96 Cards in this Set
| Front | Back |
|---|---|
|
mental representation
|
a) structure information that has meaning because it refers to something else. Representation comes from the back of the brain.
B) example: map bc spacial relations on a map are similar to the real world
c) give our minds an efficient way of computing real world stimuli through these si…
|
|
Marr's levels (including the 3 levels: computation, algorithm, and implementation). Why is Marr's formulation relevant to cognitive science?
|
(Highest to lowest)
a)Computational theory- what the goal of the computation is, the purpose/function of something.
Representation and algorithm- how can this theory be implemented? What is the representation for the input and the output? What is the algorithm for the transformation?
H…
|
|
turing machine
|
a) a hypothetical device for instantiating an algorithm. There are symbols on the tape, the tape is unlimited. Uses a finite set of rules and purely mechanical procedures. It can do algorithms (deterministic)
b) a minimal computer
c) Universal turing machines can compute anything comput…
|
|
algorithm
|
a) Purely mechanical procedure for solving a problem
b) instructions for programing a DVD player
c) The efficiency of the algorithm is the basis for the efficiency of the computation.
|
|
mental imagery
|
a)There is a linear relationship between the length of time the subject takes to solve the problem and the degree of rotation between the two figures.
b) we are able to figure out if a box can fit into our car
c) Gave rise about the format in which information is stored and the way in w…
|
|
parts of a turing machine
|
Tape: no constraints on how long it is (infinite storage)
Head: reads/writes/changes symbols
Table: finite (limited) set of rules/functions
State
|
|
universal turing machine
|
a) Can compute anything that is possible to be computed.
b) a computer
c) proof that neurons are capable of doing all the information processing to support thought (does not necessarily prove that neurons actually do it, just that they are capable of doing so). Shows there is no need f…
|
|
Four aspects of PSS (Physical Symbol System)
|
1) Symbols
2) Expressions / symbol structures
3) Rules & processes for manipulating symbols
4) Rules & processes can be represented within the system
|
|
PSS hypothesis
|
a) A PSS has the necessary and sufficient means for general intelligent action. It is able to solve problems by transforming symbol structures (in sophisticated enough ways).
b) chess (when it is your turn you must make a move, what's the best move to make to win the game? Pawns/castles…
|
|
dorsal-ventral stream (what-where)
|
a) computation: what and where
Dorsal stream: where (going from back of brain upward)
Ventral stream: what (going from back of the brain downward)
c) Makes the brain more efficient
|
|
specialization of function
|
a) reduces interference that arises from computational incompatibility
b) if the brain had one specific area that only inputted and outputted one specific thing
c) makes the brain more efficient
|
|
modularity
|
a) have things in modules for special purposes and specific types of input. There is no communication among the modules.
b) if the brain had one specific area that only inputted and outputted one specific thing
c) Makes the brain more efficient
|
|
4 lobes of the brain
|
Frontal lobe- behavior and planning
Temporal lobe- auditory perception, language, visual memory, declarative memory, and emotion
Parietal lobe- awareness of body in space, motor tasks
Occipital lobe- vision
|
|
computationally incompatible
|
a) Specialization of function reduces interference that arises from computational incompatibility.
b) talking on the phone and driving
c) One computation causes problems for the other computation, can make less clear and efficient and disrupt other processes.
|
|
computationally efficient
|
a) Modularity is more computationally efficient.
b) Chewing gum and walking
c) The two computations together make the tasks more clear, causes less problems.
|
|
4 criteria proposed by Fodor
|
Modularity (FODOR)
-Domain Specificity
- one of a kind info as input
- selective
-Informational encapsulation
- protected from other ongoing events
- no interference possible
-Obligatory firing
- start the computation as rapidly as possible if conditions are met
-Fast speed
- ef…
|
|
parallel
|
a) Two task are completed at the same time
b) processing shape and color
c) Makes the brain more efficient
|
|
serial
|
a) Completing one tasks then completing the other
b) reading a book then watching TV
c) Makes the brain less efficient
|
|
feature map
|
a) a form of representation that emphasizes a particular feature
b) traffic street view is only about traffic
c) they are low level features
|
|
cognitive map
|
a) First proposal for explaining behavior in terms of representations (stored information about the environment)
more complex
b) a mental map that represents many things at once.
c) representations are one of the fundamental explanatory tools of cog sci
able to be highly processed
…
|
|
memory taxonomy
|
long term
- declarative
- episodic
- semantic
- non declarative
- procedural
- perceptual representation system
- classical conditioning
- non-associative learning
short term
|
|
What does it mean that "memory is a process"?
|
a) Memory is a process, not a thing (it's not tangible).
b) Encoding - consolidation - retrieval
c) Important if we want to remember anything we learn
|
|
consolidation
|
Stabilizes a single memory and retains it for later retrieval
|
|
encoding
|
a) the process of acquiring info and transferring it into memory
b) in class taking note and learning
|
|
retrieval
|
a) remembering something after it has been encoded
b) losing your keys - think back to the last time you had them, retracing your steps to find them
|
|
Broadbent's Model of Attention
|
Info comes through senses and passes through a short-term before passing through a selective filter
- makes it through selective filter, goes to the limited capacity channel
cocktail party phenomenon
- know little about conversation they are not attending to - they hear words but don't…
|
|
representation
|
a) very general term; structured information such as an abstraction or symbol (e.g., a word, a value in a computer program).
b) Example: a map
c) Important because it has meaning because it refers to something else.
|
|
distributed representation
|
a) When a concept is represented by a network that is part of a larger group of related things. Not located in one space, is distributed among many different places.
b) Example: multiple nodes creating a concept.
c) Importance: (core challenge) to explain emergent properties (higher le…
|
|
computation
|
a) a function that, when given an input, will produce an output. A purely mechanical procedure for manipulating information.
b) Example: neural network.
c) Importance: computations are able to use simulations to explain how systems are structured.
|
|
simulation
|
a) A way of imitating a process or change in the real world to predict what will happen or explain what did happen and why.
b) Example: imagining chess moves by going through them in your mind (not actually moving the pieces) to decide what move to make.
c) importance: can be used to m…
|
|
category
|
a) a group of related things. Computationally straight-forward as an emergent property of networks.
b) an example would be placing a thing in a group such as 'friend or foe?' or 'guilty or not guilty.'
c) Important because the representations of inputs coming into a network enables the…
|
|
exemplar
|
a) specific item or instance. Members of a category that a person has encountered in the past. Member of a category that is the most representative of the category as a whole.
b) Example when you think of a bird you think robin (canonical) not penguin (unusual). A golden retriever would…
|
|
emergent property
|
a) a "big picture" property of a system that is determined by the combination of behaviors of lower-level features interacting.
b) Example: the heart is made of heart cells but the heart cells on their own do not have the property of pumping blood. The heart is only able to pump blood m…
|
|
semantic network
|
a) concepts stored in the mind in terms of connected ideas. Direct links from higher order categories to exemplars. All the ideas in the mind are connected together. Its activity is spreading activation.
b) Example: The word "bird" is associated with different concepts, such as robin, o…
|
|
spreading activation
|
a) activating a concept that pulls up related, available concepts along with it. Certain connections between nodes can be stronger than other connections.
b) Example would be saying the word "nurse" and then the words with the strongest connection such as "doctor," or "medicine" are act…
|
|
attractor network
|
a) Has a specific structure of interconnections that can settle into a stable pattern. You can train the model meaning it can learn multiple exemplars.
b) Example: Coins rolling down an attractor basin have high energy and low harmony but when they hit the bottom they have very low ener…
|
|
attractor state
|
a) the network self organizes and wants to be stable. If changes occur, the pattern in the network reorganizes itself to regain stability.
b) Example: a coin rolling down an incline, when it reaches the bottom it's at a stable state.
c) Importance is that higher level things computatio…
|
|
pattern completion
|
a) Neurons that are activated to represent a stored memory.
b) an example is spreading activation (attractor network).
c) Important for memory retrieval.
|
|
neural network model
|
a) a computing system made up of simple, interconnected units which process information in response to external inputs.
b) A simple example of this is a perceptron for a more complex model there would be hidden layers.
c) Importance: can be used to establish the minimum structure and p…
|
|
node/unit
|
a) elemental unit of processing. Represents a concept (one thing).
b) An example of nodes would be like a marching band, each one of the band member's activity contributes to the overall pattern.
c) They're important because each one does its job so the system works.
|
|
layer
|
a) multiple nodes that represent a concept.
b) An example would be an input layer, output layer and a hidden unit.
c) Important because it's distributed representation. Making a simple computation into a more complex one.
|
|
input layer
|
a) What's put in and operated on by any process or system.
b) Example is vision -the retina in the eye.
c) Without the input there would be no computation for hidden layers or output layer.
|
|
output layer
|
a) A place where information leaves the system.
b) Example would be vision - the image you're seeing.
c) Without the output layer the hidden layers wouldn't be able to be interpreted leaving a system with no result at the end.
|
|
hidden units
|
a) in multiple layer networks there are hidden units. They learn/adapt by changing their weights ~ but architecture does not change. (can only go one way X-> X-> X not X-> <-X).
b) An example is a turing machine.
c) Hidden units are important because they can learn to compute anything …
|
|
perceptron
|
a) models with only input and output layers. Can only compute linearly separable (simple) computations. They are able to learn. Multiple perceptrons create multi-layer networks.
b) An example is a AND/OR function.
c) Important because to have a multilayer network you need an input and …
|
|
linearly separable function
|
a) Classes that are separated with a single line. Refers to the fact that classes of patterns can be separated with a single linear line.
b) An example would be relatively simple computations such as a OR/AND functions.
c) Important for perceptrons so they categorize sets of patterns.
…
|
|
XOR function
|
a) (not linearly separable) operates on truth values. Can be computed by a perceptron. True if one and only one is true.
b) example: do you want a reeses or a snickers? In logic xor: one or the other, not both.
c) Important because we can train XOR
|
|
learning in neural networks
|
a) When teaching a neural network the connections between nodes are able to change but not the number of nodes or structure. The connection between the nodes allows more or less activation to flow from one node to the next. It has no idea what you're trying to teach it but it guesses. It …
|
|
feed-forward activation
|
a) Activation spreads forward through networks. This is how to change hidden units to map input/outputs. Depends on activity in earlier layer/nodes and activation of function.
b) Example: if the input layer is stimulated then it activates the hidden layer which activates the following l…
|
|
connection weights
|
a) these regulate how much activity is necessary for one layer to activate the next. Uses discrepancy to reduce error.
b) Example: the connections between two layers.
c) Important because without it networks would not be able to learn.
|
|
back-propagation
|
a) after each training cycle, the output mismatch (error) is used to adjust the weights.
b) An example is learning a language, the weights must be changed. You don't understand the language right away.
c) Its importance is it can train mappings between any number of layers so it can co…
|
|
reward
|
a) indicator of value, useful for "shaping behavior."
b) Example: classical conditioning.
c) important because it's useful for computing adaptive/intelligent actions.
|
|
errors of reward-prediction
|
a) indicator of value, useful for "shaping behavior."
b) Example: classical conditioning.
c) important because it's useful for computing adaptive/intelligent actions.
|
|
orbitofrontal cortex
|
a) part of the frontal cortex used for primary and secondary representations of reward
|
|
nucleus accumbens / striatum
|
a) part of the brain very important for reward
|
|
dopamine (DA)
|
a) Important in reward, carries a reward-relevant information signal: errors of reward-prediction. Can be released in bursts, with firing rates going above or below a baseline rate to signal unexpected rewards being present, or absent.
|
|
dopamine responds to different stimuli
|
- no prediction/reward occurs
- DA neural activity is most active when the reward is presented
- DA activity before and after the reward is presented is neutral
- reward predicted/reward occurs
- DA neural activity is most active when the conditioned stimulus (when they're told that…
|
|
computational role of DA
|
a) carry a temporal-difference (TD) error information across learning trials.
b) Example: computer models have used TD methods to learn to play backgammon.
c) Importance: errors of reward prediction allow the system to learn/predict which features in the environment are associated with…
|
|
human social cognition
|
a) 5 parts
Species general:
- ability to recognize conspecifics generally and individuate (recognize others in their own species)
- ability to monitor others actions, goals, and intentions
- ability to engage others contingently
Uniquely human:
- theory of mind skills
- joint atte…
|
|
cultural intelligence hypothesis
|
a) your brain size correlates to the size of your social group. human number is 150 on the scale
b) example: humans have a wider social network than animals do - lions have a pride, and that's it.
c)importance: an explanation for brain size and makes us uniquely human.
|
|
social learning in humans
|
a) solving a simple but not obvious problem by observing a demonstrated solution.
b) example: baby claps hands after mother does so.
c) importance: a theory of how people learn.
|
|
mirror neurons
|
a) Fire when you do an action, and fire just as strongly when you observe someone do the same action.
b) example: if someone is moving their hand, your brain fires neurons in the same place as when you move your own hand.
c) Importance: use the system for self to understand others; un…
|
|
the self (from cog sci perspective):
|
a) the self is just a computation; it is an attractor state rising from constraint satisfaction. Not a thing, but an idea.
physical body:
represent, recognize (can recognize self in the mirror, can identify pain, etc)
Memory
episodic (autobiographical) memories (memory over time)
se…
|
|
what is hard for cog sci to explain about the "self"?
|
a) The main thing that is challenging is consciousness. Philosophical issues make it tricky. Attribution of consciousness (we don't actually have consciousness we just think we do). We can't really observe consciousness. We can think our pets are conscious be we really don't know.
b) Exa…
|
|
Proprioception
|
a) awareness of position in space.
b) Example: touch nose eyes closed, blind people can eat fine.
c) Importance: reinforces that "this body is me".
|
|
constraint satisfaction
|
a) the process of finding a solution to a set of constraints that impose conditions that the variables must satisfy.
Very general and can be done by attractor neural networks (same as we saw for categorization)
Constraint: a fact, clue, partial info; satisfaction: account for as many fa…
|
|
the rubber hand illusion
|
a) our body has representations and a visual map, and these two senses usually tell us the same story. keep hand very still, propricental sense fades away, so visual sense sees rubber hand and makes you think that the rubber hand is yours
b) example: real arm is hidden out of sight and a…
|
|
autobiographical memory
|
a) memory from specific points in the past; a combination of episodic and semantic memory.
b) example: remembering buying your first car.
c) importance: it allows us to learn from experience and provides us with identity.
|
|
subjective continuity
|
a) we are the same person over time.
b) example: when you go into unconscious sleep, you still wake up with self awareness.
c) Importance: don't have to restart every day, don't wake up thinking "who am I?"
|
|
self-related encoding
|
a) going about life and evaluating in respect to yourself. kind of mental processing that brings the self into a decisions
b) example making a judgement on a word - sporty ... am I sporty?
c)Importance: it makes memory stronger, aids memory
|
|
agency
|
a) interpreting one's thoughts as the cause of one's actions.
b) example: a puppet doesn't have agency, but a puppeteer does; voluntary action.
c) importance: sense of free will & capable of doing things
|
|
illusion of agency
|
a) interpreting one's thoughts as the cause of one's actions.
b) example: two people controlling mouse pointer. Felt like they had agency, but didn't really have any. Someone says it's going to rain - and all of a sudden it rains, they think it made it happen. Superstitions
|
|
self-recognition
|
a) mental process of identifying yourself
b) example: recognizing my own voice
c) importance: re enforces the concept of yourself
|
|
internal representations (maps) of the body
|
a) the representation is inside our brain; when we see things, we have internal representations.
b) example: map inside cognitive system contains a representation of our arm. Write MSU on piece of paper, and that paper represents the university.
c) importance: we know we have an arm, …
|
|
default mode network
|
a) network of interacting brain regions known to have activity highly correlated with each other and distinct from other networks in the brain. key brain regions: medial prefrontal, interior parietal, posterior angulate, temporal poles
b) Example: daydreaming, mind wandering.
c) Import…
|
|
change blindness (demos)
|
a) one things vs. another. For a while you can't tell the difference, then later think "I can't believe i didn't see that'. The information is present but we are not conscious of it.
b) the airplane wing with the engine disappearing. You know there is a difference but you can't tell what…
|
|
qualia
|
a) the "what its like" subjective experience part of being a human.
b) example: the redness of red
c) importance: juicy hard part of consciousness.
|
|
mind/body problem
|
a) two facts; there is your body and there's what it's like. Mind and body are closely related, but it's difficult to understand the relationship between the two.
b) example: brain damage can change one's personality
c) importance: very compelling that there are two things that are fu…
|
|
supervenience
|
a) Relationship of body and mind. Brain causes mental states. The mind is an emergent property of the brain. Supervenience does not hold when there is one brain state with two or more mental states - something else is needed to explain the difference between the mental states.
b) exampl…
|
|
Graziano: attribution of consciousness
|
a) we have attention and a representation of attention. it's the representation that is the basis for knowing we're really conscious because we can identify the representation. qualia is an illusion.
the brain attributes the property of awareness to itself; we only believe we are consci…
|
|
global workspace
|
a) Emphasizes consciousness. We can combine a lot of information together in novel ways which gives us a workspace to transform things.
b) Example: my social psych class relates to this class so consciousness is a workspace for me to relate the two together if I'm actively thinking about…
|
|
interpreter (narration / interpretation of ongoing experience)
|
a) your consciousness writes an explanation for all that you do.
b) example: split brain patients; sign says laugh, so she laughs, but she can't explain why she laughs so she fills in narration with a lie (but she doesn't think she's lying).
c) an explanation for how your conscious wo…
|
|
visual neglect (neuropsych condition)
|
a) unaware that you're not paying attention. Ignore a side of space and/or objects, usually recover quickly.
b) example: right parietal damage; can only draw numbers on one side of the clock, even though you know they are there and are unaware of deficiency
c) Importance: reveals that …
|
|
signaling
|
a) reflexive reaction to the presence of a stimulus.
b) example: alarm calls in vervet monkeys.
c) importance: another way to communicate other than verbalization
|
|
Chimpanzee signing and dolphin signing
|
chimpanzee: able to teach many signs, but they level off , not sensitive to word order.
dolphin: can do more than chimps - sensitive to word order.
|
|
Chomsky's criteria for language
|
a) language has structured principles, used for the expression of thought, establishing social relationships, and communication of ideas. It has various physical mechanisms of which little is known. It has been integrated into a system of cognitive structure
b) Example: of structured pri…
|
|
evidence for universality / innateness of language
|
a) deaf children invent sign language
all cultures have a language
language development similar across cultures
we can identify sounds even in our mom's womb
|
|
tacit / implicit learning of language
|
a) Not taught, and people don't really know what they learn. Implicit learning is skill learning.
b) example: grammar.
c) importance: a way of acquiring a skill without realizing you have a skill.
|
|
what is a "grammar"?
|
a) simple rules that produce endless variability.
b) example: people agree on grammaticality of sentence they never heard before.
c) importance: it makes language special.
|
|
computational problem of language: ambiguity at multiple levels
|
-noisy signal: noise can mask language. example: when someone coughs and it makes narration more difficult to understand (background noise in general)
-hard to segment into speech parts - word ambiguity: really hard to segment morphemes, words, etc because it all blends together becau…
|
|
phonemes
|
a) A distinct unit of sound that distinguishes one word from another.
b) Example: the letters p and b in the words pad and bad.
c) Importance: They are an abstract underlying representation for segments of words. They are the basis of a language.
|
|
how do visual cues help disambiguate phonemes?
|
a) It is easier to understand speaker when you can see them as you hear them talk.
b) example: talking to someone on the phone vs. talking to them in person.
c) Importance: training children with visual and auditory cues together gives them a deeper level of phonological processing.
…
|
|
coarticulation
|
a) blending language together.
b) example: what are you doing becomes whuddya doing
c) importance: allows for super speedy quick talking
|
|
McGurk effect
|
a) It's harder to understand people without the visual. When the sound is ambiguous, you hear the sound that corresponds with the vision. Seeing him make a different sound legitimately makes you hear things differently.
b) Example: pairing one sound with another visual representation of…
|
|
mental rotation
|
a) computation: rotate something in mind
it has 2 possible algorithms:
- Representation - Just measure the angle: "mental protractor"; angle actual does not affect time taken
- Transformation - Simulate rotation until things line up: estimate the angular distance rotated; the actual an…
|
|
value map
|
a) A value map is in our head, and we assign a value to something in order to pick the right thing to do in a given situation.
b) getting opinions of the best restaurants in town and picking the best one based on collected data.
c) important because we can shape our behavior based on …
|
LIN 463: EXAM 1