Invitation to Cognitive Science. A Reading Memo

Machine Learning Artificial Intelligence Natural Language Processing Semantic Web Ontology Knowledge Information Processing Digital Transformation Probabilistic Generative Model Deep Learning Navigation of this blog

Invitation to Cognitive Science. A Reading Memo

Cognitive science will be the discipline that studies cognitive processes such as human perception, thought, language, learning, memory, and decision making. Cognitive science is located at the intersection of diverse disciplines and integrates knowledge from psychology, neuroscience, computer science, philosophy, linguistics, and artificial intelligence studies.

By elucidating human cognitive processes, cognitive science aims to understand the mechanisms of perception and thought, and to develop theories about human behavior, decision-making, and language acquisition and processing. Cognitive science research results also play an important role in examining the differences and similarities between human and machine intelligence.

Some of the major areas of research in cognitive science include the following

• Memory research Research on how humans store information and on memory retention and reproduction.
• Language research Research on the acquisition and processing of human language. This includes grammar, understanding of meaning, and mechanisms of speech and listening.
• Research on decision-making Research on the mechanisms by which humans make decisions and the factors that influence these decisions.
• Research on the senses Research on the mechanisms of the human senses and perception.
• Attention research Research on the mechanisms of human attention and how we direct and switch our attention.

Cognitive science, through its many ongoing studies, will reveal the diversity and complexity of human cognitive functions and provide useful insights into human intellectual capacities.

In this issue, we present reading notes from this reference book on cognitive science, Invitation to Cognitive Science.

Cognitive Science, which has developed along with Artificial Intelligence, helps us understand how thinking works and how the brain works. How do machines understand logic and reasoning? How is the human brain different from machines? The author, who studied cutting-edge cognitive science at an American graduate school, explains in an easy-to-understand manner. He even approaches the author’s new theory, the “Super Information Field Hypothesis.

Chapter 1: I and Cognitive Science

Cognitive Science is the Paradigm that Replaced Structuralism
Structuralism
Provide external stimuli, observe the responses, and probe the internal structure of the mind from there
Difficult to infer the interior from the relationship between inputs and outputs to a black box
Cognitive science
Focus on “functions” of the mind and brain
Focus on how the mind and brain work rather than how they work
Functionalism
A function that is inherent to the mind or brain
Expressed in terms of “function
Encounter with cognitive science
The Cognitive Computer
Roger Shank
Cognitive
The human mind can be replaced by abstract descriptions
Noam Chomsky’s linguistics
There is a universal grammar in the universe (like a mathematical theorem)
Change the value of a variable and it becomes English or Japanese
Language barriers can also be clarified by mathematics
Roger Shank and Marvin Minsky
Shank.
CD theory (Conceptual Dependency)
Semantic representations depend on concepts
If there is a linguistic representation that has the same meaning, it will be represented the same way on a computer
Script Theory
MOP Theory
Case-Based Theory
XP Theory
Marvin Minsky
Frame Theory
People perceive things in a kind of pattern called a “frame”
Learn from Karl Pollard
Robin Clark
Karl Pollard
Head-Driven Phrase Structure Grammar (HPSG)
Connectionists in Action
Connectionist
Neural networks can learn grammar by learning
Syntax
If you know the grammar, you know the meaning of the sentence
Chomsky
Semantics
If you don’t know the meaning, you can’t determine the structure of a sentence
Shank

Chapter 2: Cognitive Science and Strong AI

Behaviorism and Functionalism
Artificial Intelligence has developed from a variety of disciplines, including not only computer science, but also psychology, linguistics, philosophy, brain medicine, and mathematics. It is a development intertwined with various disciplines such as psychology, linguistics, philosophy, brain medicine, mathematics, and so on.
S-R Model
S:Stimulus, R:Responce (stimulus-response model)
PS Model
Production System Model
Collects rules called production rules and applies them to make the artificial intelligence reason.
Inference is performed by accumulating patterns such as Frame Theory
Minsky’s “Frame Theory
Frame: A unit of knowledge and its combination method to represent a certain knowledge
In order to have an artificial intelligence understand some kind of tree, several knowledge (elements) related to the knowledge are combined to form a “frame”.
Knowledge is understood based on whether it fits into the “frame” or not.
Example: restaurant
People recognize events in the lower world by evoking frames
Let a computer perform frame inference
If some elements of a frame match, it is considered to be that frame
Shank’s “script theory
A methodology for recognizing procedural sequences of human actions
Script: a script, a sequence of actions associated as a single entity based on a scenario
Example: restaurant
Shank’s “CD Theory
CD: Conceptual Dependency
How to make Artificial Intelligence understand semantic representation
The core of a semantic representation system is always the representation of events
The structure for representing the core of an event
An EVENT
An ACTOR
An ACTION (action) performed by that action
A concrete basic action
An important basic action
ATRANS (transfer of possession)
MTRANS (transfer of information)
INGEST (internalization)
PTRANS (movement of physical objects)
MBUILD (formation of information)
PROPEL (imposition of power)
Means of action
SPEAK (Speak)
GRASP (Grasping an object)
EXPEL (move it out of the body)
ATTEND (directing attention)
MOVE (move the body)
An OBJECT to which the action is directed
A DIRECTION (direction) to which the action is directed
Each element is called a “slot,” and assigning a value to a “slot” is called “filling a slot.
EXAMPLE
Hohn amazed Mary
ACTOR” John
amaze” is not “ACTION”, but amaze because you did some action
Example 2
John saw Mary
ACTOR John
ACTION” MTRANS
OBJECT” Mary
Scripts that complement CD Theory
Script” theory takes context into account
When sentences contain cause-and-effect relationships, it is difficult to answer questions that require reading the context
Example.
John was hungry. He went into Goldstein’s and ordered a sandwich. The order came quickly. He tipped the waitress a lot.
Question
What is Goldstein’s?
What did John eat?
Who made the sandwich
Who took John’s order
Who brought the sandwiches
Why did John tip so much?
Can be interpreted as having a script behind it
Programs with pre-built script knowledge
SAM: Script Applier Mechanism
Something that can ask and answer questions and make inferences
PAM: Plan Applier Mechanism
Incorporates context
MOP Theory and XP Theory
Is it possible for humans to understand the information that constitutes human intelligence?
Even if it can be clarified, can it be described by humans?
CD theory is the analogy for atomistic units and script theory for molecular units
Dynamic Memory
Shank’s book
The story of the description of human knowledge is brought into the human brain
Humans have different experiences from orphan to orphan and process personal experiential knowledge as learning
Mass of experiential knowledge
Memory Organization Packet MOP
When something goes wrong, humans try to explain it with a hypothesis, which becomes an index and a memory
Explanation Patterns:XP
Case-Based Theory” and After Shank
Scripts and MOPs are replaced by the idea of “cases
Scripts and MOPs are pre-taught, but cases are self-learned
Case-Based reasoning: CBR Case-Based Theory
Minimum concepts are necessary, such as how case explanation patterns are created
Abstraction is necessary
To match patterns
Reuse of past “cases” and “Explanation Patterns
Tweekung: Tweaking
E-learning and military applications

Chapter 3: Artificial Intelligence and Neural Networks

Neural Networks in Artificial Intelligence
Learnable Artificial Intelligence
Backpropagation”, a broadening of learning in Artificial Intelligence
How to avoid local solutions
A Neural Net called PDP Model
Society of the Mind
Combination of agents that make up the mind
Subsumption? Translated with www.DeepL.com/Translator (free version)

Chapter 4: Grammar First or Meaning First?

Syntax and Semantics
Grammars were created when neural nets were analyzed
Karl Pollard’s HPSG
HPSG
Decompose sentences into a tree structure with phrases and words as units
Decompose phrases and words from the bottom to the top with the main part called “Head
Words are not just symbols for meaning, but contain a lot of information within them.
This information is intertwined with other words and phrases, and the same elements can be “unified.
Words and phrases can contain contextual information
Functions and Arguments
Arguments
Parameters
Variables in a function
If f(x,y)=x+y, then x and y are parameters
Real argument
If x=1,y=2, then it is a real argument
Natural language is also considered a function
Natural language is a world of higher-order predicate logic
HPSG and Higher-Order Predicate Logic
HPSG “unifies” using attributes of words
HPSG realizes higher-order predicate logic
HPSG is analyzed according to structures called “partial information structures
Partial information structures
Structures in which function information is written on several attributes of a word.
Examples
PHON: Abbreviation for phonology
SEM: SEMANTICS (semantics)
Tomata algorithm allows fast unification
Soft unification” using probabilistic models of non-monotonic logic
Egg first or chicken first?

Chapter 5: Beyond Cognitive Science

The “Frame Problem” that Cognitive Science Cannot Solve
How do humans know that this is a restaurant?
How to climb the stairs of abstraction
The “Super Information Field Hypothesis” for Overcoming the Paradigm of Cognitive Science
The “Super Information Field Hypothesis” for “Restaurant Frames
Religious People and the Super Information Field
What is Reality?
Reality is one of the ways to share the super-information field
The alternative paradigm to “Functionalism” is “Homeostasis