Pragmatic action Kirsh and Maglio 1994

1
Pragmatic action (Kirsh and Maglio 1994)

• Task-performance is often described in terms of
  step-by-step progression through a range of
  states of the world to reach a goal state. The
  parameters of variation are defined, and the
  cognitive system
• --generates a representation of the way the world
  is
• --calculates the next move such that, when its
  taken, the physical world will be closer to the
  goal state
• --implements that move by generating an action
• The quality of an action is determined by the
  extent to which its effects bring the physical
  system closer to the goal state (winning a game,
  for example). Each such action counts as a
  pragmatic action.

2
Simple example exit the room

• --Identify the doors coordinates and ones own
  coordinates
• --Minimize the distance between the two, one
  physical step at a time

3
More complicated example Tower of Hanoi
4
Rules of ToH

• --Move all three disks from left to right, ending
  up with the same graduated configuration
• --Move only one disk at a time
• --A larger disk may never be on top of a smaller
  one
• Here, moving closer to the goal is defined in
  terms of a decision tree, moving closer to the
  action is to move the system further along a
  successful branch on the tree

5
Playing Tetris

• --Zoids fall from above
• --Buttons to rotate/drop
• --Zoids are rigid
• --If they stack up to top,
• game over
• --During the game,
• filled rows disappear

6
Playing Tetris the orthodox cognitivist model of
action generation
7
Problems with the orthodox approach

• Among other things, it doesnt explain
• --the number of rotations
• --early rotation
• --translation (movement to the wall and back)

8
Kirsh and Maglio, p. 20

• If ones theory of the agent assumes that
  thinking precedes action, and that, at best,
  action can lead one to re-evaluate ones
  conclusions, then action can never be undertaken
  in order to alter the way cognition proceeds.
• This, they claim, is misguided. Instead agents
  perform, epistemic actions, ways of altering the
  information available to the agent and thereby
  altering the way cognition proceeds.

9
CCs Three Ways to Play Tetris

• 1. Internal imagery only
• 2. Choice between use of internal imagery and
  external rotation
• 3. Choice between the use of internal imagery and
  use of internal rotation implant.

10
The moral

• Theres no principled difference between the
  external rotation process in case 2 and the
  implants processing in case 3. But there is no
  principled difference between the internal
  processes in cases 3 and 1. So, given that case 1
  is clearly cognition, case 2 must be as well.

11
Clark and Chalmerss Otto

• -Bad memory, so stores info in notebook
• -Entries in notebook guide action in the same way
  that memories (or nonoccurrent beliefs) stored in
  cortex do.
• -If Inga (the normal) has nonoccurrent beliefs
  stored in cortex, then Otto has nonoccurrent
  beliefs stored in his notebook (so his mind is
  extended beyond the boundary of his skin and
  skull).

12
At least three things going on

• --Attempt to break down internalist prejudice
  (what Clark sometimes calls skin-bag prejudice).
• --A functionalist (or related computation-based)
  argument for extended minds.
• --Argument about the best way to do cognitive
  science

13
Parity Principle

• If, as we confront some task, a part of the world
  functions as a process which, were it done in the
  head, we would have no hesitation in recognizing
  as part of the cognitive process, then that part
  of the world is (so we claim) part of the
  cognitive process. (Clark and Chalmers, p. 8)

14
Doubts about the Parity Principle

• Doesnt explain what makes something cognitive.
• Why should cognition be whatever strikes us as
  cognitive (if its in the head)?
• Worst of all, it gives the wrong answer in some
  cases. (Neuron on the table.)

15
Functionalism about mental states

• To be in a given mental state is to be in some
  (physical) state or other that plays the
  appropriate causal role, where causal roles are
  to be defined in terms of relations to inputs,
  outputs, and other mental states.
• But isnt this circular, given the reference to
  other mental states?

16
Put formally

• (x)x is in mental state M if and only if
• F1 FnT(F1Fn, I1Im, O1Ol) Fix
• where the Is and Os are antecedently understood
  predicates, typically taken to express input and
  output properties, and where T(F1Fn, I1Im,
  O1Ol) represents the best theory of the causal
  relations that obtain between mental properties
  and input- and output-properties (sensory inputs
  and behavioral outputs on most functionalist
  accounts of mental properties).

17
Realization

• (x)Fix if and only if
• P1PnT(P1Pn, I1In, O1On) Pix
• Where Ps denote physical properties. In other
  words, given a theory of how the mental states
  are related to each other (and to inputs and
  outputs), whether something is in one of those
  mental states is determined by the presence of a
  corresponding pattern of interrelated physical
  states.

18
Clark and Chalmerss Functionalist argument

• Premise 1. Functionalism about mental states is
  true.
• Premise 2. Some externally realized states play
  the distinctive causal role of what we already
  recognize to be mental states.
• Therefore, some externally realized states are
  mental states (and thus the mind is extended, so
  long as we assume the mind is where the realizers
  of its states are).
• In support of Premise 2, consider Otto and his
  notebook

19
Computational-processing version

• Premise 1. The essence of a given type of
  cognitive operation (such as manipulating an
  image) is its distinctive causal profile.
• Premise 2. Some (at least partly) external
  processes have the same cognitive profiles as
  those we already recognize to be cognitive
  processes.
• Therefore, some cognitive processes are at least
  partly external (so cognitive processing extends
  beyond the boundary of the skin).
• In support of Premise 2, consider Tetris (Kirsh
  and Maglio and CCs hypothetical case)

20
The Causal-Explanatory Kinds Argument

• --Grouping the internal and (at least partly)
  external states (and processes) together yields
  more scientifically useful kinds.
• --Thus, according to our best cognitive science,
  mental and cognitive kinds are realized at least
  partly by the external environment.
• --Thus, the mind (or cognitive system or
  cognitive processes) are extended.