Cognitive Science Midterm 1
Gold's Theorem 1967
standard (simplified) interpretation: successful language learning requiress either 1) negative evidence or 2) innate structure constraining the set of possible languages or (universal grammar) --But we know that negative evidence is not used in children so it must be innate knowledge
Language acquisition device (universal grammar)
suggests a simple unified explanation of 3 important things 1) why language is uniquely human 2) why language is learnable 3) why there are language universals
Language is infinite
you can always create new sentences. It relies on mental representations (to give words meanings) which is trouble for behaviorism.
Georg Cantor (1845-1918)
"there is more than 1 size of infinity" (There is countably infinte and uncountably infinite) --Developed The Diagonal Method: to show that there is more than one size of infinity. This is directly related to Leibniz's dream. "You can always create a package unlike all the others" Sketch of Proof = Assume only countably infinite subsets of N. Use diagonal method to create new subset of N, beyond the countable ones. Existence of new subset means our assumption was false. Thus, "uncountably infinite". -- Go down the diagonal and flip all positives to negatives and all negatives to positives.
Classical Conditioning
(e.g. Pavlov- cats do what they want to) Passive: associate stimuli that are paired in your experience, but are not directly under your control. --a learning process in which an innate response to a conditioned stimulus comes to be elicited in response to a previously neutral stimulus; this is achieved by repeated pairings of the neutral stimulus with the conditioned stimulus.
Herman Ebbinghaus
(objective measures) measured memory retention: do you or do you not remember a given experience? Exponential forgetting curve. --Spacing Effect: a list of items is better remembered when studied at space intervals, rather than all at once.
Alison Gopnik
- Children are a source of answers to classical philosophical questions - Plato to piaget to probabilistic models (the idea of the child as the scientist) - The reset button: causal learning as an example of understanding the unseen. - Humans have an especially long period of immaturity: cultural learning; exploration.
Mahesh Srinivasan guest lecture summary:
- Flexible language. Reflects a fit between language and thinking (kids quickly generalize such uses to new words, suggesting a good cognitive fit. - Cross language similarities, suggesting a universal cognitive basis, but also some differences: source verbs. - "verbing weirds language"
Alan Turing: the world's first and most influential computational scientist
-- appears to have had a poetic streak, not just a technical one, was involved in a British code-breaking effort during World War II, and had interests in a wide range of topics, with mathematics as the apparent unifying strand. -- died by suicide from cyanide poisoning at age 42. Half eaten apple by bedside (apple logo) --Interested in Fibonacci numbers (Fibonacci numbers are a sequence of numbers in which numbers later in the sequence are defined in terms of numbers earlier in the sequence). --The number of clockwise and counterclockwise spirals on the head of a sunflower are generally two consecutive numbers in the Fibonacci sequence
John Searle's Chinese Room Argument
-- is the same as the imitation game but in Chinese --The Chinese room argument can be seen as an ironic inversion of Turing's original idea: whereas the Turing test began with the imitation game and replaced a human in that game by a machine, the Chinese room argument began with the Turing test and replaced part of a machine in that test by a human -- it is a critique of the Turing Test -- The original Chinese room argument is based on hypothetical transmission of written linguistic communications. Chinese Room Argument: Imagine a physical symbol system passing the Turing test, I'll show it doesn't understand (turing test is an inadequate criterion) -- challenges the claim that it's possible for a computer running a program to have a "mind" and "consciousness" in the same sense that people do, simply by virtue of running the right program (refutes strong AI)
The Turing test
--is a proposed criterion for determining whether a machine can think, assessing their ability to use language - rather than some other human ability. -- was inspired by a game (the imitation game) that requires someone to guess whether they are interacting with a man or a woman. --was held by Turing to have the advantage that it draws a fairly sharp line between the physical and the intellectual capacities of a human being --Eliza, SHRDLU, Watson = all attempts to pass the turing test --
2 Sobering Notes:
1) Bertrand Russel's letter to Frege (1902): "I have permitted myself to express deep respect to you... but I have encountered a difficulty". Frege used sets of sets which opens a possibility that a set can be a member of itself. So Frege's system is flawed 2) Leibniz wanted a language that could also be an efficient instrument of calculation - but in Frege's logic most deductions are very complicated and time consuming (you might not be able to reach a conclusion simply because you are lazy)
David Marr's 3 levels of Analysis:
1) Computational Theory (GOAL): ie) to fly 2) Representation & Algorithm (WHAT ARE THE REPRESENTATIONS AND PROCESSES): ie) curved wings, aerodynamics 3) Hardware Implementation: (HOW CAN IT BE IMPLEMENTED): ie) feathers steel, balsa wood. MARR: "Trying to understand perception by studying only neurons is like trying to understand bird flight by studying only feathers. It just cannot be done" -- the functional goal of a given cognitive task is specified at the the computational level
3 Main Questions of Cognitive Science
1) Where does knowledge come from 2) What is the nature of thought? 3) Are there uniquely human aspects of cognition?
Infinity
A rationalist challenge to the empiricist view that all knowledge comes from experience. Our experience is finite, yet we can conceive of the infinite. --Thus we must go beyond the data given. -- Infinity comes in more than one size
Syllogisms (Aristotle)
A syllogism is a kind of logical argument that applies deductive reasoning to arrive at a conclusion based on two or more propositions that are asserted or assumed to be true. -- Syllogisms are a specific form of deductive inference -- Aristotle's project; to catalogue the valid syllogisms (how the space of syllogisms in which Aristotle catalogued was of a very specific form) -- no valid syllogisms have 2 negative premises --Aristotle systematically enumerated the syllogisms and identified which ones are valid
What was Leibniz's dream
A universal scientific language that perfectly represented our thoughts
Operant Conditioning:
Active: associate your actions with resulting reward or punishment. Cats in little puzzle boxes.
Leibniz and the "wonderful idea" (1646-1716)
Classic rationalist (veins in the marble). Fundamentally optimistic. Believes world has be planned by a benevolent God (and that we live in the best possible world) The "Wonderful Idea": "people are largely benevolent and cooperative but are hampered by language which is an imperfect mirror of intelligible thoughts and often makes reasoning obscure". What we need is an ideal language that will "perfectly represent the relationships between us and our thoughts" --Argues 3 main things: 1) People are fundamentally well-meaning 2) Human discord results from imperfect communication through language 3) An ideal language would solve many human problems
According to Leibniz: It requires a large but finite number of instances confirming a general truth to establish the universal necessity of that same truth
False
True or False: Locke's system is aligned more with Plato's, whereas Leibniz's is aligned more with Aristotle's
False
4 lobes of the cerebral cortex
Frontal: motor control, speech, planning parietal: spacial, sensory integration Temporal: objects, auditory Occipital: vision (separation between frontal and parietal = central sulcus separation between Frontal & Temporal = lateral sulcus)
David's Marr's model of vision
Image >>> Primal sketch >>> 2.5 D sketch >>> 3D sketch (continuum from imagistic to symbolic) -- hierarchical structure in perception -- involved imagistic involved imagistic representations, hierarchically structured representations, among others -So far, this tells you what representations and processes the visual system uses (on this model) it doesn't tell you what functional goal the system has or how it is implemented.
Leibniz's Veins in the Marble
It suggests that ideas may be innate in us as natural predispositions, but some labor (experience) may still be required to allow those predispositions to be realized.
What's difficult about cognitive science?
It's inherently recursive and it's deals with cognition and the mind, things that you cannot physically study (it's abstract)
John Locke
John Locke = empiricist --Neither principles nor ideas are innate. Believes in 2 types of experiences: 1) Sensation (Yellow, white, heat, cold, soft, hard, bitter, sweet) 2) Reflection (Perception, thinking, doubting, believing, reasoning, knowing)
George Miller (1920-2012)
Kept seeing the number 7 --Studies suggest that people can hold about seven "chunks" of info in memory. A constraint on human mental representation. Opens the possibility of hierarchical embedding of chunks inside one another. --Not only there mental representations but they have structure. Some thought is hierarchically structured.
The 2 Hemispheres of the brain
LEFT & RIGHT -- There are some cerebral asymmetries but not as dramatic as the stories tell -- one of the main asymmetries is in language: right handed people have more areas in their left hemisphere associated with language --The corpus callosum connects the two cerebral hemispheres.
Herb Simon and Allen Newell
The first AI System-- found proofs of mathematical facts, expressed in terms of inference rules. Used heuristics inspired by human problem-solving strategies.
In Marr's terms, neuroscience investigates the implementational level of information processing
TRUE
When a driver tries to determine whether they can fit their car around a tight corner and into a tight parking spot, a natural way to do this is through mental transformation of a mental image, echoing Shepard and Metzler's (1971) experiments.
TRUE
Multiple Realizability
The idea of multiple realizability is the notion that a given functional system may be realized by multiple different physical structures. It has been used to argue that studying the brain could in principle highlight inessential rather than than essential properties of the mind In short, a theory of mind that includes multiple realizability allows for the existence of strong AI.
David Marr
The marr prize named in honor of the late David Marr, is given annually to the best student paper at the Cognitive Science Society Conference.
According to Leibniz
The notation will do the work, by analogy with Arabic numerals. "Let us calculate who is right...." (Leibniz has led to artificial intelligence)
A model of a formula is....
The set of possible worlds that makes the formula true
Alexander Newell & Herbert Simon 1975
The Physical Symbol System Hypothesis: 1) Symbols are physical patterns. 2) Symbols can be combined to form complex symbol structures 3) The physical symbol system contains processes for manipulating complex symbol structures 4) The processes for generating & transforming complex symbol structures can themselves be represented by symbols and symbol structures --Syntax: the identification and manipulation of such symbols based purely on their shape --Semantics: the meaning and changes in meaning that these syntactic manipulations are meant to correspond to -- The PSSH holds that a physical symbol system has the necessary AND sufficient means for general intelligent action (ie) problem solving) --The PSSH supports the view that computation over symbolic representations is an appropriate means for modeling the mind, and that intelligence can be viewed as the capacity for search through a symbol structure
Checkerboard Illusion
The checkerboard illusion illustrates that perception is shaped by expectations and context
The Diagonal Method
The diagonal method is standardly explained in terms of labeled packages that contain things of the same type as the labels The diagonal method can be used to show that there are more sets of integers than integers. Used by Turing to show the attainability of Leibniz's dream of yielding valid inferences
Behaviorism isn't sufficient
Tolman shows that even non-humans behave as if they have mental representations
"The logic theory machine" is a paper that reported a complete proof of a theorem that had been carried out on a computing machine.
True
It is sometimes claimed that one of the distinguishing features of cognitive science is that it is rooted in classical philosophical problems.
True
True or false: The passage of St. Paul in which Leibniz states that the law of God is written in our hearts can be read as an endorsement of the doctrine of innate knowledge.
True
Alan Turing (1912-1954)
Turing used the diagonal method to show that no such procedure exists. --Turing machine: formal model of computation (preforms specific tasks). Turing machines are abstract objects, not physical. --Universal Turing Machine: A model of general purpose computation, like today's laptops and computers. --making possible the view of mind as a symbolic system, a processor of symbolically expressed information. The mind and the computer: different hardware, same software.
What happens to visual information once is reaches visual cortex in the occipital lobe?
Two Cortical Visual Systems (Ungerleider & Mishkin) -- "where = dorsal stream -- "what" = ventral stream
Helmholtz
advanced the idea of unconscious inference
Modes ponens, modus tollens
an inference rule modus ponens = the rule of logic stating that if a conditional statement ("if p then q ") is accepted, and the antecedent ( p ) holds, then the consequent ( q ) may be inferred.
The Dinner Conversation 1934
as recalled by skinner: professor whitehead was equally earnest, he agreed that science might be successful in accounting for human behavior provided one made an exception= VERBAL BEHAVIOR (aka Language). By understanding language we understand something important about ourselves.
The Halting Problem
brought up by Turing (deals with the Computational-Representational Understanding of the Mind) --The Halting Problem is the problem of determining, from a description of an arbitrary computer program and an input, whether the program will finish running or continue to run forever.
Edward C. Tolman
challenges behaviorism. Insight in rats: a form of learning is not driven only by reinforcement (latent learning) -- Cognitive maps & mental representations: Rats were not merely associating actions with rewards-- they had a "cognitive map" of the maze or a mental representation of the world. They are central to cognitive science. It is a key feature that distinguishes it from behaviorism which sought to disperse with "the mental" altogether"
Mechanical Turk von Kempelen, 1770 (Actually hid a human chess player) & Digesting duck Jacques de Vaucanson, 1739 (also fake)
clockwork of enlightenment is reflected in the concept of "mechanical man"
Joseph Weizenbaum (1923-2008)
developed ELIZA, the original chatterbot: applied pattern matching rules to human statements and replied accordingly. It had no representation of the meaning of statements (and not much about the form). ELIZA usually just spits out appropriate responses as output but even though its symbolic responses it's shallow (it's input is examined, keywords are matched, and usually an appropriate response is given)
Terry Winograd
developed a natural language understanding computer program called SHRDLU (in contrast to ELIZA). --SHRDLU: an early AI program that linked language understanding to action in an artificial micro-world of blocks. -- The name comes from a linotype machine for newspaper publishing where the keys are ordered by English frequency. --What SHRDLU does: 3 interacting components. 1) Syntactic Analysis: grammatical analysis of input sentences. 2) Semantic analysis: determines what the sentence means. 3) Perception & Interference: consults the "world" for answers to questions posed. -- All implemented as procedures --Links language understanding symbolically represented with observable action in the world. Language understanding is dependent on multiple interacting processes
John Watson (1878-1958)
focused on the analysis of behavior, rather than mental states. Emphasized the role of the environment in explaining behavior. --He saw no dividing line between man and brute. BEHAVIORISM: "psychology as behaviorists views it is a purely objective natural science. It's theoretical goal is the prediction and control of behavior. Introspection forms no essential part of its method" according to John Watson. --Behaviorists are essentially empiricists.
Getting oriented to the brain
forebrain midbrain: primitive sensory/motor hindbrain: responsible for balance, posture, some motor control (cerebellum, pons, medulla) --In general, the hindbrain is evolutionarily older than the forebrain.
2 year old girl
had a cyst in right hemisphere (basically only had left hemisphere) but was still able to function pretty normally. Shows neural plasticity of the brain
Lashley
hierarchical structure of plans
Metaphors of the mind
historically reflected the technology of the age --have at some point in the past been based on water technology --have historically been based on information-recording technology such as writing instruments --come and go, but can be useful as a basis for concrete hypotheses and theories.
Language
is uniquely human. Language is rule-driven. (it is a classic challenge to behaviorism). It is hierarchically structured - not just a string of words. It has a infinite generative capacity from finite means.
Noam Chomsky (1928- present)
language seems to be yet another domain that is governed by hierarchically structured mental representations-- not just a string of words. Poverty of the Stimulus: same as Plato's problem only Noam Chomsky applied it to linguistics in particular (children have innate linguistic capabilities)
Hodgkin & Huxley (1952)
model of neuron action potential propagation drew on "telegraphers equation" (ie) the transatlantic undersea cable
Language universals
patterns that are found in all or many language, including unrelated ones (suggests an innate bias). Word order: if a language has verb-final word order, it will tend to have postpositions rather than prepositions
Universal Grammar
Boils down to 2 claims: 1) Language & language learning are constrained: some logically possible hypothesis are never considered. (ACCEPTED) 2) The Constraints are specifically linguistic in nature-rather than stemming from more general cognitive forces (CONTESTED) --the idea of an innate endowment that helps you learn language and specifies what real human languages can and can not look like.
Boole held that the mind was unlike other objects of scientific inquiry. On what basis did he make this claim?
Boole based this claim on his view that the general truth is seen in the particular instance. --Boole was a dualist, focusing on deduction not induction
Humors
4 substances that balance each other out in a healthy person's body: -- Black bile: despondent -- yellow bile: bad tempered -- blood: amorous -- phlegm: calm
Turing Machines
A formal model of computation that preforms specific tasks. It is an abstract machine M that has... 1) An internal state, one with finite member of possible states Q 2) An infinite tape on which symbols can be read and written 3) A finite set of rules that specify what M should do as a function of what it sees on the top.
Computational-Representational Understanding of the Mind (CRUM)
A hypothesis which proposes that thinking is performed by computations operating on representations. This hypothesis assumes that the mind has mental representations analogous to data structures and computational procedures analogous to algorithms, such that computer programs using algorithms applied to data structures can model the mind and its processes. (Artificial Intelligence) "Information processing" The computational-representational understanding of mind holds that cognition is information processing, which can be formalized in terms of symbolic representations.
Birth of Cognitive Science
Birth of cognitive science at 1956 MIT symposium.
The Little Albert Experiement
An attempt to see if conditioning worked with humans. Watson tried to form a phobia in an 11 month old infant. Classical conditioning: pairing a white rat with a loud (aversive) noise. Tested whether classical conditioning occurs in humans.
Aristotle and the Blank Slate
Aristotle = empiricist (384-322 BC) Blank Slate (tabula rasa) = individuals are born without built-in knowledge and therefore all knowledge comes from experience or perception.
Rene Descartes
Asks "what can we know with certainty?" His skepticism: "I can doubt the existence of my body and my senses" but "I cannot doubt the existence of my mind" (by thinking I establish its existence. "I am a thing that thinks". Therefore... DUALISM: the idea that mind and body (or more generally, the world) are different kinds of entities. You are a dualist if you believe that mental states cannot be reduced to neural states. And you think that the physical sciences cannot explain the operation of the mind MATERIALISM: You are a materialist if you think we can explain human minds using sciences like physics and biology. If you think that any kind of mind can be explained in physical terms.
Gottlob Frege (1848-1925)
Begriffsschrift (Concept Script) --instead of propositions we use predicates to denote properties of objects (P(a) where p = predicate and a = object). --Truth formulas: the truth of a formula depends on the truth of its parts (each formula has a truth formula). Assignments of truth values to a proposition is a possible world -- Frege did not provide a procedure that will tell you whether or not a conclusion C follows from premises P.
Difference between Behaviorism and Cognitive science
Behaviorism = environment >>>>>>>>> behavior Cognitive Science = Environment>> representation (processing) >>> behavior
Dualism and connection to computational view of the mind
Contrast to dualism is MONISM: which is the belief that the mind and body are the same kind of thing. But mores specifically, the contrast to dualism is MATERIALISM
"I wish that you would consider all of these following altogether naturally in this machine from the disposition of its organs alone, either more or less than do the movements of a clock or other automation from that of its counterweight and wheels.."
Decartes (1664)
David Hilbert (1862-1943)
Decision problem: want a procedure that accepts premises P and conclusion C and determines whether C follows from P. The problem of finding a way to decide whether a formula or class of formulas is true or provable within a given system of axioms. --entscheidungsproblem is german for "decision problem" --such procedure would fulfill Liebniz's dream
Induction vs Deduction
Induction = characterized by the inference of general laws from particular instances (inductive conclusions may not necessarily be true though (generalizing beyond the data given, guided by biases, conclusions could always be wrong) Deduction = goes from known truths to derive new truths (characterized by the inference of particular instances from general laws)
Contralateral Organization
Information from left visual field is processed in the right hemisphere and vice versa --Each of the two eyes takes in visual information from both the left and right visual fields.
Phineas Gage & the Horrible Accident
Iron entered side of his face shattering the upper jaw and passing back of the left eye and out of the top of his head, -- Born = 1823 --Accident = 1848 --Died = 1860 -- Doctor's report: "He is capricious vacillating, devising many pans of future operations, which are no sooner arranged then they are abandoned" (AKA accident ruined ability to plan) --TAKE HOME MESSAGE: frontal lobe = connected with planning (executive functions) and accidents can yield to knowledge (though unsystematic and uncertain) concerning the functional organization of the brain
Latent Learning
Is a name for a kind of learning that is held to occur without reinforcement. Was explored by Edward C. Tolman
Leibniz's metaphor of veins in the marble
Is the same basic idea as the tabula rasa or blank slate, but with different technology. --It suggests that ideas may be innate in us as natural predispositions, but some labor (experience) may still be required to allow those predispositions to be realized --It is meant to support Locke's stance on the notion of innate ideas.
Leibniz & his view on Locke
Lock is just wrong because experience is necessary but not sufficient to account for our knowledge (example: how we know the angles in a triangle add to 180 degrees). ANSWER: "All the instances confirming general truth... are not sufficient to establish the universal necessity of the same truth. Thus our minds must contain 'seeds of eternity'" --Beasts are purely empirical --Veins in the marble -- All ideas come from sensation or reflection
Caveat of Marr's 3 levels of Analysis:
Marr's framework suggests we can usefully investigate the different levels largely independently of each other (ie) we can understand the algorithm without understanding the implementation) -- this is an assumption which could lead us astray -- tension between top down and bottom up approaches
Alfred North Whitehead (1861-1947)
Mentor of Russel. Co-authored Principia Mathematica with Russel which was built on Frege which sought to ground all mathematic truths in logic
Symbol Grounding Problem
Need to know when our symbolic representations are true of the world. --related to the problem of how words (symbols) get their meanings, and hence to the problem of what meaning itself really is.
Behaviorism isn't necessary
Newell & Simon, Miller, and Chomsky show that rigorous science of mind is possible without the extreme assumptions of behaviorism
Inference Rules
Operations that depend only on syntax, but have semantic implications
Plato and Innate Knowledge
Plato = rationalist (& teacher of Aristotle) -- We know things that we would not have known otherwise without innate knowledge. -- plato's problem = how we know so much, given so little evidence -- Poverty of the stimulus: children have innate linguistic capacity
How do Plato's problem and the "poverty of stimulus" related to one another
Plato's problem was "How do we know so much given so little" and the "poverty of stimulus" is this same question only it was term coined by Noam Chomsky many years later, specifically in regards to language.
George Boole (1816-1854)
Propositional logic, classes or sets, xy= x and y. xx= x is the foundational rule of Boole's system: in ordinary algebra it is true only for x=0 and x=1. Can cast Aristotle's "Principle of Contradiction" (nothing can both belong to and not belong to a class x) -- "All plants are alive": P = PL. Can verify validity of Aristotelian syllogisms using Boole's logic.
B.F. Skinner (1904-1990)
Radical Behaviorism: theories aim to explain behavior in terms of environment. "there is no evidence that we have acquired knowledge" according to Skinner. Sought to dispense with talk of mental states entirely. --Argued that if we can describe what happened at a football game we watched, that is not evidence that we have actually acquired knowledge of what happened - it shows only that our behavior has changed. --An inversion to Cartesian skepticism: privilege the senses, not the mind -- Skinner Box (devices for studying learning and reinforcement)
Rationalist vs Empiricist
Rationalist = knowledge comes partly from reason, some of which may be innate Empiricist = all knowledge comes from experience
Skinner's Response published in the book Verbal Behavior in 1957, 23 years after dinner conversation with whitehead
Skinner = classic empiricist Chomsky = classic rationalist Whitehead = Rationalist
2 extremes
Solipsism: is the view that no mind exists except for your own (A consideration of solipsism can be found in the writings of Descartes) -- The philosophical appeal of solipsism is that it allows you to impute mental states to only those entities for which you have direct first-person evidence of those mental states Anthropomorphism: the attribution of human abilities to anything
Introspection
Was an aspect of earlier psychology that behaviorism sought to avoid. -- Wundt and others at the time sometimes relied on people's reports of their experience. Not everyone relied on introspection.. for instance Ebbinghaus Introspection is the examination of one's own conscious thought and feeling. In psychology the process of introspection relies exclusively on observation of one's mental state, while in a spiritual context it may refer to the examination of one's soul.
The notion of a completed or actual infinity
Was held by Leibniz to be frequently used by Nature
Plato's Meno. What does "like the sting ray we have made him numb" mean?
We have cause someone who falsely thought he knew something to realize that he in fact did not know it (John Locke does not agree, he agrees with the Blank Slate) -- Plato's Meno.... learning is remembering, some aspects of human knowledge are innate, Someone who does not know a particular thing may nonetheless have true opinions about that thing
Father of experimental psychology: William Wundt (1832-1920)
When asked to press a button when people hear a tone they respond late; but when asked to focus on the stimulus (the tone) rather than the response (the button press) people responded more accurately. Wundt's interpretation: APPERCEPTION: (subjective introspection) the process of making an experience clear in consciousness
Shepard & Metzler (1971) experiments: What evidence is there for imagistic or experiential representations?
You are given 3 sets of cube-like figures. Is it 6 different ones or 3 copies. FINDINGS = time to confirm "same figure" is a linear function of the angle of rotation- whether in picture plane or in depth INTERPRETATION: People mentally rotate one figure onto the other. Imagistic mental representations.
