Lecture Notes | Second Midterm Flashcards Preview

🚫 PSY100H1: Introduction to Psychology (Winter 2016) with J. Vervaeke > Lecture Notes | Second Midterm > Flashcards

Flashcards in Lecture Notes | Second Midterm Deck (54):
1

John Locke: Blank Slate

  • John Locke, empiricism, and tabula rasca (blank slate); completely bottom-up from experience rather than top-down from concepts and expectations
  • problem of blank slate: the paradox of learning, i.e. something must be innate; how much can you start with absolutely no knowledge?
  • Locke argued that there are four principles of association:
    • similarity; e.g. apples and oranges
    • contrast; e.g. black and white
    • contiguity; e.g. students and seats
    • frequency; e.g. salt and pepper
  • events occur to an organism that starts out (innately) with the four principles of association
  • by associating events together the organism organizes experience into knowledge
  • this is completely bottom up view of learning was taken into psychology by behaviorists
  • they wanted to explain behaviour only using stimulus and response, i.e. no inner (non-scientific) processes; so no top down processes
  • but something must be innate—four principles of association identified with what Pavlov and Thorndike had found

2

Classical Conditioning: Similarity, Contrast, Contiguity, and Frequency

  • work done by Pavlov, so also known as Pavlovian conditioning
  • what was the nature of the pairing?
    • similarity, contrast, contiguity, and frequency
  • stimulus generalization: similarity between two stimuli
    • e.g. a bell with a similar tone to the first will also trigger the conditioned response.
    • but if we gave a shock every time the word “car” was said, would you generalize to “tar” or “truck”? you’d actually generalize to “truck” because you associate it with the meaning
  • stimulus discrimination: contrast between two stimuli
    • e.g. a bell with a very different tone won’t trigger the CR
  • contiguity; the presentation of the neutral stimulus and the unconditioned stimulus occur close together in time and space
  • frequency; Pavlov repeatedly paired the bell with the meat powder
  • these four models explain some types of conditioning, but not all; association is not enough to explain learning

3

Logical Problems with Classical Conditioning

  • similarity is not in the world, i.e. not in the stimulus
  • any two things can be very similar, if “similar” meanings sharing properties
    • e.g. “blue” and “shoe”, “blue” and “red”, “blue” and “blend”, etc.
  • there must be a top-down process that selects which properties to pay attention to when determining the relevant similarity.
  • associations are symmetrical but a lot of your thoughts are not; association = co-activation
    • e.g. Mary; Bill; loves—“Mary loves Bill” not the same as “Bill loves Mary”
    • your thoughts have a specific direction to them that associations do not
    • this direction is important since it affects the truth or falsity of your thoughts
  • so there seems to be a top-down process that imposes a logical structure on the stimuli in order to turn it into a thought (something that could be true or false, unlike an association such as salt and pepper)
  • Perceptrons: machines that simulated stimulus-response networks.
    • Minsky and Papert (1969) mathematically proved that there were many functions Perceptrons could not learn, such as exclusive “or”, yet even rats are capable of exclusive “or”, so something’s wrong
    • looks like behaviourist machines are too simple

4

Experimental Problems with Classical Conditioning

  • issues with frequency as a driver of conditioning
  • appetitive conditioning: when the US is an event that is usually considered pleasant, and that the organism seeks out
  • aversive conditioning: when the US is an event that the organism considers unpleasant and seeks to avoid
    • occurs much more rapidly than appetitive conditioning
    • often one or two pairings of the NS and the US are needed to turn the NS into a CS
  • it’s not simple frequency that’s at work; instead of association between the NS and US, the organism is using the NS as a signal for the US, i.e. using the NS to predict that the US will appear
  • prediction is not symmetrical like association
    • A predicting B doesn’t mean that B predicts A; no direction
  • organisms often use the NS as a predictive signal (predictive pattern detection) then signal detection theory applies
    • e.g. remember the gazelle and the noise in the bush
  • it’s not just simple frequency; the brain is doing signal detection criterion setting in order to deal with risk in prediction
  • there’s a lot of internal processing
  • the temporal arrangement of stimuli: delay, trace, simultaneous, and backward
    • if contiguity drives conditioning, the more contiguity between the US and CS should produce the most association
    • trace is often the best type of conditioning; it gives you time to predict and prepare for the US

5

Blocking | Experimental Problems with Classical Conditioning

  • an experiment by Kamin (1968) demonstrated blocking
    • 1st group: light signals US; light and tone signal US; tone does not become a CS
    • 2nd group: no light signal training; light and tone signal US; tone becomes a CS
    • for the first group, the light and the tone are redundant; the light already predicts the US, so they don't need the tone
  • latent inhibition: unfamiliar stimulus more readily conditions than a familiar stimulus
  • overshadowing: more salient member of a compound more readily conditioned as CS and interferes with condition of less salient member
  • however, not just prediction, but also preparation
  • Rescorla (1988) and Hollis (1997): argued that the CR is not a copy of the UR
    • e.g. rat sees a snake and it runs, but if it’s conditioned with a bell noise for snakes it freezes when it hears the bell
    • prediction + preparation = anticipation; the bell is being used to predict the appearance of a snake
    • the CR is the exact opposite of the UR
  • classical conditioning is not association but anticipation, it’s a much more intelligent process

6

Beginnings of Operant Conditioning

  • Thorndike (1911): did experiments where he put a cat in a box and had it pull a lever; this lead to the discovery of:
    • law of effect: the idea that responses followed by satisfaction will occur again and those that are not followed by satisfaction become less likely
  • Kohler and Sultan: from the gestalt tradition emphasizes top down processing and the importance of insight learning
    • Sultan the monkey learned how to pile up boxes in order to get to the bananas
    • he demonstrates an S-curve with his learning, i.e. insight; once he learns how to do it, he keeps doing it
  • B.F. Skinner: was the most important of the behaviourists and greatly refined the puzzle boxes of Thorndike
    • these boxes were called operant conditioning chambers, or The Skinner Box

7

Rambaugh et al.: Insight and Abstraction

  • the presence of both insight and abstraction in learning also emphasized by recent work in operant conditioning specifically in the salience based theory of learning of Rumbough et al. (2007)
  • conditioning is based on what the organism finds salient
    • e.g. rats trained to run a three dimensional maze will spontaneously find short cuts by jumping between levels, for which they were never trained
    • the rats restructure what information is salient to them in order to produce insightful solutions beyond what they were trained/conditioned
  • so again, what we have is the intelligent, even insightful, anticipation of the world in which internal processes of information selection (salience) are interacting with internal processes of abstraction and insight in order to produce behaviour
  • Rumbough et al.: “What is trained does not necessarily equal to what is learned.”
  • all of this points to conditioning not following well the behaviourists model of learning
  • instead learning seems to be a process involving the interaction of top down and bottom up processes

8

Bandura: Observational Learning and Modelling

mirror neurons and imitation

  • when children human and chimps are given boxes that contain candy, humans will try to do every single step whereas the chimps skip to the last one to get to the candy faster
  • this is because the children are assuming that the adults know more than they do — they gain more in the long run, about culture and social norms; the chimps are getting short-term rewards of candy

empathy, mindsight, and mindsight resonance

  • your ability to introspect what’s happening within yourself makes you better at telling what’s going on with other people
  • when two people do the same things together (e.g. nod your head, doing the wave) they tend to like and trust each other more

9

Harlow: Monkeys and Mindfulness

  • Harry Harlow (1949): monkeys capable of more abstract learning
  • +stimulus covers a raisin; -stimulus does not
  • the monkeys had to move the object to uncover a raisin
  • problems:
    • only six trials; too short for operant conditioning
    • also a previous +S can become a -S, and vice versa 

10

Shiffrin and Atkinson Model of Memory

11

Sensory Memory

  • sensory memory: a memory system that momentarily preserves extremely accurate representations of sensory information
  • information that is not quickly passed to short-term memory is gone forever
  • iconic memory lasts about 1/3 of a second
  • echoic memory lasts about 2 seconds
  • main function is to give us continuous scenes and sentences
    • iconic memory: when you look around, you only pick up on a few things you focus on, but having this memory allows you to put things in a big picture; evolutionary advantage that would allow us to know what are in our surroundings (e.g. the tiger creeping up behind you)
    • echoic memory: when you’re listening to things, the sounds coming out of someone else’s mouth needs to stay in your mind long enough for you to process things

12

Implicit Learning and "Psychic Powers" | Sensory Memory

Reber and artificial grammar learning, and ‘psychic’ powers

  • e.g. give people a random string of numbers and letters: e.g. 1aaddf3jkll
  • you ask people to tell the similarities between the sets of strings they’re given: they’ll say either (1) ‘I don’t know’ or (2) give you a rule that doesn’t actually do it (e.g. there’s always two vowels in a row)
  • if you then give them the rule, people do worse at telling if it works than if they’re doing it implicitly
  • ‘psychic’ phenomena; e.g. ‘I can feel when I’m being stared at’
    • they bring in people to stare at them and they were scoring well above chance
    • in a replication, people were not given feedback (‘yes, you’re right’ or ‘no, you’re wrong’) and their performance dropped to chance level
    • this is because the experimenters were bringing people in using a complex pattern, and the participants were using implicit learning to pick up on the pattern
    • most instances of psychic experiences are a result of implicit learning

13

STM: Working Memory

  • at most, you can remember 4 ± 2 things
  • the number of things you can hold varies considerably based on the chunking phenomenon; chunking makes information co-relevant to one another
    • e.g. UNICEFFBICIANATO is harder to remember than if you broke it down into UNICEF, FBI, CIA, and NATO; it’s more relevant to you
  • chunking indicates that working memory can process a lot more information if it has been patterned
  • working memory is a relevance filter more than a holding space
  • main job of working memory is not so much to hold things for offline manipulation but more to select from all the information, what the relevant information is that is going to get into long term memory
  • things lines up with four things:
    • measures of working memory correlate very strongly with measures of high general intelligence
    • working memory has a high overlap with consciousness
    • (two above things—many people argue has to do with filtering for relevant information)
    • learning is salience based,
    • and what information, the ‘meaning’, gets into long term memory
  • with time and age, your working memory begins to decline; it seems harder for them to stay on topic, their filters aren’t working as well as your younger one
  • but keeping a lot of stuff that seems irrelevant can help older people see things deeper than younger people don’t; they tend to seem wiser
  • there’s some good evidence that long-term mindfulness practice will improve your working memory

14

Long Term Memory

  • it doesn’t seem that there’s a capacity; e.g. nobody says ‘Don’t tell me anything else, I won’t be able to remember it!’
  • your memory is not about an accurate remembering of the past, but an intelligent prediction of the future
    • you don’t want to be an accurate but stupid video recorder, rather a relatively inaccurate but highly intelligent problem solver
  • the ‘common-sense’ idea that memory is about recalling things perfectly is false
  • 49% of confident eye-witness testimonies are false
    • this is why we can’t have capital punishment; our memory isn’t good enough to be able to say that people are truly guilty

15

Sachs (1967): Sentence Recognition | Long Term Memory

  • Sachs, 1967: experiment with sentence recognition
  • you have people reading a text, stop them, and ask ‘Is this the sentence you just read?’
  • change in form noticed but not change in meaning
  • this is why your brain is very misleading

16

Bransford and Franks (et al.) | Long Term Memory

  • Bransford and Franks (et al.): work shows that people often ‘remember’ a more meaningful thing that they never encountered
  • you show people a bunch of random dot patterns
  • then show them a pattern that represents the average of all the patterns they saw, but which they never saw
  • people will report strongly remembering having seen the average pattern
  • that pattern wasn’t seen, but is the best anticipation for what might come in the future; it’s the best generalization for the future

17

Lotus and Palmer | Long Term Memory

  • transfer appropriate processing means that the brain will adjust to solve a project
  • Loftus and Palmer: participants viewed a video of a car crash
  • two conditions:
    • control question: How fast were the two cars going when they contacted each other?
    • leading question: How fast were the two cars going when they smashed into each other?
  • smashed = 65.7 km/h ; contacted = 51.2 km/h
  • the wording of questions can alter memories; children are particular susceptible to this 

18

Martensville, Saskatchewan (1992) | Long Term Memory

  • one woman accused another who ran a daycare of sexually abusing her child
  • no parents had actually seen evidence of mistreatment
  • none of the children had complained to their parents
  • police investigated, questioning the children who went to the daycare; nothing suspicious was reported
  • after repeated questioning by a rookie officer with only 7 months of experience, children began to report sexual abuse; allegations began to snowball
  • during the investigation, the children at the day-care center were repeatedly questioned about the abuse by the police, who used suggestive techniques to obtain testimony of abuse
  • rumors of child mutilation and other obscene things surfaced; rumors of a Devil Church and cult where police were abusing children undercover
  • the RCMP arrested 9 people, 5 of them police officers, and charged them with over 100 counts of sexual and physical abuse
  • a young man visited children at their preschool, read them a story handed out treats, did nothing aggressive, inappropriate, or insulting
  • they came back a week later and asked the children two sets of leading questions (1) whether or not something happened (2) questions about what other children had said
  • the number of children who said "yes" to leading questions skyrockets to 70-80% in the second condition 

19

Craik and Lockhart: Levels of Processing | Long Term Memory

  • the more meaning oriented someone’s processing is, the more likely they will remember the information processed
    • the intention to remember does nothing to actually help you remember things
  • this was the central claim of Craik and Lockhart’s theory called Levels of Processing; consider these three sentences:
    • Is the world table in capital letters?
    • Does the word blue rhyme with the word shoe?
    • Does the word friend make sense in the phrase "he really likes his friend"?
  • then, unexpectedly they were given a word recognition task; they’re asked if they saw the bold words in previous sentences ‘Did you see the word table, friend, blue, etc.?’
  • people remembered words much better in the third type of sentence because they were processing meaning in a deeper fashion
  • problem is, what does term ‘deeply’ mean?
    • it’s a spatial metaphor for being more meaningful, but what is the independent measure that something is more meaningful?
    • if meaningful just means remembered better, then we have a circular explanation — ‘things are remembered better because they’re remembered better’
  • is how meaningful something is a constant property of that thing?
    • e.g. there’s someone who ‘means the world to you’ but you drift apart and five years later you see a photo of them and you think, ‘What? What did I see in them?’
    • so, meaning is probably not a constant property

20

Transfer Appropriate Processing | Levels of Processing

  • Morris, Bransford, and Franks (1977): repeated the standard levels of the test (‘did you see the words?’), but had another condition:
    • e.g. ‘did you see a word that rhymes with blue?’
    • now, participants remembered more of the words that had been processed in the second type of sentence
  • memory seems to work more in terms of transfer appropriate processing, i.e. the brain is trying to pick up on patterns that will transfer well to the future
    • i.e. the more how you’re remembering is similar to how you’re going to have to recall it in the future, the better you’ll remember it
  • the brain’s trying to find information in the current context that will be relevant to its retrieval later on
  • the problem facing your brain is that it can’t always tell how it will have to process things in the future, so it grabs onto the most general thing it can
  • memory as reconstructive rather than reproductive
    • your brain doesn’t store all of the information, it just gets the key information and some rules/principles that allow it to reconstruct things later
    • this is adaptive; instead of carrying around all the materials for your shelter every time you move, you only need to carry the essential ones to reconstruct it so that it won’t weigh you down

21

Carmichael et al. (1932) | Levels of Processing

In this experiment, people were given pictures with one set of words. They went away for a week, and when they came back were asked to draw the pictures they saw the week before, as accurately as possible.

22

State Dependent Memory: Bridging Between Concepts | Long Term Memory

places are retrieval cues

  • place characteristics like sound, visual features, room size, odors, etc. all get encoded with the material you’re remembering
  • e.g. memories when you go back to your childhood home, school, street, etc.
  • e.g. traumatic memories and triggers

Smith et al. (1978): people were given 80 words

  • you allow people to return to the same room they memorized them, and they remembered 49 words
  • you make people go to a different room after they’ve memorized these words, and they remembered 35 words

even state dependent memory effects have been found

  • be careful when you’re hiding something; the state you're in now may change, and you may not remember where you hid something later
  • e.g. if you lose your keys when you're drunk, just try getting drunk again to find them

23

Types of Long Term Memory

24

Different Kinds of Knowing

  • propositional knowing: something that’s the case, works in terms of truth conditions; great for learning rules
  • procedural knowing: how to do something, works in terms of appropriacy conditions; great for learning routines
  • perspectival knowing: what it’s like to be x, works in terms of identity assumption conditions; great for learning roles

25

Memory and Problem Solving

  • Weisberg and Alba (1981): told people to "think outside the box" in the nine dot problem, but it wasn’t effective
  • Casselman (1970): tried a different clue; simply drew a huge box around the nine dot problem on the page, which worked!
  • learning, memory, and problem solving are all interlinked in terms of the brain trying to find relevant information for achieving its goals
  • memory is very intertwined with learning, and problem solving
  • memory has a lot more to do with the future than the past

26

Newel and Simon: The Formulization of Problem Solving

a problem is made up of four components:

  • the goal state; e.g. I want to be not hungry
  • and the initial state; e.g. I’m hungry
  • (there’s a problem when the goal state and initial state aren’t the same)
  • search-space (or problem space): the number of different paths you could take in order to get from initial state to the goal state
    • the problem with this model is that it’s misleading; you’re not looking at things from a birds-eye view, you’re in one of those points
    • F^D: size of a search space, where F is the number of operators and D is the number of steps
    • e.g. average chess game: 4.239 x 10^88
    • compared with 10^10, which is the approximate number of neurons in the human brain, versus 5 x 10^14 which is the approximate number of synaptic connections in your brain, versus approximately 10^80 number of atoms in the universe
    • so there aren’t even enough atoms in the universe to represent the number of moves you could make in a chess game
  • operators
  • path constraints: there’s always more than one problem
    • if your problem is cooking dinner, you could solve it by burning your house down; but you probably don’t want to do that because your house solves a lot of other problems for you
  • a problem solution is a sequence of operations that turns the initial state into the goal state while obeying the path constraints
  • a problem solving method is any technique that finds a problem solution

27

Algorithms and Heuristics | Memory and Problem Solving

  • algorithm: a problem solving technique or method that is guaranteed to find a solution
    • so there’s an algorithm for determining the number of people in the room (e.g. counting)
  • heuristic: a problem solving technique/method that increases the chance of finding a solution
    • think of some heuristics in chess; they can increase your chances of winning, but there’s still a chance you can lose
  • for many problems we can’t use algorithms because we face what Cherniak called the finitary predicament
  • first related point: given combinatorial explosion, and an algorithm requiring an exhaustive search space, the use of an algorithm commits the problem solver to a solution attempt that for all practical purposes is impossible; we don’t have enough time or resources (think back to the size of the search space of a game of chess)
  • second related point: since most of the time we have to use heuristic, we face the consequence of their use
    • heuristics try to pre-specify what information is relevant, they try to pre-judge what information will be relevant; source of prejudice or bias
    • e.g. you drive your friend to the airport and tell them to have a safe trip, that you hope they don’t crash; and then you get back into your car, which is the number one killer in America, but don’t worry about dying
  • deep fact: relevant is not in the world, so no piece of information is intrinsically, or always, relevant; think of the nine dot problem and how the shape was actually irrelevant
  • so something other than heuristics at work when we are problem solving
  • what seems to be at work is problem formulation: how the initial state, goal state, operators, and path constraints are represented, this seems to really effect the size and shape of the search space

28

Kaplan and Simon (1980): The Mutilated Chessboard Problem

  • problem formulation/framing needs to be recursively self-correcting; this may have connections to working memory as a higher order relevance filter and also to intelligence
  • humour is a quick marker of social awareness and insight problem solving, which are really adaptive, which makes people seem more attractive
  • what improves insight? insight is mostly a matter of attention and salience re-construal

29

Knoblich (1999, 2001): Chunk Decomposition | Insight

  • Knoblich and associates 1999, 2001 chunk decomposition and constraint relaxation
  • reordering matches to make a true statement; people tend to move the matches associated with the numbers rather than the operations
  • if you’re good at chunk decomposition, you’re also probably good at insight

30

DeYoung, Flanders, and Peterson (2008): Frame Breaking | Insight

  • frame breaking: breaking up a problem formulation
  • the anomalous card sorting task—flash a few anomalous cards (e.g. black cards that are of the heart suit, which are supposed to be read) among regular cards and identify when an anomalous card goes by
    • the better you are at identifying anomalous cards (frame breaking), the better you are at solving the 9-dot or mutilated chess board problem
  • note how this isn’t related to how well you reason, but how well you re-construe what information is relevant
  • global processing of gestalts --> local processing of features

31

Baker-Sennet and Ceci (1996): Leaping | Insight

  • Baker-Sennet and Ceci (1996): evidence for the facilitation of insight through the redirection of attention in the opposite direction
  • leaping; when you can use fewer cues to get the right answer, e.g. seeing a sofa from only a collection of dots
  • the better of a leaper you are, the better you are at solving insight problems
  • further evidence for the facilitation of insight through the redirection of attention in the opposite direction; so, do we have a contradiction?
  • however, we’ve already established that directing attention “up” hampers insight problem solving
    • e.g. if you “leap” and perceive a square in the 9-dot problem, then it’s hindering

32

MacCrae and Lewis (2002): Navon Letters | Insight

  • MacCrae and Lewis (2002): used Navon letters and facial recognition; evidence that moving to the featural level can impair gestalt-level processing
    • congruent: a big E made up of little Es; global E, local E
    • incongruent: a big H made up of little Es; global H, local E
  • experiment: you give people a bunch of photos and then ask if they’ve seen these same faces in a second round
    • if you get shown a face, it’s taken away, and then you’re asked to described it, then your ability to identify the person accurately actually goes down
  • is this because you’re using language; is that the issue?
  • MacCrae and Lewis ask you to look at a face and then Navon letter; this also impairs your ability to recognize the face
  • what’s happening when you have to describe a face is that you drop down to the featural level, which ruins your attention; so it has nothing to do with language
  • if you’re asked to describe a face and then have to listen to music, you improve your ability because music involves a global level of processing

33

Phase-Function Fit | Insight

  • how do we explain the contradictory results of Young, Flanders, and Peterson, and Baker-Sennet and Ceci?
  • let’s use the same strategy that Rummelhart employed to resolve the chicken-and-egg reading problem (when you had to read the letter to read the word but read the word to read the letter, which made reading impossible)
  • there are multiple levels of processing that are going both top-down from the gestalt and down-up from the features at the same time in order to process these features
  • Förster and Dannenberg (2010): parallel opponent processing
  • Lewis (2005): dynamical systems

phase/function fit; timing is everything

  • insight requires that you break in an inappropriate frame and make an appropriate one
  • when you’re in the breaking phase, then you should be going down, to the featural level
  • when you’re in the making phase, then you should be going up, to the gestalt level
  • if you do these in the wrong way, then you’ll impair it

34

Mindfulness | Insight

  • because timing is everything, you have to gain skills to be able to direct your attention appropriately; is there any way to train this skill?
  • yes; mindfulness as the training of attention in frame breaking and frame making
  • meditation and the scaling down of attention by focusing on sensations of the breath
    • do this to get better at breaking the frame
  • contemplative practices by scaling up attention and realizing how everything impermanent and/or interconnected
  • non-dual practices that synergize the two movements

Moore and Linowski (2009): reducing the Stroop effect

  • read out the colour that each of the words are printed in: green, yellow, blue (each printed in different colours)
  • by doing mindfulness practices, people were able to do this better; the Stroop effect was reduced

35

Insight Summary

  • at the core of problem solving is problem formulation/framing
  • a central characteristic of such framing is insight which has to do with recursive relevance realization, i.e. re-realizing what we find relevant/salient
  • insight is probably a dynamically self-organizing top-down and bottom-up processes of attentional scaling
  • such attentional scaling is trained to mindfulness practices which improve insight; increase cognitive flexibility, range, and phase-function fitting

36

Reasoning and Rationality

  • rational: using the most reliable means to obtaining your goals
    • that’s why you always want to be rational; why wouldn’t you?
    • rationality and logic are not the same thing
    • so if your goal is truth then you should use methods that have been worked out to be the most reliable for achieving the goal of truth

reasoning; a conjunction is always less probable than a conjunct

  • e.g. which is more probable? having hair or having blonde hair?
  • e.g. Linda is a 31-year-old, single, outspoken, and very bright. she majored in philosophy, she’s concerned with issues of discrimination of social justice, and participated in the anti-nuclear movement. which is more likely? (a) Linda is a bank teller or (b) Linda is a bank teller and active in the feminist movement.
  • most people want to choose (b), even though (a) is more likely
  • this is because your brain wants to make more sense of all the information; i.e. why does it matter if Linda is all those things if we can’t apply it? 
  • the conjunction fallacy: people consider the conjunction to be more probable than the conjunct
  • participants consider the description of Linda relevant, and then are trying to predict a causal relationship between those facts and how Linda will turn out as a person
  • so when people are reasoning they’re still using the learning machinery, the memory machinery, and the problem solving machinery, not just the formal system of probability

37

Using Heuristics in Reasoning and Rationality

  • people are not just pursuing truth but are after making sense of a situation
  • people are using heuristics when they reason, rather than using the algorithms of probability theory
    • they’re using the representativeness heuristic: how well an example represents a category;
    • and the availability heuristic: how easily examples come to mind (determines how probable your mind finds it)
    • e.g. plane crashes vs. driving home from the airport
  • again, we can’t use algorithms most of the time or we’d be lost in combinatorial explosion; that would be cognitive suicide; we’d never solve any of our problems
  • but when we use heuristics, we often come to the wrong conclusion; it achieves goals that the person doesn’t want to achieve
  • so being logical doesn’t equal being rational; but this doesn’t mean you should be illogical
  • so how do we get clear about rationality?

38

Computational Limitations | Reasoning and Rationality

  • Keith Stanovich from U of T—intelligence: dealing with computational limitations; that is what we’re measuring in I.Q. tests
    • what we’re actually measuring in an I.Q. test is how well your brain is able to sort through the combinatorial explosion and pick out the relevant information
    • strong positive manifold in those tests pointing to a core ability; i.e. how well you do on Test A is very predictive of how well you do on Test B and Test C, and every combination of those
  • Verveake and Ferraro (2013): computational limitations means avoiding combinatorial explosion, which means relevance realization
  • remember high correlation between measures of working memory and I.Q.; Hasher’s work arguing that working memory is a relevance filter
  • so perhaps intelligence is the brain’s ability to realize relevance

39

Strong Positive Manifold | Reasoning and Rationality

  • if rationality just means dealing with computational limitations (avoiding combinatorial explosions with relevance realization) then rationality would just equal intelligence
    • intelligence = dealing with computational limitations = rationality
    • so if rationality isn’t logic is it intelligence?
  • no; measures of rationality (e.g. critical detachment, making good probability judgements, avoiding the confirmation bias, etc.) also have a strong positive manifold
    • all of the different mistakes you can make is highly correlated/predictive of how likely you are to make the other mistakes
  • what’s the correlation between the strong positive manifold of intelligence and the strong positive manifold of rationality?
    • the answer on average is 0.3
    • this means intelligence is necessary but not sufficient for rationality; something else important is at work
    • you can be highly intelligent but irrational at the same time—there’s no contradiction there
    • what we should be measuring in society is rationality, not intelligence

40

Intelligence and Rationality

  • intelligence is about relevance, but perhaps rationality is about making sure relevance tracks the truth, i.e. what what you find salient is what will most reliably get you to the truth
    • the art of bullshitting; people don’t care about what they’re saying, they’re just trying to make it salient to you
    • most advertising is trying to get you to bullshit yourself
  • intelligence has many goals, such as making sense and explaining situations for adaptive prediction of the world; maybe rationality is trying to make truth-seeking the most relevant goal for certain tasks
  • when what we care about most is obtaining truth then we need to train our relevance realization machinery, our intelligence, to zero in on the information that leads to truth
  • we have to apply our intelligence to our intelligence; we have to adopt a particular cognitive style
  • cognitive style is a set of skills, sensitivities, and motivations, that effects what one finds relevant/salient/important; many cognitive styles can be learned
  • for example, one can learn to pay more attention to, and find more important, one’s cognitive process as opposed to one’s cognitive product, i.e. what one’s cognition produces
    • what one believes is either true or false, you can’t say it’s (ir)rational; how rational it is comes about as a result of how the belief’s produced

41

Active Open-mindedness | Reasoning and Rationality

  • Stanovich argues and has evidence that this importantly contributes to rationality; notice that this is a kind of mindfulness
    • has a ton of experimental evidence that the cognitive style of active open mindedness is what is most productive of performance in rationality tests
  • active open mindedness: a learnable cognitive style in which one first learns biases and framing and then one practices actively looking for them in one’s thinking
  • one then practices the habit of actively counteracting them when one is engaged in tasks that are focused on obtaining true information as their primary goal
  • so one uses one’s intelligence to acquire a cognitive style that better fits one’s intelligence to the goal of obtaining truth when that’s the primacy goal
  • so to be rational means to have active open mindedness telling you when and how to be more logical

42

Categorization

  • categorization: the mental ability to group things together so that they are sensed as belonging together; needed for encoding, abstracting, inferring, and communicating
  • it is that sense of belonging together that is the crucial psychology aspect
  • Smith (1995); Resemblance Theory of Categorization: we sense things as belonging together because we find them similar to each other
  • main problem with the Resemblance Theory is that, as we’ve seen, similarity is not a property of the world; it depends on what we consider the relevant factors of comparison
  • so instead, finding things similar in the relevant manner seems to depend on our concept of those things (not looking out and seeing similarities naturally occurring in the world)
  • there are three main psychological theories about what concepts are, and how they function

43

Classical Theory of Concepts

  • classical theory of concepts: concepts are mental definitions that list the necessary and sufficient features for being a particular thing; all the members of a category share an essence (i.e. the necessarily and sufficient properties, or essential properties)
  • given the notion of a concept as a conjunctive list of necessary features the classical theory is committed to some central predictions
  • what counts as a member of a concept’s category and what doesn’t will be clear, all members can be organized into taxonomic hierarchies with necessary deductions
  • e.g. if something is a closed, three-sided, straight-lined figure then it’s a triangle; it needs each feature, and if it lacks any of them then it’s not a triangle, and once it has all of them it’s definitely a triangle
    • something is either a triangle or it isn’t
    • if something is a right-angle triangle, then it’s necessarily a triangle
    • any other property is merely accidental
  • first problem: the theory claims that concepts are definitions, but as Wittgentstein pointed out, most concepts do not have definitions or essences
    • e.g. define “game”
    • there is no definition that will include all and only the things we call games; most things don’t have essences
    • instead, categories are related by what Wittgenstein called family resemblance; we tend to categorize things in terms of overlap, not essence
    • interestingly, science is about finding the things that do have an essence
  • second problem: whether something is in or outside a category is not always as clear as the classical theory predicts
    • furniture: chair – in, cucumber – out, grandfather clock – maybe, a throw pillow – maybe
    • all members are not equally good; using verification times it is clear that a robin is a better instance of a bird than a penguin
  • Hampton (1982): showed that taxonomic deductions break down
    • deck chairs are chairs – yes
    • chairs are furniture – yes
    • so deck chairs are furniture – no
    • what’s going on there? people reliably do that, but why?
  • many psychologists have abandoned the classical theory in favour of the prototype theory of concepts

44

Prototype Theory

  • prototype theory: there is no set of necessary and sufficient features, instead there’s a prototype which is a set of typical features, membership is determined by similarity to the prototype; not all members are equally good, and category types are fuzzy

first problem: how is the prototype generated if similarly is based on the prototype?

  • Roth and Shoben (1983): participants rate tea as more typical than milk for secretaries taking a break while milk is more typical than tea for truck drivers taking a break
  • so what’s the prototype for a beverage? tea or milk?
  • or do we have a prototype for beverage-for-secretary, beverage-for-truck-driver; so then you’d have to have beverage-for-athlete, beverage-for-child, beverage-for-someone-study, etc. to infinity?

second problem: typicality gradients do not seem to be stable

  • Medin and Shoben (1988): people reliably relate small spoons as more typical than large spoons, and metal spoons are more typical than wooden spoons
  • therefore, the typicality gradient should be SM, SW, LM, then LW
  • yet people overwhelming rate LW as more typical than either SW or LM spoons
  • Gleitman and Gleitman (1983): found the opposite; typicality effects for concepts that are definitional (and therefore shouldn’t have them), e.g. an odd number—5 is a more typical odd number than 7, why?

third problem: difficulty facing conceptual combination

  • conceptual combination is necessary to explain our ability to have some many new thoughts
  • so take the prototype for pet – furry, cat or dog sized, black or brown
  • then the prototype for fish – trout or salmon sized, grey colour, etc.
  • so a pet fish should prototypical be furry, cat or dog sized or trout sized, black brown, or grey?
  • prototype for city, maybe for American city, but for American city in the state of Washington not on the coast?
  • both of these are doubtful; prototype theory can’t explain how to combine concepts together, which is really important for a theory
  • we can’t even combine prototypes together the way we combine concepts

45

Missing the Gestalt: Problems with Classical and Prototype Theory

  • shared problems for both classical theory and prototype theory: feature lists and description
  • feature lists and the gestalt problem for concepts;
    • features of a bird – if I throw a pile of feathers and wings into the air, is it considered a bird? no
  • the gestalt: the structural-functional organization such that something behaves as a causal whole or unified thing
    • what’s interesting is that we know that this gestalt is what defines a concept, yet we can’t put it into words
  • classical and prototype theory assume that the main function of a concept is to describe or label the world
  • however, what if people are primarily trying to predict and explain the world?
    • they’re not just trying to describe it, they’re trying to predict and explain it
  • maybe concepts are more about trying to find gestalts that help to predict and explain the world; maybe concepts are more like theories than like pictures

46

The Micro-Theory of Concepts

  • the micro-theory of concepts: concepts are like theories that organize features in an attempt to explain/predict the world
  • can explain most of what the other theories could; theories can explain when we have definitions
  • it is when our explanations turn out to be law-like, e.g. F=ma, that we have essences
    • i.e. essences are what science finds when our theories turn out to be deeply true
  • can explain fuzzy boundaries; no theory applies cleanly to everything
    • e.g. Darwin’s theory of evolution can’t explain space travel
  • can explain while not all members equally good, theory fits better to some data rather than others
  • it can explain conceptual combination: consider engine repair vs. expert repair
    • e.g. how come you don’t think that “expert repair” means repairing your expert?
  • it can explain why people have relations between features
    • e.g. do small birds or large birds sing? how do you know this? (did you go to bird school where someone taught you that this was important) you know this because you don’t just have a feature list

problems:

  • verges on circularity: theories are made out of concepts which are made out of theories?
  • perhaps we have multiple levels of top-down bottom-up processing trying to make sense of the world
  • perhaps language causes us to fixate on certain levels
  • Matson’s concept of sizing up (revised by Vervaeke et al.)
  • we think that things have set features

but if that’s the case then our theory of categorization has just dissolved into our theory of learning, memory, problem solving, and in fact has just dissolved into our theory of intelligence

  • the problem of causal relevance: what caused the sinking of the Titanic? combinatorial explosion
  • the problem of prediction: which formulations/patterns are transfer appropriate?
  • how to find patterns? how to chunk data? how are pieces of information relevant to each other? what are the relevant similarities?
  • two issues that intersect with categorization: one is language and the other is intelligence
  • we’ve also seen the issue of intelligence intersecting with problem solving and reasoning

47

Language

  • five levels at which language is studied: phonetics, morphology, syntax, semantics, and pragmatics
  • phonetics: the articulation of speech sounds
    • e.g. why do you hear the “b” and “p” in “bat” and “pat” more distinctly than the “l” in “look” and “bottle”?
  • morphology: how speech sounds are put together into words
    • e.g. why can you have “ttle” in “battle” or “bottle” but not start words with those?
  • neither one of these overlaps very strongly with the topics investigated by psychology
  • syntax: how we structure words together into sentences/utterances
    • attempts to explain the amazing productivity of language;
    • that you have the capacity to understand and/or produce an astronomically vast number of sentences is amazing

48

Compositionality

  • compositionality: we have a finite number of elements (e.g. words) and then a finite number of rules (i.e. a grammar) for how words can be put together, i.e. how we can compose sentences
    • also exists for music, logic, mathematics, and thought
  • understanding compositionality in syntax is essential to understanding how we generate music, logic, math, or thoughts
    • so the same words can be used to mean different things (remember this from problems facing association); e.g. Peter loves Mary vs. Mary loves Peter
    • syntax therefore also takes us beyond association to more specific thoughts/propositions about the world
    • it gives us accuracy, i.e. make specific truths about the world
    • syntax both vastly increases the range of what we can think about because of compositionality and the accuracy with which we can think it because of the specificity of thought it affords
  • a lot of work in linguistics is to try and find out the basic elements of productivity and specificity of language
  • can we find the rules for how they produce the sentences they use and how they understand the sentences we produce, not just English but all languages?
  • can psychologists add anything to that?
  • evolutionary psychology can give us some ideas on how language arose and how it’s organize in the brain; this in turn may tell us how language is related to other things like music and gesture

49

Musilanguage

Stephen Mithen: in a book called the Singing Neanderthal, he, along with many other theorists, argues that language and music evolved out of a common source called musilanguage

  • this would help to explain their commonalities and that both are equally universal
  • it would help to explain why they go so naturally together in singing
  • but his hypothesis is controversial
  • however, it does point to some very recent and interesting psychology research

Thompson and Schlaug: made use of melodic intonation therapy to help restore language use to individuals who have lost it with stroke, or to autistic individuals who didn’t fully develop it

  • it involves having individuals sing words or short phrases while tapping each syllable with their left hand
  • what seems to be going on is that the music uses areas of the right hemisphere to help rewire the left hemisphere for language use

contrary to what the media says, language isn’t just in the left hemisphere;

  • a lot of very important features of language use are in the right hemisphere, these tend to be the more musical aspects of language such as tempo, intonation, volume, expression, etc.
  • this does point to an overlap, and possible common ancestor for music and language
  • however, it should be noted that melodic intonation theory has helped been used to help recover motion and memory; but that may still point to a shared evolutionary past

50

Arbib: Pantomime and Musilanguage

Arbib: recent work on the origin of language which comes right out of learning theory

  • his theory is based on the idea that we’re powerful imitative learners — mirror neurons that we have already discussed
  • we can trigger that imitative learning machinery in others through pantomime
  • what is interesting about pantomime is that it turns a learning process into a communication process; i.e. you take your imitation ability, and instead of learning, you just do it and communicates to someone what it is
  • what is even more impressive is that pantomime has a lot of features of language — productivity and displacement
    • displacement: the ability to make you think of things removed in time, space and modality (possibility)
  • is there any evidence for this hypothesis?
  • we’re constantly pantomiming even to ourselves all day long ­— it’s called gesturing
  • Goldin-Meadow: produced a lot of experimental work showing that gesture significantly effects learning, problem, solving, and communication
  • gesture is not an add on ornamentation but is doing serious cognitive work
  • there’s also the evidence in development for gesture and spoken language to be co-developing together and supporting each other; e.g. in the one-word stage young children supplement the spoken word with gesture in order to make more complete communication

it’s quite possible that musilanguage and pantomime co-evolved together; gesture and rhythm and tempo go very strongly together

  • there is also the universal existence of dance which combines gesture and musilanguage together
  • it’s quite possible that this was not just singing Neanderthals but dancing Neanderthals
  • pragmatics has two core theories: speech act theory, and conversational implicature theory

51

Speech Act Theory

Austin and Searle: speech act theory; based on the idea that language is not only about describing or explaining the world, but also about making things happen

  • so consider when a minister at a wedding says “I now pronounce you husband and wife”; he’s not describing something, he’s making something happen
  • same thing for a judge saying “I find you guilty as charged”
  • now in addition to such explicit speech acts there are implicit ones we’re making throughout the day
  • consider the utterance: “I’ll be at your place at 2”
    • can be used to promise, threaten, warn, plead, confirm, etc.
    • what’s interesting is how the speech act is not reducible to the words since the same words can be used for so many different acts
    • so something psychological is going on as well
  • what seems to be going on is an assessment of the appropriate conditions for a particular action have been met
  • what are the relevant features of the context for determining which action is being performed?
    • e.g. a minister can’t just run down the street shouting: “you! you! you’re married!”
    • what would have to be the case in order for “I will be at your place at 2” to be a threat?
  • speech acts require a lot of intelligence to determine what the relevant information is and to make the relevant predictions from it

52

Grice: Conversational Implicature

  • Grice; based on the fact that in communication we have to convey much more than we say
  • we have to rely on people to read between the lines, make connections, and extend beyond what we say
  • why? why don’t we always say what we want to say? because there would be too much to say
  • e.g. a person drives up in a car, rolls down the window, and says to someone on the sidewalk: “excuse me, I am still out of gas,” and the other person says, “oh, there’s a gas station at the corner”
    • there’s a lot that’s being conveyed by these two sentences; just take a few examples; anything being conveyed is violated, you’d be pretty angry
    • it’s close enough that you can still get there with the amount of gas you have, it’s not in the next city over
    • at the gas station, there is gas you can purchase for Canadian currency, you will not be asked for your first-born child
  • implicatures are different from logical implications because they assume the speaker is trying to co-operate with the listener to get the right conveyance
    • e.g. if someone asks how many kids you have and you say that you have one, then someone else asks you and you say two, the first person would probably be angry at you; although having two necessarily means that you have one, and you haven’t said anything false, why wouldn’t you just say that you have two in the first place?
  • Grice argued that the assumption of cooperation is governed by four maxims
    • quantity: give the right amount of information
    • quality: give true information
    • manner: use the proper format
    • relevance: give relevant information

53

Sperber and Wilson: Conversational Implicature

Sperber and Wilson: argue that all of the maxims collapse into the maxim of relevance

  • quantity: give the relevant amount of information
  • quality: not speak the truth, but be honest; but honest doesn’t mean speak everything going through your mind, but only what’s relevant to the speaker and listener
  • manner: use the relevant format
  • relevance: give relevant information
  • this theory collapses into your ability to zero in on relevance, i.e. it ultimately relies on your intelligence, reasoning, and problem solving abilities
  • so we see how central intelligence is, however we’ve also seen that it’s not the same as rationality, and we should independently pay attention to, and value, rationality

54

Spearman (1920): General Intelligence Factor

  • Spearman (1920s): showed strong positive manifold between performance on various different tests in various subjects
  • this has been born out by the fact that in psychometric tests of intelligence there’s a strong positive manifold between measures of the individual sub-tests
  • the marshmallow test: put a marshmallow in front of a 4-year-old and tell them, “you can have it whenever you want, but the longer you wait the more chances I will give you another marshmallow” (i.e. delayed gratification)
    • your ability to self-regulate and delay gratification is predictive of your chances of success
  • this lead Spearman to postulate the existence of a general intelligence factor
  • this is abbreviated as g and measures of g have found to be good predictors of a lot of academic and life outcome (see textbook)
  • so the questions emerge: what is g measuring, and is that all there is to intelligence?

Decks in 🚫 PSY100H1: Introduction to Psychology (Winter 2016) with J. Vervaeke Class (50):