computational modelling

Question 1

why is modelling important (5 reasons)

Answer 1

data never speaks for self,
multiple models to select from
verbal theorising not sub for quantitative analysis
model selection based on quantitative and intellectual/qualitative judgement
psych moving rapidly towards this level of specification

Question 2

what is a model

Answer 2

abstract framework capturing data’s structure (e.g mean)

simpler version of what’s being explained

Question 3

Lewandowsky & Farrells 2011 model classification

3 elements

Answer 3

data description = describes relationship b/w variables
process characterisation = explanatory power lies in hypothetical constructs of mind (neutral regarding implementations of processes
process explanation = implement HOW processes occur, not neutral

Question 4

3 levels of model

Answer 4

computational = what system does & why
algorithmic/representational level = how system does what it does (what rep used, processes to build &manipulate reps)
physical = how systems physically realised

Question 5

benefits of modelling

Answer 5

model helps understand/explain data
instantiation of quantitative model = all assumptions of theory ID’d and tested
force theorists to be explicit

Question 6

what is emergence of understanding

Answer 6

models (as comp programs) cannot generate novel ideas - E-Z reader &skipping costs

Question 7

wrong models and verisimilitude

Answer 7

all models are inherently wrong, but can still be useful

verisimilitude = partial truth value

Question 8

problems with models

Answer 8

scope and testability - hard to falsify

Question 9

application of neural networks

Answer 9

txt to speech - NetTalk (Sejnowski and Rosenberg 1988)
medical images 
credit checks
stock market 
national defense 
handwritten character recognition

Question 10

what is connectionism (neural networks)

Answer 10

sub symbolic
mimics biology
parallel not serial
emergent properties ( model = grown not built, back propagation learning