Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
115 Cards in this Set
- Front
- Back
- 3rd side (hint)
|
What is cognitive science
|
the interdisciplinary studyof mind and intelligence, embracing philosophy, psychology, artificial intelligence, neuroscience,
linguistics, and anthropology. |
|
|
|
Cog Sci History
|
Ancient Greek Philosophers- study of the mind- metaphors, tabula rasa, knowledge comes from innately known concepts
Descartes- Dualism- mind and body separate. 1860's Wundt- laboratory methods for studying mental operations systematically. Behaviorism (Pavlov) through 1950s. 1956- Miller short term memorory limits and chunking. AI- McCarthy, Minsky, Newell, Simon Chomsky- mental grammars. 1980s connectionist theories. |
|
|
|
Rationalism
|
Knowledge only gained by thinking and reasoning. Held by Plato, Leibniz, Descartes,
|
|
|
|
Empiricism
|
Knowledge in terms of rules. Aristotle, Locke, Hume,
|
|
|
|
Kant
|
Attempted to combine Empiricism and Rationalism by arguing that human knowledge depends on both sense experience adn the innate capacities of the mind.
|
|
|
|
mental representation
|
a structure or process in the mind that stands for something. Knowledge in the mind consists of mental representations. Non-mental representations are like words on a page.
|
|
|
|
mental procedures
|
processes that happen on mental representations. multiplication for example.
|
|
|
|
Wilhelm Wundt
|
laboratory methods for studying mental operations.
|
|
|
|
behaviorism
|
denies existence of the mind- Pavlov, Watson, Skinner- belief that psychology should be restricted to the study of relationship between observable stimuli and observable responses. Talk of consciousness and mental representations was banished.
|
|
|
|
George Miller
|
Cog Sci founder- 1956- human capacity for short-term memory is limited to 7 items, more can be kept by breaking into chunks.
|
|
|
|
John McCarthy
|
Cog Sci founder- leader of AI in formal logic
|
|
|
|
Marvin Minsky
|
Cog Sci founder- concept like frames are central form of knowledge representations.
|
|
|
|
Allen Newell
|
Cog Sci founder- proofs in formal logic, power of rules in accounting for aspects of human intelligence.
|
|
|
|
Herbert Simon
|
Cog Sci founder- proofs in formal logic, power of rules in accounting for aspects of human intelligence.
|
|
|
|
Noam Chomsky
|
Cog Sci founder- rejected behaviorist assumptions about language as a learned habit and proposed instead peoples ability to understand language in terms of mental grammars consisting of rules.
|
|
|
|
Case-based reasoning
|
analogical thinking
|
|
|
|
1980s developments
|
connectionist theories of mental representation and processing based loosely on neural networks.
|
|
|
|
methods
|
1. Experimentation with human participants, computational modeling,
|
|
|
|
Central hypothesis of Cog Sci
|
Thinking can best be understood in terms of representational structures in the mind and computational procedures that operate on those structures.
|
|
|
|
CRUM
|
Computational-Representatioal Understanding of the Mind
|
|
|
|
CRUM analogy
|
Computer program: data structures + algorithms= running programs; Mind: mental representations+computational procedures = thinking.
|
|
|
|
4 elements in cog sci models
|
theory (postulate), model (design test), program (perform test), platform
|
|
|
|
criterion for evaluating theories of mental representation.
|
1 representational power 2 computational power, (problem solving; i planning, ii decision making, iii explanation 3 psychological plausibility, 4 neurological plausibility, 5 practical applicability
|
|
|
|
Wilhelm Wundt
|
1870's laboratory methods for studying mental operations systematically- from philosophy to psychology.
|
|
|
|
Christopher Longuet-Higgins
|
coined term Cognitive Science in 1973- Edinburgh- Dept of Machine intelligence perception.
|
|
|
|
Lighthill report
|
AI winter- scalability problem- stopped funding for AI work in UK for years.
|
|
|
|
intersection vs. union
|
disciplines working together to validate results, but not merging into one.
|
|
|
|
3 circles of cog sci
|
–Brain-the understanding of neurobiologicalprocesses and phenomena. –Behavior-the experimental methods and findings from the study of psychology, language, and the socioculturalenvironment –Computation-the powers and limits of various representations, coupled with studies of computational mechanisms
|
|
|
|
interdisciplinary problems
|
breadth vs. depth, importance in having a strong foundation in a core discipline.
|
|
|
|
CS requires hypothesis
|
• suggested explanation of a phenomenon • reasoned proposal suggesting a possible correlation between multiple phenomena
|
|
|
|
mind
|
refers to the collective aspects of intellect and consciousness which are manifest in some combination of thought, perception, emotion, will and imagination. –My definition: Ethereal entity that accounts for all phenomena that we can not empirically explain (yet!).
|
|
|
|
brain
|
control center of the central nervous system –Extremely complex structure, with more than ~100 billion neurons each of which is connected to ~10,000 others.
|
|
|
|
computation
|
refers to information processing –from simple calculations to human thinking
|
|
|
|
Universal Turing machine:
|
symbol-manipulating device capable of simulating the logic of any computer
|
|
|
|
Von-Neumann architecture
|
implementation of Turning machine. Input, cpu (instruction adddress register, instruction unit, arithmatic unit, accumulator) memory, output.
|
|
|
|
representation
|
about how people store and process information- knowledge representation
|
|
|
|
internal symbol in cog. Psychology
|
refers to external reality
|
|
|
|
80s revolution
|
Rodney Brooks proposed intelligence without representation.
|
|
|
|
CRUM hypothesis
|
thinking is performed by computationsoperating on representations
|
|
|
|
CRUM shortfallings, criticisms
|
the mind not well explained by representations and computations.
|
|
|
|
Cognitive theory
|
Set of representational structures and set of process that operate in these structures
|
|
|
|
Computational Model
|
Interprets structures and processes by analogy with computer programs
|
|
|
|
3 stages in process
|
discovery, modification, evaluation
|
|
|
|
Criteria- representational power
|
How much information is expressed by this representation?e.g. University guidelines on enrolment
|
|
|
|
Criteria- computational power
|
Problem solving: how do we accomplish our goals? * Planning how(find a successful sequence of actions) * Decision which(chose the best plan) * Explanation why(did it happened)?-Learning: generalization(through experience)
|
|
|
|
Criteria- psychological plausability
|
Since we aim to understand human cognition, any theory must address how people think (as opposed to machine learning). Use of quantitative measures in human subjects
|
|
|
|
What is Cog Sci?
|
the interdisciplinary studyof mind andintelligence, embracing philosophy, psychology, artificial intelligence, neuroscience, linguistics, and anthropology.
|
|
|
|
syllogism
|
Aristotle logic- two premises & a conclusion. All students are overworked, Mary is a student; therefore mary is overworked. Can analyze form rather than content.
|
|
|
|
deductive inference
|
conclusion follows necessarily from premise- if premise is true, conclusion is true. Syllogism is deductive.
|
|
|
|
inductive inference
|
reasoning that introduces uncertainty. For example, premise of syllogism is taken from a sampling rather than whole population (all the students I know are overworked- does not conclude that Mary is overworked.)
|
|
|
|
Gottlob Frege
|
1879- began modern logic (post aristotle) with much more general logic than aristotles
|
|
|
|
p ➔ q
|
Prepositional logic- if p, then q. If paula is in the library, then quincy is in the library.
|
|
|
|
p v q
|
Prepositional logic-p or q.
|
|
|
|
(p v q) ➔ ~d
|
Prepositional logic-If paula or quincy are in the library, then debra is not.
|
|
|
|
S(p)
|
Predicate calculus- Paula is a student ( item(s) in parenthesis is subset or member of capitol letter)
|
|
|
|
(for-all x) (student (x) ➔ overworked (x))
|
Predicate calculus- all students are overworked. (for any x, if x is a student, then x is overworked.
|
|
|
|
(for-all x) (for-all y)[(student (x) & course (y) & take (x,y)) ➔ get-credit for (x,y)]
|
Predicate calculus- students who take courses get credit for them. (For any x and y, if x is a student, y is a course and x takes y, x will get credit for y.
|
|
|
|
First order logic (FOL)
|
aka Predicate calculus: system of deduction extending propositional logic by the ability to express relations between individuals (e.g. people, numbers, and "things" more generally).
|
|
|
|
modus ponens
|
p ➔ q; p, therefore, q.
|
|
|
|
modus tollens
|
p ➔ q; not q, therefore not p
|
|
|
|
monotonic deductive planning
|
problem with deductive planning: can only infer new deductions, cannot learn from and reject previous ones. Doesn't allow you to discard previous beliefs found to be wrong. When programmed to make nonmonotonic, it becomes processor expensive. Also, cannot learn from experience- it goes through the same process over and over again, when presented with the same problem.
|
|
|
|
deduction
|
knowing all, and given a member
|
|
|
|
induction
|
knowing some and inducing based on experience.
|
|
|
|
abduction
|
inference where you form a hypothesis in order to generate an explanation.
|
|
|
|
Criteria for Evaluating Approaches
to Mental Representations |
slide
|
|
|
|
induction
|
2 senses. 1. Any inference that introduces uncertainty. 2. Only inductive generalization which general conclusions are reached from particular examples.
|
|
|
|
pragmatic reasoning schemas
|
how people use logic to solve problems- instead of if p, then q- they translated to real life situations.
|
|
|
|
Why do people make the inferences that they do?
|
People have mental representations similar to sentences in predicate logic. People have deductive and inductive procedures that operate on those sentences.The deductive and inductive procedures, applied to the sentences, produce inferences.
|
|
|
|
Hippocrates
|
thoughts/feelings arise ONLY from the brain
|
|
|
|
Crick
|
Astonishing hypothesis: You–your joys and sorrows, your memories and your ambitions, your sense of personal identity and free will, are in fact no more than the behavior of a vast assembly of nerve cells and their associated molecules.F. Crick
|
|
|
|
logic
|
Formal science that investigates the structure of statements and arguments through inference
|
|
|
|
inference
|
Process of deriving a conclusion based on existing knowledge
|
|
|
|
formal logic
|
is the study of inference with purely formal content, where that content is made explicit.
|
|
|
|
modal logic
|
allows manipulation of necessity and possibility • “It is possible that Jose is tired today”
|
|
|
|
epistimic logic
|
knowledge and belief • e.g. operator “Kp”: it is known that p
|
|
|
|
deontic logic
|
morality • e.g. pis permissible or forbidden
|
|
|
|
problems with predicate logic
|
–Hard to deal with time(“now”, ”later”...) –Metapropositions(“asks”involves and asker and the proposition that is asked)
|
|
|
|
streghts and faults of Propositional & Predicate logic
|
are good at true/false statements. –Can’t handle uncertainty • Use probability theory instead • e.g. P(j)= 0.9, high probability of Jose being tired
|
|
|
|
rules of inference
|
modus ponens, modus tollens
|
|
|
|
Johnson-Laird
|
Deductive reasoning not carried out by formal logic rules nor schemas, but by mental models • mental representations that correspond in structureto the situations they represent
|
|
|
|
categorical syllogisms
|
–Universal / affirmative • AllX are Y –Particular / affirmative • SomeX are Y–Universal / negative • NoX are Y–Particular / negative • SomeX are notY
|
|
|
|
Affirming the antecedent
|
if p, then q; p therefore q
|
|
|
|
denying the consequent
|
if p, then q; not q therefore not p.
|
|
|
|
affirming the consequent
|
INVALID Reference: if p, then q: q therefore p
|
|
|
|
denying the anticedent
|
INVALID Reference: if p, then q; not p therefore not q
|
|
|
|
Logic Theorists
|
Newell, Simon, Shaw 1Starts with a series of logic propositions from which a theorem is to be proven• Analogy with a one person board game –Propositions are the starting positions –Rules of inference are the permitted moves –Proving a theorem = asking whether a given state of the board is reachable by a sequence of legal moves from one of the starting positions Addressed the combinatorial explosionproblem, and the need to use heuristics
|
|
|
|
heuristics
|
rules of thumb that contribute to satisfactory solutions without considering all possibilities. May help in processing, but doesn't guarantee a solution
|
|
|
|
General problem solver
|
(Newell and Simon, 1972)
• Universal problem solver machine – i.e. any formalized symbolic problem could be solved, in principle, by GPS – e.g. chess playing, theorems proof, geometric problems...• User defines objects and operations that can be done on the objects• GPS generates heuristicsin order to solve problems • It then creates subgoalsto get closer and closer to the goal• Problem: Can not solve any real-world problems (i.e. non- symbolic)• Eventually evolved into SOAR |
None
|
|
|
SOAR
|
SOAR
(Newell, 1990, 1993) • SOAR’smain element–the idea of a problem space: all cognitive acts are some form of search task• Memory is unitary and procedural –no distinction between procedural (skills) and declarative memory(facts and experiences) • Chunking: primary mechanism for learning –represents the conversion of problem-solving acts into long-term memory • e.g. IF you want to get from campus to home, THEN drive |
None
|
|
|
ACT-R
|
“Cognitive skills are realized by production rules”
• These rules operate on problems –e.g. addition• Production rules organized around a set of goals • One goal always active at any given time • Working memory: knowledge that the system is currently attending to |
None
|
|
|
rules
|
natural way of describing human knowledge-
–Innate • biological circuitry, e.g. vision–Learned by inductive generalization • formed from examples • formed by chunking(SOAR) or composition(ACT)–Learned by specialization • specific for a given situation–Learned by abduction • rules run backward to provide explanation –Learned by their performance • Incremental learning through associated usefulness value |
None
|
|
|
rule based systems
|
–Had psychological aims from the beginning
• as opposed to logic-based models–Abandon FOL‘sexpressiveness for simple if-thenrules –Computational advantages • concise and independent representations–Psychological plausibility • strongest among all CRUM based approaches |
None
|
|
|
FOL vs. rules
|
–It is apparent that logic is not the way to go in cognitive science because of the lack of core irepresentation and computation
|
|
|
|
Concepts
|
Mental representations of stereotypical situations or entities (objects) in a given category. Can be used as a single entity.
|
|
|
|
concept acquisition
|
old school- innately (plato, chomsky)- contemporary- through experience and from other concepts.
|
|
|
|
when did concepts come into Cog Sci
|
Minsky(1975): “Thinking should be understood as frame application rather than logical deduction”
|
|
|
|
concepts comprised of;
|
schemas (Rumelhardt), frames (Minsky), Scripts (Schank and Abeson)
|
|
|
|
frames
|
Data-structure for representing a stereotyped situation- have slots
|
|
|
|
frame slots contain
|
–Frame identification
–Relation with other frames • e.g. hotel phone phone communication device–Requirements for frame match • e.g. object specifications –Default information • e.g. chairs have 4 legs, hotel beds made by the staff –New instance (unspecified) • e.g. color of bedspread |
None
|
|
|
multiple frame issues
|
Ambiguous multiple inheritance (which frame to use when multiple apply)- new class is created.
|
|
|
|
scripts
|
Frame-like structures representing a stereotyped sequence of eventsin a particular context
|
|
|
|
script components
|
"–Entry conditions
"–Entry conditions –Results –Props –Roles –Scenes " |
None
|
|
|
script problems
|
"1. Impossibility to know ahead of time what occurrences will break a script
–e.g. “John was eating dinner at his favorite restaurant when a large piece of plaster fell from the ceiling and landed on his date...”–Q’s: Was John eating a date salad? What did John do next?Was John’s date plastered? –Structured representations can be inflexible! ... 2Script match –e.g. “John visited his favorite restaurant on the way to the concert. He was pleased by the bill because he liked Mozart” –“Bill”can refer to dinner check orconcert playbill –Which script should be called?" |
None
|
|
|
concepts, like rules- can be
|
"• Innate: basic concepts + mechanisms to form new ones
–“hardwired”to recognize faces, properties of physical objects...• Formed from examples –Sample sizeneeded for generalization varies–Plasticity (tuning) from further examples, i.e. online adaptation • Formed from other concepts" |
None
|
|
|
analogy
|
Analogyis a correspondence or partial similarity between two situations (analogs) – It usually uses “like”or “as”. John is like a pig
|
|
|
|
methaphor
|
is an indirect comparison between two (or more) seemingly unrelated subjects – It usually uses “is a” john is a pig
|
|
|
|
analogy in AI
|
case-based reasoning (AI)
|
|
|
|
how to distinguish superficial similarities from important ones in analogies
|
Find the causal relations that give the right and relevant (to the goal) outcome • i.e. analogy representation requires causality
|
|
|
|
Framework for computational models of analogical reasoning
|
retrieva(remembering), comparison, adaptation
|
|
|
|
3 kinds of learningin analogical reasoning
|
storage, adaptation, generalization
|
|
|
|
Keysar
|
Literal and metaphorical processing interact with each other (Keysar, 1990) – i.e. metaphorical interpretation happens simultaneouslywith literal processing, not after
|
|
|
|
Glucksberg& Keysar
|
subjects asked to signal when sentence is literally false
|
|
|
|
Boroojerdi
|
zapping left prefrontal cortex (PFC) speeds up analogical reasoning (!)
|
|
|
|
Glasgow & Papadias
|
Knowledge representation scheme that brings to the foreground the most important visual and spatial properties of an image
|
|
|
|
Computational Imagery
|
Glasgow & Papadia’sscheme on computational imagery: – 3 representations for 3 kinds of processing Deep, spacial and shape
|
|
