CogSci_13OrganizOfMind PDF

Organization of Mind: Modular and Non-Modular Information Processing I. Fodor’s modular and nonmodular processing Characteristics of modular processing Visual illusions as evidence of automatic specialized modules Nonmodular central processing II. Massive modularity hypothesis − Cheater detection mo...

Organization of Mind: Modular and Non-Modular Information Processing I. Fodor’s modular and nonmodular processing Characteristics of modular processing Visual illusions as evidence of automatic specialized modules Nonmodular central processing II. Massive modularity hypothesis − Cheater detection module and Wason card problem III.ACT-R: Hybrid cognitive architecture A. Symbolic modular architecture B. Subsymbolic base Modular & Nonmodular Processing Jerry Fodor’s The Modularity of Mind (1983) Rejected idea that the mind is organized in terms of faculties such as memory and attention that can process any type of information Instead, proposed the existence of specialized information-processing modules for things like − Color perception − Shape analysis − Visual guidance of bodily motions − Grammatical analysis of heard utterances − Detecting melodic or rhythmic structure of acoustic arrays − Recognizing the voices of conspecifics v Modules operate at a low level and work quickly to provide rapid solutions to highly determinate problems v In addition, there are high-level, open-ended nonmodular processes that can bring a wide range of information to bear on very general problems Modular processes are characterized by Domain-specificity − Modules are designed to carry out very specific and circumscribed information processing tasks − Can only operate on limited range of inputs Informational encapsulation − Not affected by what is going on elsewhere in the mind − Cannot be “infiltrated” by background knowledge and expectations or knowledge from other modules Speed − Modular processing operates quickly and efficiently, e.g., shape analysis Mandatory application − Respond automatically − Cannot be “switched off,” e.g., perception of visual illusion Visual illusions provide support for theory that certain aspects of visual perception may be modular Ø Café wall illusion: are the gray lines slanted or straight? ➜ ➜ Not clear how this illusion works, but one theory is that neurons in the brain interpret a brightness contrast between tiles as a small wedge, making the lines appear slanted Ø Muller-Lyer illusion: Is line AB or line BC longer? ➜ Our brains are used to perceiving angles as corners that are near or far away and sees the inwardfacing corners as more distant and therefore smaller Ø Is Tile A or Tile B darker or are they the same color? ➜ ➜ Illusion results from visual system’s attempt to maintain lightness constancy: we perceive an object as having a constant color, even if changing illumination alters the wavelengths reflected by the object Modular processes are also usually characterized by Fixed neural architecture − It is sometimes possible to identify determinate regions of the brain that are associated with particular types of modular processing Ø Ex: fusiform face area for face recognition Specific breakdown patterns − Modules can fail in highly determinate ways, which provide clues on the form and structure of processing Ø Ex: prosopagnosia Ø Prosopagnosia: failure to recognize particular people by the sight of their faces – After stroke, sheep rancher could not recognize people but could recognize sheep However, according to Fodor, not all cognition is carried out by modular mechanisms Modules provide inputs to nonmodular central processing The latter can evaluate and correct the outputs of cognitive modules Ø Ex: evaluating beliefs and decision-making Massive Modularity Hypothesis According to the massive modularity hypothesis, The mind does not do any central processing; instead, all information processing is modular The human mind is a collection of Darwinian modules, that is, specialized modules, each of which evolved to solve a specific set of problems encountered by our primitive ancestors Typically, Darwinian modules engage in more complex types of informationprocessing than Fodorean ones and are not informationally encapsulated Ø Examples include: − Emotion detection − Intuitive mechanics or folk physics o Innate understanding of basic principles governing the behavior of physical objects, e.g., that objects don’t “magically” appear and disappear − Folk psychology − Cheater detection Evidence for the massive modularity hypothesis comes in part from research indicating that humans tend to be much better at reasoning with deontic conditional than they are with ordinary, nondeontic conditionals Deontic conditionals have to do with permissions, prohibitions etc. Ø Ex: If you are drinking beer, then you must be over 21 years of age This phenomenon is illustrated by the Wason card problem: Statement: “If a card has vowel on one side, then it has an even number on the other side” Which card or cards below would you need to turn over in order to find out whether this rule is valid? Answer: A and 9 « Reframing the selection task in such a way that what is being checked is a condition having to do with permissions, entitlements, and/or prohibitions improves performance Statement: “If a person is drinking beer, they must be over 21 years of age” Each card has the drink on one side, and the person’s age on the other side: You’re the bouncer… Which of the cards below should you check out? Answer: ➜ 73% of students who tried drinking-age problem made correct selections, as opposed to 0% in the standard, abstract form of task One theory of why people are better at reasoning with deontic conditionals than with nondeontic conditionals is that when they solve problems with the deontic conditionals, they are using a specialized module for monitoring social exchanges and detecting cheaters v This is the cheater detection module (Cosmides and Tooby) Why should there be a cheater detection module? Cooperative behavior presumably has a genetic basis However, an individual who takes advantage of cooperators without reciprocating will likely do better than one who cooperates Ø Ex: They gain your trust, then steal all your bananas So how could the genes for cooperative behavior ever have become established? ➜Enter the cheater detection module… ACT-R: Hybrid Cognitive Architecture ✧ “Adaptive control of thought – rational” ✧ Cognitive architecture with modular organization that was developed by John R. Anderson in 1976 ✧ It is hybrid in sense that it incorporates both symbolic and subsymbolic information processing Perceptual-motor layer Perceptual module in turn consists of a visual module, audition module, etc. Motor module consists of speech module, manual module, etc. Communication between modules on different layers takes place via buffers (workspaces) Cognitive layer Declarative memory is organized in “chunks” Procedural memory is encoded as production rules: actions for the system to perform, e.g., retrieve a chunk from memory, send a command to the motor module to perform an action − Production rules can be nested within each other, so that output of a given production rule will trigger firing of another production rule ★ All of the above modules are encoded in the form of physical symbols What makes ACT-R a hybrid architecture is that the symbolic, modular architecture is run on a subsymbolic base ACT-R is designed to operate serially, so that at any given moment, only one production rule can be active, but how does it select that one? − The pattern-matching module controls which production rule gains access to the buffer by working out which production rule has the highest utility at the moment of selection, as determined by o How likely the system is to achieve its current goal if the production rule is activated o The cost of activating the production rule − These calculations are subsymbolic and use an artificial neural networks approach Subsymbolic equations are also used to model how accessible information is in declarative memory − The basic units of declarative memory are chunks, but each chunk is associated with a particular activation level, which is determined by o How useful the chunk has been in the past o How relevant the chunk is to the current situation and context ✧ The general information processing that takes place in the buffers is symbolic ✧ In contrast, the calculations that determine whether or not a particular item of knowledge ends up in a buffer are subsymbolic Video Reference Video excerpted from: It's White and Gold! https://www.youtube.com/watch?v=wIZHRO0MZvI