probability distribution is a function that describes the
likelihood of different outcomes in an experiment or process.
Conditional Probability: This refers to the probability of an event occurring given that
another event has occurred. The notation is P(A∣B)), which is the probability of A given
B.
Bayes’ Theorem is a key concept in probability theory:
P(A∣B)=P(B∣A)P(A)/P(B)
Where:
P(A∣B): Posterior probability (the probability of A given B
P(B∣A): Likelihood (the probability of B given A
P(A) Prior probability (the initial probability of A)
P(B): Marginal likelihood (the total probability of B)
2. Bayes' Rule
Bayes' rule is a fundamental result in probability theory that describes how to update the
probability of a hypothesis, H, based on new evidence E. This rule is particularly useful in
situations of uncertain knowledge.
In AI, Bayes’ Rule is often used for:
Updating beliefs: As new data or evidence comes in, Bayes' Rule helps adjust the
probability of different hypotheses.
Machine Learning: Naive Bayes classifiers use Bayes' Rule to classify data points into
categories based on prior probabilities and the likelihood of the data.
3. Belief Networks (Bayesian Networks)
A belief network (or Bayesian network) is a graphical model that represents a set of variables
and their probabilistic dependencies. Nodes represent variables, and edges between nodes
represent conditional dependencies.
Nodes: Variables (can be random or deterministic).
Edges: Conditional dependencies between variables.
CPT (Conditional Probability Table): Each node has an associated table showing the
conditional probability distribution of the variable, given its parents in the network.
Belief networks allow efficient computation of joint distributions and reasoning about
uncertainty. They are used in areas such as:
Medical diagnosis: Predicting diseases based on symptoms.
Risk management: Calculating the risk associated with different decisions.
4. Default Reasoning
Default reasoning is a method of reasoning that deals with the assumption of typical or
common situations in the face of incomplete knowledge. When there is no direct evidence, the
system assumes the most likely scenario.
For example:
Default assumption: Birds fly. If no information is available to contradict this (e.g., a
penguin is a bird), the system assumes the bird can fly.
However, default reasoning is not always reliable since exceptions exist (e.g., penguins and
ostriches).
5. Fuzzy Sets and Fuzzy Logic
Fuzzy sets and fuzzy logic are used to handle reasoning that involves approximate or imprecise
information, unlike classical binary logic where a statement is either true or false.
Fuzzy Set: A fuzzy set allows partial membership. An element can belong to a set to a
certain degree, which is expressed as a value between 0 and 1.
o Example: "Tall" could be represented as a fuzzy set, where a person 6 feet tall
might have a membership value of 0.9 in the "Tall" set, while a person 5 feet 5
inches tall might have a membership value of 0.3.
Fuzzy Logic: Fuzzy logic extends classical logic by allowing for truth values to range
between 0 and 1 (rather than just true/false). Fuzzy inference systems use fuzzy sets
and logic to make decisions or infer conclusions from imprecise input data.
o Fuzzy Rules: Rules like "IF temperature IS high THEN fan speed IS fast" can
be interpreted in a fuzzy manner, meaning the temperature can be somewhat
high and the fan speed somewhat fast.
Fuzzy logic is widely used in control systems, such as thermostats, traffic control, and
automatic braking systems.
Structured Knowledge
Structured knowledge refers to knowledge representation methods that explicitly define the
relationships and attributes of objects, events, or concepts. Structured knowledge allows for
better organization and reasoning about the relationships between elements.
1. Associative Networks
An associative network is a structure that represents concepts as nodes and relationships
between those concepts as edges. Each node represents an object or concept, and the edges
define the relationships between them.
Semantic Networks: A form of associative network where nodes represent concepts
and edges represent semantic relationships (e.g., "is a," "has part," "can fly").
Associative networks allow for efficient information retrieval and reasoning, as they can model
how concepts are related and help infer new information based on existing knowledge.
2. Frame Structures
Frames are data structures used to represent knowledge in AI. They represent "schemes" or
"templates" for various concepts or objects and define their attributes and the relationships
between those attributes.
A frame is composed of:
o Slots: Attributes or properties of the concept.
o Facets: Rules for filling the slots or additional constraints on the values of the
slots.
Frames are used to represent structured information, like a "car" frame that might include slots
for "engine type," "color," and "wheels."
Example: A frame for "Person" might include slots like "name," "age," "address," and "phone
number."
3. Conceptual Dependencies
Conceptual dependency theory is a representation of the meaning of a sentence based on its
conceptual components, rather than its syntactic structure. It focuses on the underlying
concepts and the relationships between them.
Example: The sentence "John gave Mary a book" can be represented as:
o Agent: John
o Action: Gave
o Recipient: Mary
o Object: Book
This kind of representation abstracts away from linguistic differences and focuses on the core
meaning of actions and objects.
4. Scripts
Scripts are structures used to represent typical sequences of events or actions that occur in a
particular context. They represent knowledge about routines or sequences of events that are
expected in specific situations.
Example: A "Restaurant Script" might include actions like "enter restaurant," "order
food," "eat food," and "pay bill." The script helps the AI reason about what typically
happens in a restaurant.
Scripts are particularly useful for common sense reasoning and understanding scenarios where
multiple events unfold in a predictable manner.
