CogSci midterm I (copy)
Terms
undefined, object
copy deck
- Instinct blindness
- Cognitive scientists speak of instinct blindness, when people perform complex mental operations so instinctively that they do not realize the complexities involved. Take the example of seeing and recognizing a chair. It's instinct blindness to think that seeing and recognizing a chair is simple, yet myriad processes are involved, from adjusting to the amount of light falling on it to conceptualizing what a chair is.
- Curing Instinct Blindness
- Computer Science: When you have to implement a process, you see how hard it really is! Neuropsychology: ie. brain damage --> when brain is functioning normally, it's hard to see recognize its parts
- Why does CogSci have so many different underlying disciplines?
- - sociological reasons - cogsci is the natural intersection of these disciplines! - each field is autonomous -methodical reasons: much more efficient
- Marr's 3 Levels
- Computational What is the problem being solved? Algorithmic What are the steps used to solve it? Implementation How are those steps implemented in underlying hardware (biological or otherwise)?
- Marr's 3 Levels: Reason
- Having all of these different disciplines is not an accident! Need multiple viewpoints in order to fully understand the mind
- 100 step rule
- - An example of constrants in cogsci (neurosci on psych) - essentially, brain can only carry out 100 steps of computation a second --> slow compared to computers. But many processes (ie. recognizing faces) far too complex to achieve in under a sec if brain worked like a computer (one algorithm at a time). Thus, brain must be wired up to do a bunch of different processes at once! :)
- Modularity of Mind: Basic Idea
- Mind is like a swiss army knife --> has specialized disassociable parts. In this way, mind is no more homogenous than body!
- Ubiquity of Specialization
- Molecular Level E.g. hemoglobin (oxygen transport) vs. rhodopsin (light transduction) Anatomical Level E.g. the heart (pumping blood) vs. the liver (detoxifying poisons) Behavioral Level Path integration in ants . . .
- Specialized systems spring from...
- ...specialized tasks! Makes sense since underlying problems are specialized too.
- Irresistibility (an example of of symmetric restriction on info flow)
- Ie. What fuels optical illusions! --> Doesn't matter what outside information is known, mind will still be fooled! Even for some higher level cognition: insurance scenario --> essentially are mental reflexes
- Informational Encapsulation (example of symmetric restriction on info flow)
- Processing internal to the module can't access external information. Ie. optical illusions --> essentially a mental reflex --> can even occur for high level cognition (ie. insurance cognition)
- Lack of access to interlevels (another example of symmetric restriction)
- Processing outside the module can’t access the intermediate states of processing inside the module — just its final output. Explains Instinct blindness!
- Symptoms resulting from symmetric restrictions on info flow
- Mandatory Processing • “I couldn’t help hearing what you were saying†• Understanding speech • Recognizing a face • Don’t think of a pink elephant! Why? Encapsulation from outside control! - also characteristic breakdown: aphasia, agnosia, autism --> if process is modular in a certain way, can lead to impairment
- informational encapsulation
- This is roughly the thesis that modules are much less open and permeable to background knowledge and beliefs on the part of the individual.
- Another resulting symptom from encapsulation
- Domain-Specificity • Only computes certain functions • “Specialized systems for specialized tasks†Not equal to encapsulation! but tend to go together
- Face Recognition!
- - instant and irresistible - when you recognize something as a face, mind assumes face is concave, not convex...even inside of mask looks like a face! - info encapsulation - no acess to interlevels - engage in it thousands of times a day, have no idea what's happening - fast, mandatory - domain-specific: highly specific to faces (no other processes these abilities would even be useful for, even for other body parts) Characteristic breakdown: prosopagnosia (people can't recognize faces) -> opposite exists too
- ‘Theory of Mind’
- The capacity to perceive, interpret, predict, and explain the behavior of others in terms of their underlying mental states. - Theory of mind appears to be an innate potential ability in humans, but one requiring social and other experience over many years to bring to fruition. -Having a theory of mind allows one to attribute thoughts, desires, and intentions to others, to predict or explain their actions, and to posit their intentions. As originally defined, it enables one to understand that mental states can be the cause of—and thus be used to explain and predict—others’ behavior.[6] Being able to attribute mental states to others and understanding them as causes of behavior implies, in part, that one must be able to conceive of the mind as a “generator of representationsâ€
- 'False belief' task
- Sally and Ann Sally has ball, goes to brush teeth Ann steals ball, puts it in box Sally comes back. Where does she look for ball? Box or basket? --> most kiddos say the box. Because its there. Duh. Conclusion: can't perceive that people have beliefs --> takes age to choose the basket, and have Theory of Mind come to fruition
- ToM and Autism
- - Children with autism have issues with automatically perceiving and interpreting the world in terms of mental states - Autistic: average to great with false photo task Normal: average to great with false beliefs Conclusions: Suggests that autism is a disorder that targets very particular modules -> eyes and face don't hold special cues for autistic
- Central Systems
- Domain-general systems for belief-fixation, which exploit information from input systems. Analogy to Science: Isotropy Everything a scientist knows is, in principle, relevant to determining what else she ought to believe. Fodor’s 1st Law of the Nonexistence of CogSci “The more global (i.e. isotropic) a cognitive process is, the less anybody understands it†Methodological encapsulation! modules: feed into central systems (not for any one thing - domain-general)
- Why MoM? (I)
- Speed! The Drawback: Ignoring context If you’re wondering whether the yellow-stripey thing you just glimpsed is a tiger, it matters whether you’re in New Haven, a yellow-stripey-chair store, or Budapest. The Tradeoff In many situations — e.g. trying to perceive a predator — a miss is far more deadly than a false hit! “A jack of all trades is master of noneâ€
- Why MoM? (II)
- Importance! “The trouble with being smart is that if you have to think before you blink when a guy sticks his finger at you, then your blink becomes “sicklied o’er with the pale cast of thought†and you end up with a finger in your eye.†(Fodor) - FAR too important to let them be domain-general! o.O
- Cognitive penetrability
- Can ideas and beliefs affect module? Only changing input (ie. closing eyes when presented with an optical illusion), not the automaticness of the module
- MoM Counterexample(s)
- High-functioning autistic individuals have learned to read faces --> but only shows that there are others ways to 'figure it out" --> doesn't mean that modularity doesn't exist
- What does it mean to implement a process formally? / Formal symbol systems
- -representation is not equal to what is represented - synatactic productivity --> combinations possible, underlying system is productive --> no. of sentences, numbers infinite - compositional semantics --> meaning of sentence is a function of words
- The 15 Game
- Basically tic tac toe! Have to make three numbers add up to 15 --> formally equivalent
- Ubiquity of Computation
- Mind is a computer Example: Tunisian Ant -after food search, knows where home is --> uses dead reckoning (have to keep track of how far they've moved, know direction, how far, how fast moving)--> comp. ultimately storing and operating
- So what? (What does the Tunisian desert ant tell us?)
- Implications for Cognitive Architecture Dead reckoning requires storing, preserving, & adding the values of a variable — not association Bees even communicate these values Dir of waggle = dir relative to sun # waggles = distance Methodological Moral Find cellular mechanisms by which nervous system stores & retrieves values of a variable, and performs elementary computational operations (e.g. +)
- Church/Turing Thesis
- Any finite mathematical or logical procedure — i.e. any computation — can be implemented on a Turing Machine (and also on just its tape!) • All modern PCs are (complicated) TMs • The difference? Architecture! • Note that the substrate doesn’t matter! (F-ism)
- Why is formality important?
- Curing instinct blindness E.g. Minsky anecdote Ubiquity in even simple cognition E.g. path integration in ants Cognition = a form of computation? Maintaining semantics w/o referring to it!
- Computational Thesis
- Organisms produce meaningful behavior by performing formal operations on symbolic structures that bear a representational relationship to the world.
- The Mind/Body Problem
- What sort of thing is a mind, and how is it related to the body? - contrast: explaining aspects of cognition - explaining how there can be cognition (quite a bit of philosophy involved)
- Dualism
- The essence of mind is independent of the physical body to which it may be temporarily attached (the belief of many people in the world, many religions, etc) - the mind is something weird, something else, something special, something we can't explain
- Arguments for dualism
- from religion - from introspection (warmth of summer doesn't feel like kinetic energy but that's what it is - -from irreducibility - from parapsychology (no actual evidence)
- Natural Selection!
- Sweeping in significance, simple in structure... - if within in a species there is variation, and one of these variations aids in survival, this trait will appear more and more in later generations. Thus, the species basic pool of hereditary traits changes
- Natural Selection: 5 Steps
- 1. Reproductive Math - Organisms have many offspring who also have many offspring. 2. Crunch Time - Population will eventually grow too large for available resources. So who wins? 3. It ain't a lottery - any advantages enjoyed by some of the contestants would count in their favor for survival 4. Inheritance - Any biasing advantages, no matter how small, would become amplified over time 5. The best idea of all time!
- Histology
- -1st year med school course - Take brain "apart", look under a microscope, look at it! Can understand structure and networks, but not how mind works
- Combinatorial expression (modern day histology variant)
- -Lights up the brain so each neuron looks different --> BRAINBOW!
- Neurons
- Underlying cells in brain
- Electrical stimulation
- organism twiches depending on which bit of brain is "poked", Or they feel a little sensation
- Neural specialization
- - Location - Contrast - Orientation - Color - Motion - Biological motion --> Serendipity in Science --> discovered individual cells for motion when there was a malfunction in slide projector with cat! --> "Holy shit, we're gonna get a nobel prize!" Hahaha!
- Single Cell Recording
- - Can measure the rate of neuron firing --> can hear the rate and study it - Very invasive
- Receptive Fields
- - neurons have very particular vision - as you get nearer to receptive field, neuron gets more and more excited - in severe cases of epilepsy (electrical storms in brain), the skull is removed to find out where it's coming from, and remove that bit...er...while that's going on: experiment time!
- EEGs and ERPs
- skulls not removed --> measure electrical waves on surface advantages: direct measurement, great temporal resolution disadvantages: need 100s of neurons, horrid spacial resolution Also, waves all interact, creating tons of interference! Because of this, EEGs are best used to measure just the timecourse of cognitive processing!
- Purpose of EEG and ERP
- EEG - reflect overall global brain activity ERP- designed to get around this, average many trials together to cancel out noise
- PET
- "Positron Emission Tomography" - inject a radioactive tracer into bloodsteam, accumulates in areas of brain that are used the most...as it decays, emits positrons, they hit electrons, gamma rays produced --> gamma rays can be detected, and scanners can reconstruct brain activity
- fMRI
- The big kahuna! -completely safe and noninvasive - safe for repeated and extended use - no special prep - great spatial resolution - signal strong enough to look at individuals -anatomical/functional scans in same session However: bad temporal resolution
- Subtraction Method
- A method used to get the measure for the thing you care about. -In this method, a person's response time for executing a simple task is determined; then the task is somehow complicated, and a new response time is measured. The difference between the two times is taken as the time necessary to deal with the complication. The method was abandoned in Wundt's lab because increasingly complicated tasks yielded unreliable time differences, sometimes even producing negative estimates for the time to execute some components of a task. Today the subtraction method is used with care in situations where it can reasonably be assumed that the addition of some task leaves all other components of the overall task unchanged. ie. time to press key w/ any stimulus v. time to press specific color key
- The FFA
- - show faces --> localize - faces v scrambled faces - human faces v animal faces
- fMRI Limitations
- - problems with methodological logic -can't look at how cognitive processes occur - fMRI can't measure overall activity - noise, stillness, claustrophobia - sociology - reliability worries (if you do a single study)
- fMRI and the Public
- People, in their ignorance, buy into any BS published in a magazine --> "Hey, we can see your dreams on a tv"
- Importance of Cognitive Architecture
- Understanding the scope of cognition The great flexibility of cognition requires us to think of CogArch not as determining specific thoughts & behaviors, but as a set of mechanisms that make possible a certain range of activities
- Comparing Architecture
- Weak Equivalence Same input/output relations. Doing the same thing, but in a different way. (Similarity = on the computational level) Strong Equivalence Same intermediate steps. Doing the same thing, in the same way. (Similarity = on the algorithmic level)
- Heritability
- The proportion of phenotypic variance attributable to genetic variance
- Why Study Infancy?
-
Looking at development itself
How and when does change occur
A tool for studying cognition
Can study certain systems in their 'pure' form, before contamination by other processes in adults
Exploring nature/nurture
To explore the initial state, get as close as you can. - Problems with methodological logic (problems of fMRI)
-
still don't know how processes occur,
conceptual problems, indirect measure (vs. EEGs), can't measure overall activity, just % change, few studies go beyond localization - 6 steps for fMRI
-
1.spin (protons); 2. magnetic field; 3. RF Pulses (tacoma Narrows bridge - resonance); 4. Blood Flow; 5. Blood & Magnetism (hemoglobin); 6. Measurement
;; protons have spin --> magnet fired --> spinning protons align themselves --> core magnetic field alligns, and then protons recess --> now, activity can be measured! Resonance is powerful! Also, blood flow is measured! --> without oxygen to shield it, hemoglobin becomes paramagnetic --> then it becomes diamagnetic...in any case, there's a change in how susceptible the blood in the brain, and that's what we measure MLSLS: More, Less, Slower, Longer, Stronger - Violation of Expectation (3 steps)
- 1. Familiarization; 2. Test; 3. Strategy; one example was the SUPPORT experiment
- Why have innateness?
- 1. Some things may not be learnable; 2. some things are too important to learn (Scottish cattle running off cliffs example); 3. importance of uniformity (language)
- Poverty of the Stimulus argument (Why we believe in innateness)
- is it possible for input to give rise to output; language acquisition, things come into ears but stimulus is too poor to judge results
- caveats to heritability
- 1. environment changes; 2. what exactly is the definition of environment?; 3. no specific genes for traits, as we know it (body size could affect IQ for all we know)
- Testing equivalence
- 1. Error Analysis (do systems make same mistakes in the same situation?) 2. Complexity Analysis (Do systems take equally long to process same inputs? ex. 9x13)
- formal systems
- tokens, start state, allowable moves, goal
- dead reckoning
- mind must be a formal system (ant example)
- piagetian methodology problems
- 1. multiple simultaneous development; 2. competence vs. performance (insensitivity of test)
- Thesis of Functionalism
- pattern matters, not the pieces