Functionalism and Functional Complexity

Minds are made of the same kind of “stuff” as bodies; no non-physical substances are required.
Two principal physicalist strategies have been surveyed so far:
- Identity Theory
- Functionalism

Claim: Every psychological (mental) state is numerically identical to a specific physical state.
- E.g., classic textbook claim: $\text{Pain} = \text{C–fiber firing}$ (in mammals).
Ontological payoff: removes "extra" mental entities—just neural (or other physical) entities.
Epistemic / explanatory ambition: if we can map every mental state to a physical state, psychology reduces to neuroscience or physics.
Limitation motivating functionalism: seems too species-specific or realizer-specific—what about octopuses, silicon aliens, or future AI?

Core slogan: "What makes a state mental is what it does, not what it is made of."
Mental kinds are individuated by their causal/functional roles:
- Inputs: perceptual stimuli, previous mental states.
- Internal changes: production of new mental states.
- Outputs: bodily behaviour, verbal reports, further neural activity.
Multiple realizability:
- Octopus pain = octopus-neural-state-X.
- Human pain = human-neural-state-Y.
- Alien pain = alien-neural-state-Z (possibly non-carbon).
- All count as "pain" because they perform the same functional role: cause aversion, trigger "Ow!", generate a phenomenal feel of hurt.
Functional analysis is substrate-neutral; it abstracts away from the underlying biochemical or physical medium.

Digital computer: defined by information-processing profile, not by silicon vs. wood.
- A wooden (hypothetical) Turing machine counts as a computer if it maps inputs to outputs according to the right algorithm.
Brains as information-processing systems:
- Psychological states = "software" or functional states.
- Neural tissue, silicon chips, or exotic alien goo can all be the "hardware" as long as they run the right functional program.
The analogy reinforces functionalism's hardware independence and offers a methodology: study cognition by specifying the algorithmic level, then let neuroscientists / engineers worry about implementation.

To have a mind = to possess a sufficiently large and intricately connected set of functionally specified psychological states.
Examples across the complexity spectrum:
- Sea sponge:
- Minimal functional repertoire: "Take in water ➜ filter nutrients ➜ expel water."
- Functional profile too simple ⇒ no mind.
- Human:
- Vast network of interdependent states (beliefs, desires, perceptions, emotions) modulated by context.
- Example:
  - If "hungry" $\land$ smell("chocolate") ➜ desire("eat chocolate").
  - If "full" $\land$ smell("chocolate") ➜ desire("remove smell").
- Same perceptual input plays different functional roles depending on other concurrent mental states.
Open research question flagged by lecturer: Where exactly is the complexity threshold?
- When does an organism (or machine) cross from mere reactivity to genuine mentality?

Ethical / moral relevance: If an entity’s moral status hinges on mentality (e.g., capacity to feel pain), functionalism suggests we look for behavioural/causal profiles, not specific biochemistry.
Artificial Intelligence: Supports possibility of conscious or minded AI if they achieve the right functional architecture.
Continuity with previous lectures: builds on rejection of Cartesian dualism; preserves physicalist framework while escaping identity theory’s narrowness.
Practical methodology: Cognitive science can model mental functions at multiple levels (computational, algorithmic, implementational) without committing to a single neural blueprint.

Key upcoming question: What is the minimum functional complexity required for mentality?
- Can it be formally measured (e.g., via information theory, state-space size, or network causal density)?
- How does this threshold apply to borderline cases: insects, simple robots, large language models?