Comp. Models of the Mind II c Flashcards
A collection of 9 advantages of computational cognitive modelling:
- increase in clarity
- sufficiency proof
- thorough evaluation possible
- unifying principles may be discovered
- complex dynamic interaction
- generative power of local rules (-> emergence of complex behavior)
- bootstrapping of development of theories
- precise quantitative predictions
- generalizability
Advantages - Increase in clarity through:
- specification of representation
- complitness and consistency of model
- all parameters and relationships must be fully specified
The following requirements in cognitive modelling can lead to …?
- specification of representation
- complitness and consistency of model
- all parameters and relationships must be fully specified
- > revealing of hidden assumptions
- > explication of hidden assumptions
- > revealing of unexpected/counter-intuitive consequences
Building a computational model is at the same time a sufficiency proof of …
… internal coherence and completness of underlying theory
A collection of 5 disadvantages of computational cognitive modeling:
- Irrelevant specification problem
- selectivity (not everything fits into models)
- skills required to access and understand programs
- Bonini’s paradox
- Assumption of invariability
- may miss creative, flexible, non-standard activity
What is the irrelavant specification problem about?
Strong impact of design decisions regarding theoretically irrelavant details, such as data structure
What does Bonini’s paradox say?
As a model of a complex system becomes more complete, it becomes less understandable or as difficult to understand as the represented real-world phenomenon
A very good practice in comp cognitive modeling is …
… Validation!!
What is validation about?
Validation is about how adequately the model reflects the aspects of the real world it has been designed to model.
3 ways to validate a model:
- explicate how much a model restrains the data
- report data variability: verify real-world data agree also with outcomes ruled out by the model
- show there are plausible results the model cannot fit
Marr’s levels can help to identify/distinguish the _________ of a model.
Marr’s levels can help to identify/distinguish the DOMAIN of a model.
In cognitivism: Cognition involves …
… computations over internal representations.
In cognitivism: Informatin about the world is abstracted by …
… perception and represented via symbolic data structures.
In cognitivism: How would a typical sequence of precessing look like?
perceive - reason - plan - act
Issues in cognitivist models:
- system blindness: constrained by (idealised) human perspective and capabilities
- brittleness of system: breakdowns due to semantic gaps
- issues with sensing: inherently uncertain, incomplete
What does PPS stand for?
the “Physical Symbol System” approach to AI
Who can be associated with the Physical Symbol System?
Newell and Simon (1976)
One statement of the PPS and two implications:
- PSSs have necessary and sufficient means for general intelligent action
- > any system exhibiting general intelligence is a PSS
- > any PSS of sufficient size can exhibit general intelligence
Two statements of the Heuristic Search Hypothesis:
- Solutions to problems are represented as symbol structures
- intelligence is exercised in problem solving by search
On the basis of the two statements of the Heuristic Search Hypothesis, one could see the task of intelligence, as the task to …
… avert the ever-present threat of the exponential explosion of the search.
According to Newell and Simon processes can to what with patterns in PSSs?
- producing
- destroying
- modifying
_____ of physical symbol systems are key, _________________ is unimportant.
DETAILS of physical symbol systems are key, PHYSICAL INSTENTIATION is unimportant.
Three main properties of Connectionist Emergent Models:
- Parallel processing of non-symbolic, distributed activation patterns
- reliance on statistical properties rather than logical rules
- dynamical systems which compute functions that best capture statistical regularities in training data