LCS 11: Cognitive Science - Modules - Claremont Colleges · Pomona College LCS 11: Cognitive...
Transcript of LCS 11: Cognitive Science - Modules - Claremont Colleges · Pomona College LCS 11: Cognitive...
Pomona College
LCS 11: Cognitive ScienceModules
Jesse A. Harris
February 5, 2013
Jesse A. Harris: LCS 11: Cognitive Science, Modules 1
Agenda
Group question 1.31. Leader discusses response to group2. Group discussion with summary of points3. Class discussion
Modules1. DeVnitions2. Examples from visual processing
Reading for next class
Response paper #1
Jesse A. Harris: LCS 11: Cognitive Science, Modules 2
GQ 1.3
GQ 1.3Take a simple action or mental process of your choosing andidentify at least three distinct subprocesses that might beinvolved in performing that action or process. Which, if any, ofthose subprocesses might be called modular in Fodor’s (1985)sense, and why?
Group leadersSam, Audrey, Lea Lynn, Daniel, Joel, Noah, Jun, Sarah, Paul
Jesse A. Harris: LCS 11: Cognitive Science, Modules 3
Cluster of concepts
1. Domain speciVcity
2. Informational encapsulation
3. Mandatory operation
4. Fast processing
5. ‘Shallow’ outputs
6. Fixed neural architecture
7. Characteristic and speciVcbreakdown patterns
8. Characteristic ontogenetic paceand sequencing
Jesse A. Harris: LCS 11: Cognitive Science, Modules 4
Subcluster 1: Modules are snobs
Domain speciVcityA module operates over or responds to only a narrow range ofinputs. A special purpose mechanism that provides answers tovery speciVc questions.
Is there an edge in a particular area of the visualVeld?
Jesse A. Harris: LCS 11: Cognitive Science, Modules 5
Subcluster 2: Modules don’t care what you think
Information encapsulationInformation processed within a module cannot be accessedfrom information outside that module.
Modules do not consider other types of informationoutside their speciVc domain, including top-downknowledge.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 6
Subcluster 3: Modules are dumb
MandatoryOnce the module receives its input, it automatically operateson the information it receives.
FastAlthough a relative term, modules Vnish their computationsquickly.
ShallowModules produce constrained output that wascomputationally cheap to produce.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 7
CAT
Jesse A. Harris: LCS 11: Cognitive Science, Modules 8
Subcluster 4: Modules are innate
Fixed neural architectureFunction associated with module is localized to a speciVc setof neural structures.
Characteristic breakdownModules are independent: when they fail, they may do soindependently of other cognitive functions.
Characteristic paceModules tend not to be learned, but rather ‘triggered’ byspeciVc developmental stages.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 9
In short
Modules are dumb snobs, who don’t care what you think. Andyou have to put up with them.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 10
Thatcher illusion (Thompson, 1980)
Jesse A. Harris: LCS 11: Cognitive Science, Modules 11
Thatcher illusion (Thompson, 1980)
Jesse A. Harris: LCS 11: Cognitive Science, Modules 12
Thatcher illusion
Jesse A. Harris: LCS 11: Cognitive Science, Modules 13
Vertical pathway Horizontal pathway
Photoreceptors (rods & cones)
Bipolar cells
Retinal ganglion cells
!"#$%"&'()*+,))-*
./(+#$&,*+,))-*
Hubel & Wiesel, 1959. Domain speciVcity.
http://www.youtube.com/watch?NR=1&v=4nwpU7GFYe8
Jesse A. Harris: LCS 11: Cognitive Science, Modules 17
Lateral geniculate nucleus (LGN
Two kinds of neuron in LGN
1. Magnocellular neurons.Deeper within LGN; thickeraxons, which allow a quickerresponse. Project onto wherepathway.
2. Parvocellular neurons.Shallower within LGN;thinner axons, respond moreslowly to input. Project ontowhat pathway.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 19
http://www.youtube.com/watch?v=nFJvXNGJsws
Split brain studies. Information encapsulation.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 21
Split brain studies. Information encapsulation.
http://www.youtube.com/watch?v=ZMLzP1VCANo
Jesse A. Harris: LCS 11: Cognitive Science, Modules 22
Stroop task. Modularity par excellence.
Stroop taskName colors of word, not the word itself. If you see the wordBLUE, say “Red” not “Blue”. If there is no word, just say thecolor that you see.
Self-demoCover up all but the Vrst column with a separate piece ofpaper. When I say START, start the task silently. When youare done, look up and jot down your response time at thebottom of that column. We’ll then do the same for the othercolumns.
http://online-stopwatch.chronme.com/
Jesse A. Harris: LCS 11: Cognitive Science, Modules 23
Stroop task. Modularity par excellence.
Stroop eUectTakes longer to name the color when the printed word isincongruent. Also, more likely to misname item.
ModularityHow is this result compatible with modularity?
Jesse A. Harris: LCS 11: Cognitive Science, Modules 24
Scope of modularity
Central processing systemsInput provided by modules is evaluated, assessed, andtransformed by more global mechanisms, which areresponsible for interpretation.
. . . Nature has contrived to have it both ways, to get thebest out of fast dumb systems and slow contemplativeones, by simply refusing to choose between them. That is, Isuppose, the way that Nature likes to operate: “I’ll havesome of each” – one damned thing piled on top of another,and nothing in moderation, ever. (Fodor, 1985: p. 4)
Jesse A. Harris: LCS 11: Cognitive Science, Modules 26
Scope of modularity
Modest modularityModularity limited to simple input-output systems, systemsthat translate an external stimulus into information to form amental percept. Such systems are sometimes called sensorytransducers.
Fodor’s claimModules are the only parts of the cognitive system that wecan ever hope to fully understand within a scientiVc theory.Non-modular components cannot ever be captured by a fullyscientiVc theory.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 27
Massive modularity
I No central processorI Mind is entirely modular, but with a
weakened notion of moduleI Information encapsulation is no longer
a property of a module
1. Biological systems are designed incrementally,requiring modular organization
2. Human minds are ‘extensions’ of animalminds, which are modular
3. Argument from computational tractability
See Cosmides and Tooby (1992) and Carruthers (2006)
Phrenology
Jesse A. Harris: LCS 11: Cognitive Science, Modules 28
Some issues with modularityIssue 1: Neuroplasticity.The brain shows some degree of Wexibility; functional deVcitscan sometimes be recovered by remapping or recruiting otherneural areas.
Issue 2: Domain speciVcity.Really part of issue 1: Multiple areas of the brain seem to takein multiple sorts of inputs.
Issue 3: Limitations on science.We’d like to be able to study more than just sensorytransducers – what happens during higher order processingand how does the mind make sense of that low levelinformation?
Jesse A. Harris: LCS 11: Cognitive Science, Modules 29
Next time: Connectionism & Neural nets
No GQ for next time
Reading is diXcult, so try to get a feel for how neuralnets work, without sweating the details.
Writing response # 1 due by Friday at 5PM – upload toDropbox on Sakai.
Jesse A. Harris: LCS 11: Cognitive Science, Modules 30