COGS 200 Final: COGS200- Fais, Mackworth, Bittner- Final Notes

8 Pages
Unlock Document

Cognitive Systems Program
COGS 200
Tom Bittner

Classical Sandwich Model- Hurley (2006), A standard model of the mind that is vertically modular with a sharp distinction drawn between perception and action  Consistent with an account of perception as a staged process that produces generally accurate perceptual beliefs about the surrounding environment Dretske (1990)- distinguishes between sensory perception (thing vision) and cognitive perception (fact vision)  Ordinary cases of indirect perception: o Seeing that a drink is cold because of condensation o Seeing someone has brown hair by viewing a photo  Asks if all cases of ordinary perception are indirect i.e. mediated by appearances o Indirect realism- perception is a process resulting in awareness of objective facts via awareness of subjective facts  Dretske applies indirect realism to cognitive perception, but not to sensory perception o Direct realist on sense perception o Indirect realist on cognitive perception Alternatives to the “classical sandwich” model  Behaviourism  Horizontal modularity  Extended mind  Haugeland Stages of perception 1. Objects/scenes (sensations) 2. Experiences(knowledge representation) 3. Beliefs Gregory (1980) “Perceptions as Hypotheses”  Perception names a certain inward subpersonal state o Allows subject to make use of limited/restricted data (perceptual input) by means of extra- and inter-polating Does science model (the process of) perception?  Both produce symbolic structures that are true when they match physical reality  Three stages o Signal  Perception- a pattern of neural events that lead to sensations  Science- e.g. output of a lie-detector, which we interpret into data points about skin conductance, blood pressure, etc. o Data- represent particular, specific states of affairs. Are less than complete, by nature. Require completion (filling in)  Interpolation and extrapolation- fitting curves to data points  May use either bottom-up or top-down processes  Science- explicit/conscious inferences  Perception- unconscious inferences  Generate hypotheses (perceptions) which may be used as predictions o Hypotheses  Structure our accepted reality  Are organized selections of data  Are not confined to the data (or to the truth)  Are the discovery and creation of objects  Perception- these are concrete objects  Science- may be either concrete or abstract objects  both require inferences drawn from data  Kanizsa figures- given a display with gaps and hypothesize an eclipsing object to explain these gaps  Postulations  Alternative hypotheses available for two reasons:  Signal may be ambiguous; the stimulus characteristics may not fully determine the data  Data may be ambiguous; they may make more than one hypothesis equally probable Error and Illusion  Science: o Signal-level: instruments not calibrated o Data-level: rival theories  Perception: o Signal-level: binocular rivalry o Data-level- ambiguous figure (rabbit/duck drawing) Perceptions and hypotheses are linked to reality very indirectly  Causally linked at signal level  Inferentially or rationally linked when data is read (or misread)  A great deal of perception (and science) is fictional: generally useful but occasionaly wrong Perception…  Is a staged process  Smoothly accommodates illusion and misperception  Vindicates our introspective evidence from the stream of consciousness  Constitutes half of the classical sandwich model The Grand Illusion  Objectively the visual image is impoverished  Subjectively, the visual image is a high-definition, full colour, wide-angle snapshot of the scene before us  We do not notice any deficits Two kinds of seeing  Thing sight- people who say they do not see a difference are wrong. They do see the difference,  Fact sight- people are correct, they do not see that there is a difference or what the difference is  The questions is: are differences objects or are they facts  Change blindness is blindness to facts, not blindness to objects 3/7 Bottom up  Thing perception, sensation, impressions impinging on sense organs, aspects of the physical world Top down  Contributions from knowledge, reasoning, inference, logic, comparison, contrast, fact perception Perceiving language: bottom up  Infants don’t have a lot of top down knowledge to work with Perceiving language  Air vibrations actually do hit your ears: “thing hearing”  Every utterance of the same word is actually different  You have the belief you heard bear: fact hearing Allophone: a token sound Phoneme: made up of allophones (abstract concept)  Distinguish meanings  Language specific It is not in our best interests to keep on discriminating the tiny differences among allophones- we want categories (phonemes)  We discriminate better across categorical boundaries Perceptual narrowing  Hypotheses: o Infants are born “universal listeners”- they can discriminate all sounds  Thing hearing o Tune their perception to discriminate only the sounds in their native language by 12 months  Fact hearing- categories of phonemes- phonemes are a construct  Mechanisms of perceptual narrowing o Distributional learning  Infants track frequency of allophones they hear  Exemplar theory  Learn to make categories depending on the distribution of allophones  phonemes  Change the distribution, and the categories change  No longer distinguish allophones Perception/action link  6 month old babies who have their tongues suppressed have their ability to discriminate disrupted o perception of non-native contrasts involving the tongue is disrupted by restraining the tongue  there is a close connection between perception and action- not a sandwich 3/9 Perceiving language- adults  adults do have world knowledge that informs top-down processing. But they still perform “thing” hearing  adults are confronted with the same wide variation in vowel sounds, but they have acquired the categories of their native language  top-down processing implies exemplar-like record of lots of info: o gender, effects of vocal tract anatomy, speaker characteristics, dialectal characteristics perception/production link  why do we perceive things? o So that we can understand the world?  However, natural selection works on behavior, not on what you understand  Analogous to our choice of rational action, and not rational thought as the goal of AI o Perception is for doing  Implies a necessary link between perception and action  Perception results in automatic imitation  perceiving and producing language are not independent processes o sender doesn’t stop being a receiver while sending; hearer doesn’t stop being a sender while hearing o perceiving is for actions, and as a consequence, hearing primes speaking (automatic imitation) Why does the tongue suppresser interfere with infants perception?  Humans are wired so that perception primes action  Infants hear a sound, and perception of those sounds primes action (imitation) o Doesn’t matter that infants don’t have the dexterity to imitate yet, their perception still primes their readiness to imitate  with teether interference, infants experience a disruption of the perception/action link o enough to disrupt their perception of sounds Perception/production link- adults  part of the function of top-down processes is to regulate imitation Effects of perception on production  pronunciations shift when linguistic environment changes: geography  pronunciations shift when linguistic environments change: social Accommodation  imitation of words and structures Effects of frequency of use  reduction- high frequency expressions change faster e.g. I dunno o easier/simpler pronunciation  conserving effect- high frequency forms resist regularization o frequency of use makes a form resist paradigmatic language change 3/14 One traditional model of the mind  vertically modular: 1. input/perception 2. processing and storage/cognition and memory 3. output/action i.e. seethinkdo  perception and action are peripheral  genuine mentality is pure cognition, not interaction with the world classical sandwich  perception-cognition-action  activity of the mind is thought to be : o representational/rational, and therefore unified in a special way o private and free/autonomous reforming the classical sandwich 1. deny the inner and the representational altogether o Dennet’s opposition to “filling in” o Noë’s “Grand Illusion” 2. retain the inner, but rearrange it, denying its unity o utilization behavior o imitation o Brooks’ subsumption archietecture o Hurley’s shared circuits model o Dennett’s Multiple Drafts Model 3. Keep cognition and representation, but expand them out into the world, loosening autonomy o Clark and Chalmers’ Extended Mind Hypothesis 4. Haugeland (1998): Combine all three of these approaches and add some considerations from systems theory to argue that the mind is intimately embodied and embedded in the world o New understanding of the claim that “intelligence abides in the meaningful” Part 1: Deny representation altogether  Ways in which perception seems “staged” o Snapshot conception and the grand illusion  Snapshot conception: representations are produced from a very limited amount of information. Bits and pieces of information we gather are integrated into a buffer.  We construct, from impoverished input, a complete inner picture of what is before us o Filling in the blind s
More Less

Related notes for COGS 200

Log In


Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.