Intro to Vision Science LOs Flashcards
Why Study Perception
To understand human physical structures/processes (sensory systems/brain)
To understand and overcome errors in observation (eg. Visual illusions, real-world implications of illusions)
To create substitutes/auxiliaries for impaired senses (braille,hearing aids)
To replace the real world with substitutes (stereo, surround sound)
To replace the observer with substitutes (robotic vision)
What is Perception
Conscious experience of objects; organization and interpretation of neural signals
Recognition or identification of an object based on sensations (sensory experience elicited by a stimulus)
Assume perception is information processing
More than the analysis of perceptual properties and performance
Goal of perception is to recover meaningful representations of the world on the basis of physical stimulation that originates in the world
What is Cognitive Science
Scientific study of the mind as an information processor
Interdisciplinary field consisting of: anthropology, computing science, education, linguistics, neuroscience, philosophy, psychology
Studies: AI, attention, language processing, learning/development, memory, perception and action
Research methods include: traditional behaviour experiments, brain imaging, computational models, neurobiological methods
What are the three implications of Cognitive approach to the study of perception
The mind can process info (data) according to a set of rules/procedures (program)
Implementation account
Formal Account
Interpretational Account
Implementation Account
physical structures involved in info processing
Wet ware of the brain: How can perceptual info be related to bio structures
Implementation is not enough,may lead to incomplete theory, not general enough to capture individual commonalities, describes but not explain
Perceptual phenomenon can be reduced to causal laws of neurophysiology
Formal Account
underlying functions and knowledge structures
Information processes are formal meaning they are represented in symbols and abstract structures
Symbols have no intrinsic meaning but token type is important (eg. words “run vs rum’ (verb vs noun)
Rules are formal; they only depend on token type (eg. a game of chess, pieces form is arbitrary, the game depends on the token type and how the pieces can be moved)
Focus should be on percepts/internal symbols representing the external world and on procedures used to solve info processing problems
Describes data and computer program used by perceptual system
Seeks to answer the questions: what symbolic structures are used to represent info, what procedures are used to manipulate these structures, how are these procedures coordinated and controlled
Interpretational Account
described in terms of behavior observed
Assigning meaning to a formal system
Computational description: what a system is doing
Answers questions: what info processing probleming are being solved by a perceptual system, what is the system doing and why (includes behaviours)
Describe the tri-level Hypothesis
Assumes that perception is information processing
Must describe perception at each of three levels of computation, representation/procedure and implementation
Must specify relationship amongst three levels
Must identify functional architecture for perception
3 Levels
1) Computational Theory: behavioral level, goal of computation, why appropriate, what is logic of strategy by which it can be carried out METHODS: logic, mathematics
2) Representation / Algorithm: functional level, how can goal/behavior be actualized, what is algorithm for transformation, representation for input and out
METHODS: psychophysics, computer simulation
3) Hardware Implementation: how can representation/algorithm be realized physically, structures in brain involved
METHODS: neuroscience, electrical engineering
What is the functional architecture (FA)
Basic steps that comprise a given function (primitive operations that make up info processing system)
FA is part of the procedural description of a system
Is a bridge from the representation/algorithm to the implementational description
WHY SPECIFY FA: all info processing systems have a FA, FA is required to make an info processing account an explanation, required to create strongly equivalent models
Is a functional account (not need specification of physical mechanisms)
Discovered by using functional analysis (takes complicated functional description and decomposes it into simpler subfunctions)
How does the FA provide an escape from Ryle’s Regress
Ryle’s Regress: just descriptions, not providing explanations at all
Functional Architecture meets both the subsumption criterion and the cognitive penetrability criterion so it exits Ryle’s Regress
Subsumption Criteria
Subsumption criterion: if a sub function can be explained using natural, causal laws then decomposition can stop
Eg. colour vision described functionally in trichromatic theory but was not subsumed until all three different photopigment molecules were isolated in 1960s, no need to explain how molecules are photosensitive
Cognitive Penetrability Criteria
if a perceptual phenomenon cannot be altered by change in beliefs then function is impenetrable. If conscious top down processing cannot change then it is bottom up processing which must be handled by FA
Eg. simultaneous colour contrast illusion, even if you know what the colours actually are it does not change experience of illusion, even though you know the two red crosses are the same red the one on the left on the yellow looks darker and the one on the right on the blue looks lighter
What is the difference between Strong and Weak Equivalence
The Representation / Algorithm Stage is either the same (strong) or different (weak) - The Formal Account stage
Does a system accomplish a task in the same way humans do: Yes (Strong), No (weak)
Strong equivalent: models that exhibit the same behaviour as humans for the right reasons, same underlying fundamental processes being used. Has the computation / representation/algorithm stage the same, implementation will always be different
Weakly equivalent: AI accomplishes a task but not in the same underlying way as humans, passes tests for the wrong reasons from a psychological perspective. Has the same computation but diff representation/algorithm and implementation
What are some themes in Vision Science
Myth of seeing as a faithful record: vision is not like a camera, if this was true visual illusions would not exist
Myth of vision as a passive process, percept must be constructed from multiple stimuli so even though we don’t consciously put in effort it is an active process
Myth of the seeing eye, where does perception take place (IN BRAIN NOT EYE)
Myth of imageless thought: visual imagery is important for cognition, without visual imagery you can’t answer how many windows are visible approaching the front door of your house
