Comp. Models of the Mind II a Flashcards
Computational Models of Mind - Motivations for Neuroscience:
Provide a framework for interpreting imaging data.
Computational Models of Mind - Motivations for Psychology:
Account for experimental data.
Computational Models of Mind - Motivations for Philosophy:
Provide a unified understanding of the mind.
Computational Models of Mind - Motivations for Human Computer Interaction:
Evaluate artefacts and help in their design.
Computational Models of Mind - Motivations for Applications:
e.g.:
- Cognitive models for supervised, mixed-initiative, autonomous control task
- Assistive and intelligent tutoring system
etc.
Modelling is used for systems/phenomena that are …
… too complex …
… too difficult …
… impossible …
… to deal with directly.
What’s a model?
A simpler and more abstract version of the system.
4 important properties of models:
- essential features preserved
- omission of details considered unnecessary
- results of good models can be applied to original system
- examination of model increases understanding of original system
Who is responsible for “neats” vs. “scruffies”?
R. Abelson (1981): “constraint, construal and cognitive science”
What is the substrate of cognitive models?
Prescriptions in formal mathematical/computer languages (in contrast to verbal description)
Cognitive models are derived from?
Basic principles of Cognition
Generic statistics and cognitive models?
statistical tools are used to analyze cognitive models
Neural models and cognitive models?
Cognitive models bridge between behavior and neural underpinnings
Examples for categorization of perceptual objects:
in x-ray image: cancerous, benign or no tumor
wild mushrooms: poisonous, edible, harmless + inedible
paintings: renaissance, romantic, modern, or “other period”
2 models for categorization of perceptual objects:
- Prototype Model
- Exemplar Model
2 models for categorization of perceptual objects - the prototype model:
- estimation of central tendency from all examples experienced from each category during training
- new target stimulus compared to each target prototype
2 models for categorization of perceptual objects - the exemplar model:
- memorization of all experienced training stimuli
- new target stimulus compared to each stored example
2 purposes of AI according to Herbert Simon:
- power of computers to augment human thinking
- use of AI to understand how human think
(testing programs not by what they can accomplish, but by how they do it -> cognitive science)
A synonym for “synthetic psychology” and a name:
cognitive modeling (cf. Computational Psychology, Sun 2008)
5 key elements of Computational Cognitive Modeling:
- Models implemented as computer programs
- Combination of deductive and experimental methods
- Clarity and Completeness
- Reproduction of investigated Phenomenon
- Description and prediction of performance
Reproduction of investigated phenomenon is important for …
… Exhaustive evaluation!
Behavior of models can be observed and meassured
Main point of “Neats” vs. “Scruffies”?
Difficult reconciliation of:
- remaining open to messy reality of human experience
- ordering and formalization of this messy reality
Caccioppo’s and Berntson’s example of the identification problem features …
… twins in a magic show!
4 steps of cognitive modeling:
- make a computational model of a cognitive process
- present given inputs
- let it perform internal operations to create behavior
- asses (matching model’s and real minds’ behavior: insights by success + failure!)
Two difficulties described by the “identification problem”:
- Hidden mechanisms cannot be determined exactly
- different mechanisms -> identical behavior
Who (and when) is behind the “law of uphill analysis and downhill invention”?
Braitenberg (1984)
Explain uphill analysis and downhill invention:
Induction is more laborious than
Deduction
What fallacy/tendency is depicted in the magic show twins example?
(happens also in the 2-4-6 problem)
The tendency to overestimate complexity of mechanisms.
Name the first 5 Braitenberg Vehicles:
- Alive (1 motor, 1 sensor)
- Fear and Aggression (2 motors, 2 sensors), Dislike of Source:
- Coward and
- Aggressive (exitatory)
- Liking:
- Love and Exploration (inhibitory)
- Values (multisensorial)
How are “decisions and spontaneous actions (will)” implemented in the Braitenberg Vehicles?
By bizarre kinds of dependence of speed and stimulus intensity.