Is Cognitive Psychology Doomed to Failure

1
Is Cognitive Psychology Doomed to Failure ?
Vaughan Bell vaughan_at_backspace.org
2
Outline

• Aims and assumptions of cognitive psychology
• Scientific methodology
• Information processing view of the mind
• Criticisms from
• Phenomenology
• Algorithmic information theory
• Future directions ?

3
Scientific / Positivist Approach

• Cognitive psychology has inherited experimental
  methodology from behaviourism.
• It is also positivist and reductionist, assuming
  that mental phenomena can be measured and
  fractionated.
• And although entertains ideas of mental states,
  has largely rejected the introspective methods
  and subjectivity of Freud.

4
Information Processing

• To varying degrees, cognitive psychology seeks to
  explain the mind in terms of information
  processing.
• Pylyshyn (1979, p435) argues that computation is
  not a metaphor but part of a literal description
  of cognitive activity.
• Parkin (2000) argues that cognitive psychologists
  use information processing purely as an analogy
  for mental function.

5
Has this become a dogma ?

• Still and Costall (1987) would agree and have
  argued that cognitive psychology has become
  restrictive as well as complacent.
• Even Don Norman (1981) has criticized cognitive
  psychology for sterility and for descriptions
  which do not fit actual behaviour.
• I am going to look at criticisms from two
  sources Phenomenology and Algorithmic
  Information Theory

6
Phenomenology

• Originated with Husserl and continued by the
  likes of Heidegger, Satre and Merleau-Ponty.

Martin Heidegger (1889 1976)
Edmund Hussrl (1859 - 1938)
7
Core Beliefs

• Phenomenology seeks to describe the structures of
  experience as they present themselves to
  consciousness.
• Without recourse to theory, deduction, or
  assumptions from other disciplines.
• Two of the main objections by phenomenologists to
  cognitive psychology are quantification and
  abstraction of human experience.

8
Objections to Quantification

• Phenomenologists believe that human experience
  cannot be quantified without stripping it of all
  meaning.
• They argue experience cannot be reduced to its
  component parts.
• And by doing so, we are throwing the baby out
  with the bath water...
• ...choosing a quest for scientific respectability
  over the truth of human experience.

9
Objections to Abstraction

• Similarly they stress the importance of
  intentionality and situated action
  (being-in-the-world).
• This implies that our thoughts cannot be
  meaningfully separated from the environment with
  which they interact.
• And by studying psychology in artificial
  experimental situations we will get nothing
  except an artificial understanding of the mind.

10
Information Processing

• Algorithmic and information processing models are
  ubiquitous in modern psychology.
• Philosophers (e.g. Searle) argue that psychology
  cannot be fully described by computation.
• Many psychologists would agree, however I can
  think of very few cognitive models that could not
  in principle by implemented computationally.

11
Innate Problems with Computation

• Two mathematicians have been particularly
  important in showing the limits of computation.

Kurt Gödel (1906 1978)
Alan Turing (1912 1954)
12
Gödels Incompleteness Theorem

• Much to the surprise of the mathematical world,
  Gödel discovered in 1930 that
• i. in any consistent system of mathematics, one
  can construct a statement about natural numbers
  that can be neither proven nor disproven within
  that system
• ii. any such system cannot prove its own
  consistency

13
Halting Problem and Completeness

• Turing discovered there is no way to calculate a
  priori, whether any given algorithm will complete
  or not.
• In other words, there are some things with are
  just not computable.
• He also described the minimum operations a
  computer needs to compute.
• A system that is Turing-complete can execute any
  possible computer programme.

14
Minimal Example

• Brainfuck is a programming language designed by
  Urban Müller (1993) to be the smallest
  Turing-complete language.
• It has 8 single character instructions. Every
  possible computer programme can be rewritten in
  it.
• gtgtgtgtltltltlt-
  gt.gt....gt.ltlt.
  gt..------.--------.gt.gt.
• The Hello world programme.

15
Relevance to Cognitive Science

• If we accept the information processing model
• Gödel There may be some mental states we will
  never uncover, and we cannot verify cognitive
  theories without referring to non-cognitive ones.
• Turing A priori, cognitive systems will be
  unable to solve certain problems.
• Müller All other mental functions can be coded
  into strings made from only eight characters.

16
Future Directions I

• We seem to need an idea of where information
  processing models are inappropriate.
• Past success is not always a good measure of
  appropriateness.
• Because we may be using the right model in the
  wrong way, or using it to ask the wrong
  questions.
• i.e. Difficult problems have difficult to
  conceive answers.

17
Future Directions II

• Perhaps we can look at previous examples of
  hybrid or non-cognitive models (e.g. Gibson,
  Bartlett)
• Or take inspiration from neurobiology,
  philosophy, computer science, art etc and hope
  that we can muddle through.
• Although we may have to accept parallel models
  and the waxing and waning of academic fashions.