2024 Learning [Student] PDF

Recap One of the benefits of categorisation is that it provides a basis for generalisation How do we use what we know about one object or a small set of objects to the entire category of objects There are effects of typicality and family resemblance that influence generalisati...

Recap One of the benefits of categorisation is that it provides a basis for generalisation How do we use what we know about one object or a small set of objects to the entire category of objects There are effects of typicality and family resemblance that influence generalisation Typicality and Family Resemblance Learning generalises more readily when the instances that are learned are typical of the category Dunsmoor & Murphy (2015) Typicality and “Tricket’s disease” is a made up property Generalisation that a category member can have Category Induction - All horses have tasks illustrate key Tricket’s disease principles of how typicality affects - All cows have generalisation Tricket’s disease - All mice have Induction: Tricket’s disease Generalizing from the - All lions The have subjects of each particular to the example can thought of Tricket’s examples disease of a single general category or as Given a set of categories in their own right examples, what is the - All mammals have general conclusion that one could draw Tricket’s disease The subject of the conclusion sentence can be a category or another Category Induction -Robins have a higher potassium concentration in their blood than humans -All birds have a high X potassium concentration in their blood than humans --------------------------- --------------------------- -Penguins have a higher X potassium concentration in their blood than humans -All birds have a high potassium concentration in their blood than humans Osherson et al. (1990) Category Induction: Effect of typical conclusion examples - Robins use serotonin as a neurotransmitter - Bluejays use serotonin as a neurotransmitter XX X - Sparrows use serotonin as a neurotransmitter -------------------------- ------------------------- - - Robins use serotonin as a neurotransmitter XX - Bluejays use serotonin as a neurotransmitter X - Geese use serotonin as a Osherson et al. (1990) neurotransmitter Category Induction: Effect of conclusion category size - Bluejays require Vitamin K for the liver to function - Falcons require Vitamin K for XX the liver to function - All birds require Vitamin K for the liver to function ------------------------------- ------------------------- - Bluejays require Vitamin K XX for the liver to function - Falcons require Vitamin K for the liver to function - All animals require Vitamin K for the liver to function Osherson et al. (1990) Category Induction: Effect of premise example variability - Hippopotamuses have a higher sodium concentration in their blood than humans - Hamsters have a higher sodium X concentration in their blood than humans -All mammals have a higher sodium concentration in their blood than X humans ---------------------------------- ---------------------------------- - Hippopotamuses have a higher sodium concentration in their blood than humans XX - Rhinoceroses have a higher sodium concentration in their blood than humans -All mammals have a higher sodium concentration in their blood than humans Summary Members of natural categories share differing levels of family resemblance Typical instances are verified more rapidly, learned faster, primed more easily, and generalised more readily Generalization is affected by typicality of Instances, Typicality of Category, Category Size, Category Variability Learning PSYC20007 Cognitive Psychology Semester 2 2024 A/Prof Daniel Little Learning Objectives Give examples of learning regularities and invariances in the environment Be able to explain the difference between a simple learning model and an attentional learning model Outline the evidence for the role of attention in learning Describe the difference between experts and novices using examples Describe the stages of insight problem solving Outline the evidence for the role of attention in problem solving What is Learning For? To make predictions about events in an environment and to control them. Learning exists to allow an organism to exploit and benefit from regularities in the environment  Different cues might be used to predict the weather: Will it rain tomorrow? the colour of the clouds over Melbourne the crowding of the trams the number of neighbours washing their cars What cues should an observer attend to?  Moon with halo “Hullo! Going to be rain. Bahloo building a house to keep himself dry. If there is a ring around it, count the stars in the ring; two, three or seven. If there are three stars, there will be rain on the third day” Kerkar Meb 1 & Kerkar Meb 2 by Uncle Segar Passi, senior Meriam elder and a Dauareb man of the Eastern Torres Strait  Causal Learning Seasonal Variation Increased Position Rain of Moon Learning is about regularity and invariance After locating a food source, honeybees will return to the hive and perform a “waggle dance”. The length of the waggle corresponds to distance outside of the colony, while angle of the waggle corresponds to the angle from the sun that the bee must fly Learning is about regularity and invariance This behaviour is not completely innate! Young adult bees start by performing tasks inside the hive but end their lives as foragers. Before they start leaving the hive, they spend around 4 days observing waggle dances Dong, Lin, Nieh & Tan (2023) investigated this experimentally Group 1 Group 2 waggles had higher Allowed to view variability than Group 1 waggle dances normally After 20 days, Group 2 improves Group 2 but still has overly long waggles Prevented from and long flight times viewing waggle After 20 days, Group 1 bees dances attract more watchers than Group 2 Measured first waggles Learning tunes the dance to local and again after 20 regularities days How do we learn? Simple Language Learning Imagine an anthropologist observing a native speaker of an unknown language. The native speaker sees a rabbit and says "Gavagai.“ What does Gavagai mean? William Quine’s (1960) Gavagai problem Simple Language Learning Without further context, it could mean: "Rabbit" Without further learning, it is "Animal" impossible to know which of the references were meant "Furry creature" Quine (1960) termed this “Indeterminancy of reference” "Look, a rabbit!" "There goes dinner!" Simple Learning Example using an Alien Language - Zylian Lifa ve doran With repetition, an association develops between the cue (the words) and the stimulus (the possible referent) Similar to classical conditioning (the pairing of referent and cue is reinforced through repetition) Simple Learning Example using an Alien Language - Zylian Lifa ve doran Simple Learning Example using an Alien Language - Zylian Lifa ve doran Wa nee ee rilan Simple Learning Example using an Alien Language - Zylian Co-occurrences lead to a strengthening between cues and outcomes Is this enough to account for how people learn? Categorisation Demo Follow the QR code or URL to complete a short categorisation demo This is the same demo that was shown in my lecture on categorisation. We will talk about the data on the next few slides. http://go.unimelb.edu.au/m8vi  Categorization Demo Subjects are shown 1 of 8 different stimuli to categorize on each trial Stimuli vary on their Height and the position of the inset vertical Line Kruschke (1993) What determines how stimuli Condition 1: Filtration are divided into categories?  Condition 1: Filtration CATEGORY A CATEGORY B Condition 1: Filtration  Condition 1: Filtration A RY O T EG CA Condition 2: Condensation B RY GO TE CA Condition 1: Filtration Condition 2: Condensation  Simple Learning Model Prediction Data and Predictions A simple model which just learns associations between stimuli and responses predicts no difference between the filtration and condensation conditions. Why? In both conditions, each stimulus is paired with each category the same number of times.  Simple Learning Model Prediction Data and Predictions If we also allow attention to shift to the dimensions of the stimuli that are most relevant or diagnostic, then this model Simple Learning Model + Attention predicts that the filtration condition will be learned faster Why? When we shift attention to diagnostic features, we strengthen the association with the category label only for those features  Simple Learning Model Prediction Data and Predictions Simple Learning Model + Attention Data Simple Learning Example using an Alien Language - Zylian Lifa ve doran Wa nee ee rilan Co-occurrence only – associations get strong with repetition Simple Learning Example using an Alien Language - Zylian Lifa ve doran Wa nee ee rilan Attention model – attended cue associations get strong with repetition Learning involves Attention Blocking paradigm Highlighting paradigm  Blocking Paradigm Reward: Food Early Training Red Light  Food Blocking Paradigm In the blocking paradigm, the mouse first learns that the red light predicts a food reward This is based on Classical Conditioning Food release (US)  Approach Left Food Tray (UR) Red Light (Cue) + Food release (US)  Go Left (UR) Red Light (CS)  Go Left (CR) Then the cues are paired with other cues  Blocking Paradigm Reward: Food Reward: Early Training Juice Red Light  Food Late Training Red Light + Bell  Food Blue Light + Alarm Juice Test Blue Light + Bell ? Juice Blocking Paradigm 1)Learning is not just a reflection of co-occurrence. The bell is paired with the food equally often as the blue light is paired with the juice. If co-occurrence were the sole factor, then we would expect 50:50 responses between the food and juice in the test phase 2)The early pairing of the red light and the food is important. By the late training phase, the red light is already paired with the food. The data suggest that little is learned about the bell because of this. We say that the “bell” was “blocked” by the prior learning about the red light. Attention also provides an explanation of the Blocking Effect Attention is shifted to A when paired with X alone Late in training little attention is left for B Cue D drives the final response Summary An attention account of blocking explains that because attention is shifted to the red light, there is not much attention left over for the bell The Highlighting Effect Food Juice Early Training Red Red Light Light + + Bell Bell  Food Food Late Training Red Red Light Light + + Bell Bell  Food Food Red Red Light Light + + Alarm Alarm  Juice Juice Test Bell Bell + + Alarm Alarm ?? Attention Explains the Highlighting Effect A and B are already paired with X Attention is shifted to D because it alone predicts the unusual event Y Cue D drives the final response Summary An attention account of blocking explains that because attention is shifted to the red light, there is not much attention left over for the bell An attention account of highlighting explains that because cue A is already associated with outcome X, attention is shifted to cue D, which drives the final response Other evidence for the role of attention Experts versus Novices Insight Problem Solving Expertise Represents an extreme form of learning or learning to a high-level of performance Experts have learned highly specific things about specific objects, situations, events What an expert “sees” in a situation will be different from what a novice “sees”.  Experts have different goals Goals determine which features are relevant for determining similarity Focus on perceptual features (like symmetry) Novices Focus on problem features (like adjacent solution steps) Experts Suzuki, Ohnishi & Shigemasu (1992 What trees go together? Different experts categorize things differently Landscaping experts categories trees according to their specific goals Shade trees, fast-growing trees, etc. Taxonomists sort trees into biological kinds Naïve subjects sort trees by their surface appearance Medin, Lynch, Coley & Atran (1997 How are experts different from novices? Different experience and knowledge leads to a difference in how attention is allocated Attention to deep structure versus surface structure Attention to contextually-relevant features Video of Chess Grandmaster recalling grid How are experts different from novices? Different experience and knowledge leads to a difference in how attention is allocated Attention to deep structure versus surface structure Attention to contextually-relevant features Attention to relations rather than individual parts (i.e., chunking) Problem solving Opposite end of the learning spectrum from expertise Do similar mechanisms of attention and knowledge play a role in problem solving and insight?  Duncker’s Radiation Problem Dunker (1945) reported that only 2 of 42 subjects arrived at this solution (5%)  Fortress Defense Problem How can you storm this castle, which is surrounded by a minefield? Sending too many people on one path will result in detonation of the mines and loss of your knights. If participants heard the story of the fortress first, then were given the radiation problem, the success rate on the radiation problem increased to 30% If instructed to “think about the fortress story”, the success rate increased to 70% A hint about where to attend improved problem solving  Attention to the most valid features allows for mapping between problem elements Surface Details Attention to irrelevant surface details can mask the underlying similarity Insight Problem Solving Imagine you have on a table in front of you, a candle, a box of tacks, and a book of matches. How can you mount the candle on the wall using only what’s on the table. Duncker (1945) Stages of Problem Solving Preparation Search for a solution using logic & reasoning If a solution is found, stop here Incubation Attention not devoted to the problem Illumination, Insight, AHA!! “Spontaneous” manifestation of the problem solution into consciousness Verification Use of logic and reasoning to confirm the solution If you invalidate the solution, return to the preparation stage Wallas (1926) Incubation Is it real? Out of 39 experiments, incubation improved problem solving about 75% of the time Dodds, Ward & Smith (2010) Different factors influence whether incubation is successful or not Incubation Longer periods of incubation are positively correlated with success For incubation periods less than 1 hour, 30 minutes is optimal For incubation periods over 1 hour, the longer the better Beck (1979), Forgosi & Guildford (1972 Incubation Better preparation increases the effectiveness of incubation Silveira (1971) had people solve a difficult problem in 3 different conditions Work continuously for 35 minutes Work for 3 minutes, incubation, work for 32 minutes Work for 13 minutes, incubation, work for 22 minutes More people solved the problem in the last condition than in the first two Theories of Incubation Better preparation and hints help insight problem solving Why? Recovery from Fatigue Preparation is cognitively draining Incubation allows for recuperation of cognitive abilities Forgetting of mental sets False assumptions can block the path to the solution Mental Sets Einstellung Effect – when an idea that comes to mind in a familiar context prevents alternatives being considered Luchins (1942) Related to the idea of negative transfer and “strong-but-wrong” errors Einstellung: “strong-and-right-but-not-the- best” Luchin’s Water Jug Experiment 127 21 3 How can you make 100 litres of water? Fill jug A, then subtract jug B once, then subtract jug C twice Einstellung Effect Not all learning is beneficial (can lock in patterns of thinking) Einstellung Effect: Existing knowledge or habitual ways of thinking influence our problem-solving After solving problems in which Solution 1 worked, participants were less likely to use the easier Solution 2 compared to a control condition And more likely to get stuck if Solution 1 no longer worked Crucial moves for the longer strategy are in green Crucial moves for the shorter strategy are in red Longer fixations are made to the locations associated with the familiar strategy Bilalic, McLoed & Gobet (2010 Crucial moves for the longer strategy are in green, but now the longer strategy doesn’t work Crucial moves for the shorter strategy are in red Expert chess players waste a lot of time looking for a familiar strategy that won’t work Bilalic, McLoed & Gobet (2010 Insight Not just another problem solving step Sudden jump or transition in understanding Not “knowing”  “knowing” Pols (2002) Theories of Incubation & Insight Unconscious Work Solution to the problem is developed unconsciously and “delivered” to consciousness once a goal is reached Difficult to assess experimentally Conscious Work Work on the problem takes place while attending to non-taxing activities (in the shower, driving) Then attention is shifted quickly to the problem and any activity is forgotten, only the end state is remembered leading to a feeling of “suddenness” Smith & Dodds (1999) Is insight really sudden? Eye Movements Predict Insight Attention, Learning & Problem Solving Knowing what aspects of a problem to attend to is critical to problem solving Learning and expertise illustrate the role of attention in the development of concepts Is attention everything? Betsy and Fast Mapping Fast Mapping: Inference by Exclusion Lifa ve doran Fast Mapping: Inference by Exclusion Lifa ve doran Wa Lifa ve rilan Summary Learning and Insight benefit from attention Learning also draws on prior knowledge to guide current learning and inference Prior knowledge can be useful when that knowledge aligns with the current problem Can also be unhelpful in the case of the Einstellung effect