Computationalism

1. Language of Thought (Fodor)

Mental representations have language-like structure:

• Symbols with syntax and semantics
• Compositional (complex thoughts built from simple ones)
• Productive (finite symbols → infinite thoughts)
• Systematic (if you can think “A loves B,” you can think “B loves A”)

Example: Believing “dogs chase cats” involves mental symbols for DOG, CHASE, CAT in a structured representation.

2. Physical Symbol System Hypothesis (Newell & Simon)

A physical symbol system (like a computer or brain) has necessary and sufficient means for intelligent action.

Any system that manipulates symbols according to rules can think.

3. Multiple Realizability

Same computation can run on different hardware (silicon, neurons, anything). What matters is the algorithm, not the physical substrate.

This supports functionalism - mental states are computational states, multiply realizable.

Perception → Encoding → Central Processing → Output Planning → Action
    ↓           ↓              ↓                    ↓
Symbols    Representations  Inference/Rules    Motor commands

Legitimates Cognitive Science

If mind is computational, we can study it scientifically:

• Build computational models
• Test predictions
• Reverse-engineer algorithms from behavior

AI Foundation

Computationalism justified GOFAI (Good Old-Fashioned AI):

• Expert systems
• Logic programming
• Symbolic reasoning

Level of Explanation

Cognitive science is autonomous from neuroscience. Study algorithms (computational level) independent of implementation (neural level).

Modeling Approach

• Formalize cognitive processes as algorithms
• Implement in code
• Test: Does model produce human-like behavior?

Examples:

• Production systems for problem-solving
• Semantic networks for memory
• Logic-based models of reasoning

Language Processing

Computationalism supports:

• Grammar as formal rules (Chomsky’s Universal Grammar)
• Parsing as algorithm applying grammatical rules
• Semantics as compositional computation over symbols

Bilingualism: Two lexicons, shared or separate computational systems for syntax?

Theory Structure

Build computational theories:

• What are the symbols?
• What are the operations?
• What’s the algorithm?
• What does it compute?

Symbol Grounding Problem (Searle)

How do symbols get meaning? Computer symbols are just patterns - they don’t mean anything to the computer.

Chinese Room: You can manipulate Chinese symbols by rule without understanding Chinese. Is that real understanding?

Frame Problem

How do you represent all relevant knowledge? Infinite facts potentially matter for any decision. Computationalism seems to require explicitly representing everything.

Embodiment Challenges

Body and environment matter for cognition, but classical computationalism treats them as mere input/output devices. Maybe cognition isn’t brain-bound symbol manipulation.

Connectionism

Neural networks learn patterns without explicit rules or symbols. Maybe cognition is subsymbolic - distributed patterns, not discrete symbols.

Dynamical Systems

Maybe cognition is continuous dynamics, not discrete computation. Real-time coupling with environment, not internal symbol manipulation.

Statistical/Probabilistic Cognition

Humans aren’t logic machines. We use heuristics, probabilities, approximate reasoning. Pure symbolic computation is too brittle.