"This is the first introductory text in cognitive science that provides comprehensive coverage of the field. . .[it] provides a solid foundation for the beginning student". -- Arthur C. Graesser, "Contemporary Psychology" "Cognitive Science" is a single-source undergraduate text that broadly surveys the theories and empirical results of cognitive science within a consistent computational perspective. In addition to covering the individual contributions of psychology, philosophy, linguistics, and artificial intelligence to cognitive science, the book has been revised to introduce the connectionist approach as well as the classical symbolic approach and adds a new chapter on cognitively related advances in neuroscience.
- Information processes are contentful and purposeful.
- Information processes are representational. Information must be represented somehow, and there must be rules related to that. ie "253" represents 2*100+5*10+3*1. These rules are the "syntax". some representational schemes include rules to build structures out of simple ones (combinatorial, aka productive)
- A Semantic interpretation, of a complex symbol is built up from the meanings of each syntactic symbol. A semantic interpretation of a multiplication can illustrate about the goals, but not about the implementation, the 'how'.
- Information process can be described formally. Thus, can be described and implemented without any awareness of their meaning (ie: a calculator can multiply without "knowing" what multiplication is). However, the more specialized implementations (ie complex sw) have very well defined representational relationship to their particular domains.
- In Human Language, there are multiple languages, but there is experimental proof that some "linguistic universals" do exist. See Chomsky.
- Information process can be analyzed at several levels (*like OSI layers.) ie multiplication can be analyzed:
Abstract -> what; Representative -> what does it represent; formal -> mechanics.
- The implementational mapping ties together the different layers.
Cognitive Psychology / Architecture of the mind
- Intelligence is not homogeneous, but it consists of subsystems.
- Vision and language are probably the most specialized information processing subsystems.
- The architecture of a system may or may not give it the potential to acquire information-processing capabilities that are not specifically built in.
- In humans, there are architectural differences form individual to individual. ie: a deaf person would have different architecture than a hearing one.
- GLOBAL VIEW
Think of it as a black box, with 3 stages.
Input: Sensory input is the input to the first stage.
2. Central systems: Thinking, attention, memory, learning and Language.
Those inturn provide input to
3. Motor systems.
Output: Motion, motor output.
- Sensory systems are "informationally encapsulated", they provide input to the central processes, but central processes are not engaged in the ellaboration of such input. (ie, no feedback)
- Two theories that may be complementary, may be exclusive:
a. Central Systems Theories. Classical view, connections, parallel distributed processing, artificial neural networks.
b. Physical Symbol systems. this hypothesis says that cognition can be analyzed as a formal symbol manipulation process.
- Yet, a physical symbol system does not necessarily have survival mechanisms, so is not sufficient to define intelligence at the human level.
- Distal access symbols, (ie: reliable calling subroutines, or making associations between symbol).
CHAPTER 7
7.2 Organization of Central Nervous System
Levels of description
1m > Central Nervous System
10 cm > Systems
1 cm > Maps
1 mm > Networks
100 um > Neurons
1 A > Molecules
(Sejnowki and Churchland)
- Amnesic patients don’t see affected procedural memory, only declarative memory, this would mean that there are two clearly different neurobiological implementations of these two types of memory.