SURFING THE EDGE OF CHAOS
MIND: REVIEW & EVALUATION – CHAPTER 8 (THAGARD)
THE ACHIEVEMENTS OF COGNITIVE SCIENCE
Scientific understanding of problem solving, learning, language much better now than 50
years ago
BUT still no unified theory of cognitive science that explains full range of psychological
phenomena
To understand mind we need to take all theories into account
Progress will require more than isolated efforts of researchers in a particular
discipline
Different versions of CRUM have illuminated different aspects of applied thinking
COMPARATIVE EVALUATION
REPRESENTATIONAL POWER
Formal logic – considerable representational power, operators & quantifiers
Computer models with rules, concepts, analogies, images, connections – difficulty representing
proposition such as: “No X are Y for some X”
Connectionism – more flexible in capturing broader range of sensory experience than
verbal representations
Struggled with how simple neuronlike units can represent complex relations
Logic Rules Concepts Analogies Images
Abandon Collects Collects Own special packaging
expressiveness of package of package of – intimately
logic for more info about a info about a connected with
computational kind of thing situation sensory functions
power (vision)
Concise, independent Bundle info together in organised structures
representations
NO CURRENT COMPUTATIONAL MODEL HAS REPRESENTATIONAL POWER TO CAPTURE ALL OF HUMAN THOUGHT
Unified theory of mental representation needs structures that among them have
Sensory richness of images & connections
The organising capabilities of concepts, analogues & pictures
The verbal expressiveness of rules & propositions in formal logic
COMPUTATIONAL POWER
Need to be concerned with speed & flexibility & abstract computational potential
, Rule-based systems (heuristics) superior performance in many domains
Problem Many problem solutions better described in terms of schema application,
solving analogical mapping, parallel constraint satisfaction (than as heuristics in
rule-based system)
Learning Neither rule-based chunking nor connectionist weight adjustment
captures full range of human learning capabilities
Language Cognitive science lacks comprehensive theory of language learning &
use
PSYCHOLOGICAL PLAUSIBILITY
Each of 6 approaches has inspired psychological experiments & computational models
Different experimental results fit best with different representational theories
NEUROLOGICAL PLAUSIBILITY
New scanning techniques cognitive neuroscience fastest developing part of
cognitive science
Connectionist models gain some neurological plausibility from analogy between
artificial neural networks & brain
PRACTICAL APPLICABILITY
Can proceed in more piecemeal fashion, selectively applying insights from different
approaches where they appear relevant
Education Rules, concepts, analogies contributed to understanding how people
solve problems
Connectionist ideas are starting to have impact on educational theory
& practice
Design Mostly improved through rules, concepts, analogies, images in creative
design
Intelligent Most expert-systems: rule-based systems
systems Some case-based (analogical) & connectionist systems are proving
increasingly useful
CHALLENGES OF COGNITIVE SCIENCE
7 important challenges for CRUM
1. The brain challenge – CRUM ignores crucial facts about how thinking is performed by the
brain
2. The emotion challenge – CRUM neglects important role of emotions in human thinking
MIND: REVIEW & EVALUATION – CHAPTER 8 (THAGARD)
THE ACHIEVEMENTS OF COGNITIVE SCIENCE
Scientific understanding of problem solving, learning, language much better now than 50
years ago
BUT still no unified theory of cognitive science that explains full range of psychological
phenomena
To understand mind we need to take all theories into account
Progress will require more than isolated efforts of researchers in a particular
discipline
Different versions of CRUM have illuminated different aspects of applied thinking
COMPARATIVE EVALUATION
REPRESENTATIONAL POWER
Formal logic – considerable representational power, operators & quantifiers
Computer models with rules, concepts, analogies, images, connections – difficulty representing
proposition such as: “No X are Y for some X”
Connectionism – more flexible in capturing broader range of sensory experience than
verbal representations
Struggled with how simple neuronlike units can represent complex relations
Logic Rules Concepts Analogies Images
Abandon Collects Collects Own special packaging
expressiveness of package of package of – intimately
logic for more info about a info about a connected with
computational kind of thing situation sensory functions
power (vision)
Concise, independent Bundle info together in organised structures
representations
NO CURRENT COMPUTATIONAL MODEL HAS REPRESENTATIONAL POWER TO CAPTURE ALL OF HUMAN THOUGHT
Unified theory of mental representation needs structures that among them have
Sensory richness of images & connections
The organising capabilities of concepts, analogues & pictures
The verbal expressiveness of rules & propositions in formal logic
COMPUTATIONAL POWER
Need to be concerned with speed & flexibility & abstract computational potential
, Rule-based systems (heuristics) superior performance in many domains
Problem Many problem solutions better described in terms of schema application,
solving analogical mapping, parallel constraint satisfaction (than as heuristics in
rule-based system)
Learning Neither rule-based chunking nor connectionist weight adjustment
captures full range of human learning capabilities
Language Cognitive science lacks comprehensive theory of language learning &
use
PSYCHOLOGICAL PLAUSIBILITY
Each of 6 approaches has inspired psychological experiments & computational models
Different experimental results fit best with different representational theories
NEUROLOGICAL PLAUSIBILITY
New scanning techniques cognitive neuroscience fastest developing part of
cognitive science
Connectionist models gain some neurological plausibility from analogy between
artificial neural networks & brain
PRACTICAL APPLICABILITY
Can proceed in more piecemeal fashion, selectively applying insights from different
approaches where they appear relevant
Education Rules, concepts, analogies contributed to understanding how people
solve problems
Connectionist ideas are starting to have impact on educational theory
& practice
Design Mostly improved through rules, concepts, analogies, images in creative
design
Intelligent Most expert-systems: rule-based systems
systems Some case-based (analogical) & connectionist systems are proving
increasingly useful
CHALLENGES OF COGNITIVE SCIENCE
7 important challenges for CRUM
1. The brain challenge – CRUM ignores crucial facts about how thinking is performed by the
brain
2. The emotion challenge – CRUM neglects important role of emotions in human thinking
