The Human Information Processing System
- Levels of Analysis (Marr, 1982)
- Representation
- Process
- The Modularity Hypothesis (Fodor, 1983)
- The Human Memory System
- Important Questions About Theories
Levels of Analysis (Marr, 1982)
- The Computational Theory
- Representation and Algorithm
- Implementation
- Where do we fit in?
The Computational Theory
- What information is available?
- What is the goal of the computation?
- What strategy is used to achieve the goal with the available information?
- Examples from Psychology:
- What information is available in the environment? (Gibson, 1966)
- What is the most adaptive strategy in this environment? (Anderson, 1990)
Representation and Algorithm
- What is the representation for the input?
- What is the representation for the output?
- What is the algorithm for transforming the input into the output?
- Examples from Psychology:
- What are the psychological structures and processes that underlie
intelligent behavior? (Newell, Shaw, & Simon, 1962)
Implementation
- How are the representation and algorithm realized physically?
- Examples from Psychology:
- How do neurons work? (Hebb, 1949)
- What is the functional organization of the brain? (Posner & Raichle,
1994)
Where do we fit in?
- Cognitive psychology and cognitive science have traditionally focused on
the representation and algorithm.
- Cognitive neuroscience places an equal emphasis on the implementation.
- Like cognitive psychology, we will emphasize the representation and
algorithm, but will always look for constraints from the other two levels of analysis.
Representation
- Data Structures in Computer Science
- Mental Representation
- Levels of Representation
- Local versus Distributed Representations
Data Structures in Computer Science
- Data Elements
- Relationships Among Elements
- Examples
Mental Representation
- Concepts as Data Elements?
- Associations and Semantic Relationships
- Examples
- Semantic Networks
- Proposituions
Semantic Networks
Propositions
Levels of Representation
- Cognitive scientists often assume that the world is organized into
multiple, hierarchically embedded levels of representation.
- Examples from Language:
- Discourse (Conceptual)
- Sentence
- Word
- Morpheme
- Phoneme
- Acoustic Feature (Perceptual/Motor)
Local versus Distributed Representations
Local Representation of "Bird"
Distributed Representation of "Bird"
Process
- Need for a Structure-Process Pair
- Programming Examples
- Examples of Cognitive Processes
- What do cognitive processes do?
- Important Distinctions
Need for a Structure-Process Pair
- Data Structures + Algorithms = Program
- Knowledge Structures + Cognitive Processes = Cognitive Theory
- Tradeoff Between Structure and Process
- In some theories the power is in the representations
- In others, its in the processes.
- Power and Sufficiency
Programming Examples
- Push-Pop
- Tree Traversal
- Depth First
- Breadth First
- Best First
Examples of Cognitive Processes
- Associative Learning
- Spreading Activation
What do cognitive processes do?
- Create New Representations
- Retrieve Existing Representations
- Manipulate Representations (Reasoning)
Important Distinctions
- Automatic versus Controlled
- Sequential versus Parallel
- Bottom-Up versus Top-Down versus Interactive
- Symbolic versus Sub-Symbolic
Automatic versus Controlled
- Is it intentionally initiated?
- Can it be stopped?
- Are you aware of the intermediate steps?
- Does it use STM resources?
- An Example of an Automatic Process (the Stroop Test)
Sequential versus Parallel
- In a sequential process, step n must be completed before step n+1 begins.
- In a parallel process, multiple steps are executed at the same time.
- Cascaded processes have elements of both: step n must begin (but
need not finish) before step n+1 begins.
- This distinction can be applied both between and within
levels of representation.
Bottom-Up versus Top-Down versus Interactive
- Bottom-up processes flow from perceptual/motor levels of representation
to more conceptual levels.
- Top-down process flow from conceptual levels of representation to more
perceptual/motor levels.
- Interactive processes flow in both directions simultaneously.
Symbolic versus Sub-Symbolic
- The Symbol System Hypothesis
- The Sub-Symbolic Hypothesis
- Example: Assigning Past-Tense Endings to Verbs
The Symbol System Hypothesis
- Thinking is information processing.
- Information processing is computation on symbols.
- The semantics of symbols connect thinking to the external world.
- The "implementation" is irrelevant! (optional)
The Sub-Symbolic Hypothesis
- Thinking is information processing.
- Information processing is sub-symbolic computation.
- The semantics of symbols connect thinking to the external world
- The "implementation" is relevant! (optional)
Example: Assigning Past-Tense Endings to Verbs
- Using Rules:
- verb verb + "ed"
- exceptions: run ran, etc.
- Self-Organizing Artificial Neural Network
The Modularity Hypothesis (Fodor, 1983)
- The mind consists of a general problem solver plus a set of input modules.
- Standard View: The mind is a general problem solver.
- Evolutionary Psychology: The mind is like a Swiss army knife.
- Language is one of the modules.
- Properties of Modular Systems:
- Domain Specific
- Hard-Wired and Innate
- Autonomous
- Non-Assembled
- Informationally Encapsulated
- This hypothesis is a major source of theoretical conflict in the psychology of language.
The Human Memory System
- The Organization of Human Memory
- Short- versus Long-Term Declarative Memory
- The Role of Memory in Language Processing
The Organization of Human Memory
- Procedural Memory
- Declarative Memory
- Episodic Memory
- "Where were you on your 18th birthday?"
- "Who was the first person you saw after you left home this
morning?"
- Semantic Memory
- "What is your mother’s middle name?"
- "What kind of bird is black and white, lives in the Antarctic, and
swims rather than flies?"
Short- Versus Long-Term Declarative Memory
- Long-Term Memory
- Input: Deep Processing
- Coding: Semantic/All
- Capacity: Essentially Unlimited
- Duration: Essentially Unlimited
- Speed: Potentially Slow
- Loss: Interference
- Short-Term Memory
- Input: Attention/Retrieval
- Coding: Acoustic/All
- Capacity: 7+2 Chunks (Storage/Processing Tradeoff)
- Duration: 10 - 15 seconds w/o rehearsal
- Speed: Very Fast!!!
- Loss: Interference/Maybe Decay
The Role of Memory in Language Processing
Language units (words, letters, phonemes, acoustic and
visual features) are stored in semantic memory.
Sentences that we have read or heard are stored in
episodic memory.
Knowledge of how to use language is stored in procedural
memory.
LTM (speed) and STM (capacity) impose important
processing limits.
Important Questions About Theories
- What form of mental representation is assumed?
- Is the representation local or distributed?
- How many levels of representation are assumed?
- What processes are assumed?
- Are the processes bottom-up, top-down, or interactive?
- Are the processes sequential or parallel within levels?
- Are the processes sequential or parallel between levels?
- Are the processes controlled or automatic?
- Are the processes symbolic or sub-symbolic?
- Are the structures and processes modular?
The End!
|