Introduction to Cognitive Science

Course Summary

Last Update: 30 November 2008

Note: or material is highlighted

7±2 Key Dates in Cognitive Science
What is cognitive science?
- ≈def the interdisciplinary study of cognition, i.e., of:
  - reasoning
  - remembering
  - language understanding & generation
  - perception
  - learning
  - consciousness
  - emotions
  - etc.
- computational cognitive science = cognition is computable
  - a working hypothesis:
    - how much of cognition is computable?
  - "computable" =
    - classical symbolic computation
    - connectionist computation
- methodology:
  1. "mind as machine"
    - express theories of cognition as computer programs
    - can test theory by running the program
      - philosophical problem:
        
        Do computers running such programs actually exhibit cognition?
        If not, why not?
    - contrast with behaviorism:
      - algorithms show how to get from stimulus/input to response/output
  2. interdisciplinary:
    - share methodologies across disciplines
What is the mind?
- philosophy ≈def search for truth in any field, by rational (logical, scientific) means.
- Mind-body problem: What is the relation between mind and brain?
  - dualisms (e.g., Descartes's interactionism)
  - monisms (e.g., Berkeley's idealism; materialism/physicalism & the identity theory)
  - functionalism:
    - psychological type of a brain state/process is a function of its causal role in the individual's cognitive life
    - Problems:
      - qualia (including inverted & absent)
      - implementation (cognitive architecture) may be important
- Problem of other minds: How do I know that you have a mind?
Theory of Mind, Folk Psychology, & Modularity
- Theory theory:
  - BDI folk psychology is a theory we use to explain & predict others' and our own behavior
- Simulation theory:
  - We explain & predict others' behavior by simulating the others' mind
- Fodor's theory of mental modules
- Pylyshyn's theory of cognitive impenetrability
Reasoning
- logic = normative study of reasoning correctly (i.e., truth-preserving)
- cognitive science: how people actually reason
  - not always correctly!
    - Wason card-selection task
    - Tversky & Kahneman on probabilistic reasoning
  - Johnson-Laird's theory of Mental Models
  - Rips's theory of Mental Rules
  - Simon: bounded rationality
  - AI: computational theories of non-monotonic, default, and defeasible reasoning
Rules and Connections
- AI: "computational cognition"
  - Minsky: make computers do cognitive tasks
  - Boden: understand (human) cognition using computers
  - AI as:
    - advanced CSE
    - computational psychology:
      - programs as theories/models of (human) cognition
    - computational philosophy:
      - how is cognition possible?
      - how much of cognition is computable?
- What is computation?
  - algorithms
  - Turing machines
  - Church-Turing thesis: Any algorithm can be expressed as a TM
  - GOFAI (classical symbolic AI): logic-based
    - Newell & Simon's PSSH:
      - a physical system can exhibit cognition
        iff
        it is a physical symbol system
        (= physical implementation of a TM)
      - production systems:
        
        Logic Theorist
        GPS
    - Fodor's CTM & RTM:
      - cognitive states/processes are computations over cognitive representations
      - cognitive representations are a language of thought (LOT),
        with a syntax (& maybe a semantics)
    - Dennett's Intentional Stance:
      - best to treat complex systems as if they were intentional (i.e., cognitive)
    - SNePS
    - Chomsky:
      - deep vs. surface structure;
        transformational grammar
      - universal grammar (innate LAD)
      - competence vs. performance
  - Connectionism (ANNs): statistics-based
    - Rosenblatt's Perceptron
    - Minsky & Papert's objections
    - Rumelhart & McClelland's PDP
    - representation as activation patterns of artificial neurons
    - Fodor & Pylyshyn's objections:
      - not productive
      - not systematic
    - hybrid systems
    - unconscious inference/tacit knowledge
  - Dynamical Systems theory: differential-equation-based
Concepts and Categories
- classical view: categories as sets defined by necessary & sufficient conditions
- Wittgenstein: family resemblances
- Rosch:
  - basic-level categories
  - prototypes vs. exemplars
- Lakoff & Johnson:
  - conceptual systems are metaphorical
  - based on human body.
Memory
- procedural
- declarative
  - episodic
  - semantic
- long-term, short-term, working memory
- AI memory schemata: Minsky's frames, Schank's scripts
Vision & Mental Imagery
- Pylyshyn: mental images are propositional
- Kosslyn: mental images are pictorial
- Gibson: direct perception of "affordances"
- Marr: perception requires processing:
  - primal image, 2.5D image, 3D image
  - 3 levels of psychological explanation:
    1. "computational" (functional): what (I/O)
    2. "algorithmic" (computational): how (algorithms)
    3. "implementation": multiple physical realizations
- philosophy:
  - direct vs. indirect realism
  - arguments from illusion, time lag
- Guest Lecture: Udin
Cognitive neuroscience
- anatomy (structure) & physiology (processing) of brain
- McCulloch & Pitts's logical ANNs
- is the brain a computer?
Emotion
- Sloman's information-processing theory of grief
  - suggests possible 4th level between Marr's (a) & (b):
    information-processing level
- Simon: emotion's role in guiding reason through the limitations of bounded rationality
Consciousness
- identity theory: conscious states are brain states
- Nagel: only a bat can know "what it's like to be a bat"
- McGinn: understanding consciousness is beyond our cognitive abilities
- Rosenthal: consciousness is meta-thought
- Dennett: multiple drafts theory
- Chalmers: physicalism is false
- are there neural correlates of consciousness?
- mirror neurons
- Rapaport on "Notes towards a Computational Theory of Consciousness",
  TODAY (Thurs., Dec. 3), 1:00-3:00 p.m., Bell 224
Metacognition in Non-Human Animals
- Guest lecture by J. David Smith
Situated/embodied/embedded/extended/external cognition
- importance of context/the world
- Putnam: Twin Earth: cognition is not (merely) in the head
- Fodor: Methodological Solipsism: yes it is!
- Hutchins/Clark: there are extended cognitive systems; so, no it isn't!
- Simon: yes it is
- B.C.Smith: no it isn't
Interdisciplinary cognitive science projects (at UB):
- Deixis & Narrative
- CVA
- others:
  - LIN + Neurosci: PET studes of linguistic processing (Jaeger, Lockwood)
  - LIN + PSY: lexical semantics (Koenig, Mauner)
  - LIN + AI: computational linguistics (Koenig, Shapiro, Rapaport, Pierce)
  - PHI + GEO + LIN: spatial ontology (Smith, Mark)
Turing Test & Chinese-Room Argument
- Turing Test:
  - Computers will be said to be able to think
    to the extent that we cannot distinguish their linguistic/cognitive ability from a human's
- Chinese-Room Argument (Searle):
  - It's possible to pass a TT without really thinking
    - Pat Hayes has allegedly defined cognitive science as "the ongoing research program of showing Searle's Chinese Room Argument to be false"! (Harnad, cited in Boden 2006:1384)
- ultimately, might be a moral choice!