Notes: 21 Jan (Colunga)
What's innate?
- (grammatical) categories
- parameters (a la Chomsky)
- content, the internal representation of a behavior
- the learning mechanism
- that which is inevitable
- that which is specified by the genome
- a highly probably developmental outcome
- changes within the organism, without the outside environment
- a predisposition to interact with the environment at a cellular level
- biases, constraints
In the example of the beehive - making hexagons is not innate (but it isn't
learned either). Whatever makes the bees push the cells in the way they do
is innate, the outcome (hexagonal cells) is the result of that behavior in
this world, so making hexagons is primal according to Elman et al.'s definition.
According to Elman et al., things can be innate through constraints at
three different levels:
- Representation
- Architecture
- Timing
His argument against innateness at the level of Representation is similar
to Braine's argument for the need of a bridging theory between the pre-innate
growth processes and the post-innate learning events.
Can we all just get along?
Differences between Pinker's (innate) and and Elman et al.'s (development)
approaches:
- they do not work on the same range of data
- their explanations go to a different level of detail
- top-down vs. bottom-up
- shape of behavior vs. shape of mechanism
What's at stake?
- humans are not special
although a complete explanation of language acquisition would probably be
complicated enough to make any evolution snob feel good, if the empiricists
are right the elements of that explanation would be simple and common to
other animals. No magic.
- MIT is not special
proving somebody else's axioms is bad manners.