Modularity Slides

Modularity Marr’s principle of modular analysis “Any large computation should be split up and implemented as a collection of small sub-parts that are as nearly independent of one another as the overall task allows” (Marr, 1976, p. 485). Colour vs. Shape Agent Architectures Simple ref...

Modularity Marr’s principle of modular analysis “Any large computation should be split up and implemented as a collection of small sub-parts that are as nearly independent of one another as the overall task allows” (Marr, 1976, p. 485). Colour vs. Shape Agent Architectures Simple reflex agent Goal based agent Learning agent Key characteristics of modules 1. Domain Specificity Activated by a limited range of inputs Only processes inputs of a certain sort 2. Informational Encapsulation Cannot make use of information from outside the module during processing What happens elsewhere in the mind is not accessible to a module Informational encapsulation This Photo by Unknown Author is licensed under CC BY-NC-ND Key characteristics of modules 3. Mandatory Application Respond automatically to the appropriate stimuli Cannot be ”switched off’, are not under “executive control” 4. Fast Process information quickly Needs to be fast enough to control “online” action *5. Central Inaccessibility Key characteristics of modules 6. Fixed neural architecture Modules are often realized by specific brain regions 7. Specific breakdown patterns Modules fail in predictable, determinate ways *8. Innate Example: Universal Grammar A widely influential hypothesis due to Noam Chomsky The rules for understanding grammar are domain specific They do not depend on background knowledge We process grammar automatically and quickly Is all information processing modular? Central Processing 1.Quinean Holistic 2. Isotropic Not informationally encapsulated The limit of cognitive science? “the more global (i.e. the more isotropic) a cognitive process is, the less anybody understands it. Very global processes, like analogical rea- soning, aren’t understood at all” (1983: 107). Massive Modularity Hypothesis All information processing is modular No central processing Darwinian modules: specialized modules that evolved to solve a specific set of problems More complex types of information processing than Fodorean modules Could central processing have evolved? Could a domain general, central processing system have evolved to solve the type of adaptive problems that were faced in our evolutionary history? 1) The argument from error Fitness promoting behavior varies form domain to domain, criteria for fitness embedded in domain specific modules 2) The argument from statistics and learning Special purpose kin selection law must be “built in”