Summary Learning Lecture notes PDF

Preview

Summary

These are summary notes for the learning section for this course...

Description

-

Learning Notes: Historical Antecedents and Dawn of the Modern Era: *Cartesian mind-body dualism: -Voluntary (conscious intent)—> Study of the mind (Empiricism, Ebbinghaus). -Involuntary (reflex behaviour)—> Study of reflexes (Reflexology, Pavlov). -Both lead to Primary rules of association (contiguity, contingency, similarity).

1. Contiguity= “Closeness”, they go together, 'close' together in time/space, etc. -We learn things better that are close together (two words next to each other). 2. Contingency= A predicted (dependent) relationship, when one occurs, another one always occurs. -E.g rain is contingent upon clouds.

3. Similarity= Remembering things that are similar to each other. -E.g two words that start with the same letter. *Learning: -Facilitates adaption to the environment. -Critical to everyday normal functioning. -Learning forms include declarative, episodic, implicit and explicit learning. -Instrumental conditioning= learning of relationships between behaviour and its consequences. -Classical conditioning= learning of relationships between events and the learning of emotional reactions to events in the environment (analyse behaviour in terms of the stimuli that control it). -30% of meaning comes from gestures in learning. *Associative Learning= Behaviour is controlled by stimuli (classical conditioning), and can change because of its consequences (instrumental conditioning). *"Learning is an enduring change in the mechanisms of behaviour involving specific stimuli and/or responses that results from prior experience with those or similar stimuli and responses”-Domjan, 2010. *Performance Distinction:

1. Performance: all the actions (behaviours) of an organism at a given time. 2. Learning: Measured by detecting a change in performance (behaviour), but also a change in the mechanisms of behaviour, -Other sources of changes in performance include fatigue, change in stimulus conditions, physiological or motivational state, and maturation.

-

*Learning: Levels of Analysis:

*We study learning by Methodology: -General process approach (principles that dictate learning). -Requires an experimental analysis (need to control for extraneous variables). *Learning is not possible without memory: studies of learning and memory typically involve 3 phases:

1. Acquisition (learning/training phase: studies of learning manipulate the conditions of acquisition). 2. Retention Interval (delay between training and testing). 3. Retrieval (testing phase). -Studies of memory focus manipulate the retention interval and the conditions of retrieval (e.g. time limit, how many times item is tested/viewed, when it is tested, etc). *Simple Forms of Learning: -Effects of Repeated Exposure. -Reflex (S-R: stimulus-response) pathway. *Habituation: -A decrease in the strength of response following repeated presentations of the eliciting stimulus. -The Startle Response used with mice in experiments. -To measure habituation in experiment= repeated startle stimulus in the presence of 60dB background noise (blocks of 10 loud startle tones). *Sensitisation: -An increase in the strength of response following repeated presentations of the eliciting stimulus (weak, medium, strong shock to mice in startle response experiment). -To measure sensitisation in experiment= repeated startle stimulus in the presence of 80dB background noise. *Dual process theory of habituation and sensitisation: -Distinct neural processes underpin increases (sensitisation (state (state of arousal)) process) and decreases (habituation (S-R) process) in responsiveness.

-

-Both are activated simultaneously. -Observed behaviour= sum of both (net outcome of the two processes). -Assumes that repeated presentation of an event activates a stimulus-response (habituation) and a state (sensitisation) pathway. -Habituation and Sensitisation are important determinants of human and non human behaviour. -Knowledge of these processes is crucial to understand control procedures in associative learning. *Associative Learning- Instrumental Conditioning (Thorndike): 1. Thorndike’s Law of Effect= behaviours that lead to a satisfactory state of affairs are ‘stamped in’, while behaviours leading to an unsatisfactory or annoying state of affairs are ‘stamped out’. -Instrumental conditioning is a process in which a behaviour becomes associated with its consequences, and as a result, the frequency of the behaviour changes (increases or decreases).

2. The Skinner Box (Skinner)= Rat presses lever (behaviour) to produce food (consequence)…therefore frequency of lever pressing increase (change in probability). Associations between behaviour and it’s consequences are learnt. -Reinforcement= (food) behaviour increases. -Punishment= (shock) behaviour decreases. *Schedules of Reinforcement: -“A schedule of reinforcement is a rule that determines which occurrence of a response is followed by a reinforcer”-Domjan, 2010.

1. Continuous=reinforcement is delivered after every response. 2. Intermittent (partial)= ratio schedules (fixed, variables), and interval schedules (fixed, variable). -Reinforcement depends only on the number of responses an animal has performed. -Fixed Ratio (FR)= a fixed number of responses. -Variable Ration (VR)= an average number of responses. *Interval Schedules: -Responses are reinforced only if the responses occur after a certain mount of time has passed. *Steady State Behaviour is a stable pattern of responding that is established once an animal has had sufficient experience with a given schedule of reinforcement.

-

*Shaping is a process in which successive approximations of a behaviour are reinforced. *Instrumental Conditioning and Shaping: -Are important determinants of human and non human behaviour (education, law enforcement, etc). -The instrumentally conditioned, shaped behaviour, bar pressing, is often used as a measure of classical conditioning. *Instrumental Conditioning: -Positive/Negative= used to define whether the consequence of the behaviour results in the presentation/removal of something. -Reinforcement/Punishment= used to define whether the probability of the behaviour subsequently increases/decreases.

1. Operant Behaviour. 2. Consequences of the behaviour. 3. Change in the probability of that behaviour. *Reinforcement and Punishment: -Positive reinforcement = behaviour strengthened by a positive outcome (e.g. praise is a positive reinforcer of saying please). -Negative reinforcement = behaviour strengthened by avoidance of a negative outcome (e.g. noise termination is a negative reinforcer of calling the police). -Positive Punishment = behaviour weakened by a contingent negative outcome (e.g. criticism is a positive punisher of swearing).

-

Negative Punishment =behaviour weakened by the removal of a positive outcome (e.g. time out is a negative punisher of swearing). -Extinction= cessation of behaviour when the link to the positive of negative condition is removed. -Shaping= Is a stage of the goal, to progressively reach it. *Positive reinforcement and Punishment: -The contingency between operant response and outcome is positive, (the production of the response results in the production of the outcome). -Positive reinforcement outcome is appetitive (e.g. bar press results in food delivery). -Increase in the probability of the operant response: increase of response rate. -Punishment outcome is aversive (e.g. cookie jar). -Decrease in the probability of the operant response: decrease of response rate. *Negative reinforcement: -The contingency between operant response and outcome is negative, (production of the response results in the absence of the outcome). -Negative reinforcement outcome is aversive (e.g. avoidance learning). -Increase in the probability of the operant response: increase of response rate. *Omission training: -Also called differential reinforcement procedures (DRO). -The contingency between operant response and outcome is negative, (production of the response results in the absence of the outcome). -Omission training outcome is appetitive. -Decrease in the probability of the operant response: decrease of response rate. *Simple schedules of intermittent reinforcement include: 1. Ratio schedules: -Fixed Ratio (FR): the number of response required is predictable. -Continuous reinforcement or CRF is also a fixed ratio schedule. -Measure of response includes cumulative and recorder. -On a CRF schedule, pigeons will peck for food availability steadily at first, and will slow down as it gets full. -Steady rate of response, pause after reinforcement. (diff to random pause in CRF). -Variable Ratio: the number of response required is unpredictable.

-

-Steady rate of responses, and extremely rare pause in responding. -A steadier pattern is observed with a variable ratio. 2. Interval schedules: -Responses are reinforced only if the responses occur after a certain interval. (The interval determines when the reinforcer is available, NOT when it is delivered). -Fixed Interval (FI): -Performance on a FI schedule reflects the subject accuracy in telling the time. -The post reinforcement pause followed by an acceleration of response towards the end of the interval reflects a rudimentary ability to tell time. -Variable Interval: Steady state if responding, and no regular pause. *Classical Conditioning (Pavlov)= a process in which one stimulus that does not elicit a response is associated with a second stimulus that does, and as a result, the first stimulus also come to elicit a response. *Classical Conditioning Experimental Paradigms: 1. Appetitive Conditioning: -Conditioned response increases with increased trials.

2. Fear Conditioning: -Shock in rat experiment for food. -Suppression Ratio=CS responding/ (CS responding + pre-CS responding). *Fear conditioning outside the lab involves: 1. Phobias: -Intense, irrational fears of certain objects or situations through avoidance. -Retrospective recall leads to traumatic conditioning event. -Vicarious conditioning leads to social stimuli acts as USs. -Genetic vulnerability (personality) leads to individual differences in susceptibility. -Prior experience leads to immunisation against conditioning. -Control over conditioning event leads to reduction in conditioning effect.

2. Panic disorder: -Ongoing elevated anxiety associated with having an attack.

-

-Exteroceptive conditioning followed by interoceptive conditioning leads to generalisation to novel situations, which leads to avoidance behaviour.

3. Interethnic anxiety: -Anxiety is the most commonly diagnosed mental disorder with 29% of the US population diagnosed.

4. Applications of fear conditioning in non-humans. 3. Eyeblink Conditioning: -US= puff of air/electrical stimulation. -CS= tone. -UR= eye blink to puff/electrical stimulation. -CR= eye blink to tone.

4. Learning what tastes good and what tastes bad: -Conditioning of taste preferences (CS= novel taste, US= sucrose, UR= liking evoked by sucrose, CR=increase liking of novel taste). -Conditioning of taste aversions (CS= novel taste, US= nauseating substance (radiation), UR= gesture-intestinal illness, CR= nausea evoked by novel taste). -This explains loss of appetite during cancer treatments. *Acquisition is a process of developing and strengthening a conditioned response through repeated pairings of a CS and a US. -Acquisition of a response to CS could be caused by sensitisation given the repeated exposure to the US (pseudoconditioning). *Under which conditions does Classical Conditioning occur?: -Contingency hypothesis consistent with observed data. -Organisms acquire a response to the CS if it reliably predicts the occurrence of the US. -Conditioning occurs if the CS provides info about the upcoming US..this is termed the info value of the CS. *Pseudoconditioning= acquisition of a response to CS could be caused by sensitisation given the repeated exposure to the US. -Paired-experimental vs unpaired control comparison controls for effects of nonassociative learning.

-

*Common Properties of Classical Conditioning: -Nature of the CR -Second order conditioning -Latent inhibition -CS/US magnitude effect. *Is the CR always identical to the UR?: NO! -Before conditioning= Morphine US—> pain tolerance UR & Tone CS—> no change in pain threshold. -Conditioning= Tone CS: Morphine US. -After conditioning= Tone CS—> pain sensitivity CR & The CR is opposite to the UR. *Physiological homeostasis: -Constant maintenance of body parameters within certain limits (e.g. body temp). -Disturbance (physiological challenges) evoke a reflexive compensatory response (e.g. vasoconstriction to compensate for drop in temp). *The Role of Classical Conditioning in Drug Tolerance: -Drugs of addition elicit 2 processes:

1. A-Process= euphoria. 2. B-Process=compensatory response/disphoria. -Most often, the B-Process becomes the CR and the CR is opposite to the UR. -Learning about conditioned stimuli that signal oncoming disturbances allows for disturbance effects to be minimised. -You need more and more drugs to get a euphoria experience. -Environmental factors become conditioned stimuli and have a contingent relationship with the drug. -Through classical conditioning, conditioned stimulus comes to trigger a conditioned compensatory response (=>CR opposite to UR)…these processes lead to drug tolerance. -E.g. Morphine: -Morphine has 2 effects:

1. A-Process= pain tolerance (threshold up). 2. B-Process=pain sensitivity (threshold down).

-Tone: morphine—> UR (A and B processes). *Second Order Conditioning (2 phases of conditioning): -Phase 1= Light: Food (light, followed by food-Pavlov’s Dog). -Phase 2= Tone: Light -Test= Tone (interested to see if salivation occurs to tone, which produced light). -Result (CR)= In theory, dog should salivate to tone, therefore Tone evokes salivation. -Tone evokes salivation as tone gained control over salivation response without ever being paired with food. *Latent Inhibition (CS pre-exposure): -Phase 1=Light -Phase 2= Light: Food -Test= Light. -Result (CR)= Light evokes little to no salivation. -CS pre-expsoure inhibits or delays subsequent learning of an association between the familiar CS and a US. *Manipulate US magnitude by giving dog a bland dog food, and then a really tasty one. -Learn faster (acquisition goes faster). *Manipulate CS magnitude by ringing bell really quietly, then very loudly/ or have dim light and then a very bright light. -Learn (acquire) CR faster, but final CR will be the same. -Determined by the US. *SUMMARY: Empirical Findings:

1. In classical conditioning, animals acquire a response to an initially neutral stimulus after it has been paired with a stimulus of biological significance. 2. Learning occurs when the US is contingent upon the occurrence of the CS, i.e. the CS signals or predicts the onset of the US. e.g. if drug is associated with contextual cues, we will relate to those cues. 3. Classical conditioning has many distinctive characteristics, aka properties (acquisition, extinction, compound conditioning, CS pre-exposure, CS & US magnitudes etc). *Rescorla-Wagner Model: -An underlying mechanism as to why such learning occurs (theoretical explanation). -Most influential theory of learning which generated 40 yrs of empirical research. -Elegant, simple and powerful ability to predict many learning phenomena.

-

*Blocking (Kamin): -Blocking has 2 groups and 2 phases. Group 1 experimental: -Phase 1=Tone:Shock -Phase 2= Tone_Light:Shock -Test= Light. -Result (CR)= Light evokes no fear. Group 2 Control -Phase 1=Light -Phase 2= Tone_Light:Shock -Test= Light. -Result (CR)= Light evokes fear. -The occurrence of blocking provides additional evidence that continuity isn’t sufficient to produce learning. -Group 1 doesn’t learn, Group 2 does learn. *Kamin’s Idea: -Only unexpected events have the ability to activate learning. -Once the US no longer startles the animal, it no longer stimulates the mental effort required for the formation of an association. -To be effective, a US must be unexpected. -Surprising of CS is central to classical conditioning as it motivates us to find CS’s that will predict the shock (surprising factor). *Rescorla-Wagner Model Continued: -Incorporated Kamin’s idea into a mathematical model of learning (the US is surprising to the organism, & the unexpected US activates the learning process). -Accurately predicts and explains a large collection of conditioning phenomena but fails to explain latent inhibition. -New aspects of the model include: 1. The discrepancy between the experienced US and the expected US strengthens the associative link between the CS and the US. -Prediction Error drives learning, and as you learn the US, the prediction error goes to 0. 2. Strengthening of the associative link between CS and US over successive cooccurences reduces the suprisingness of the US. 3. When the US is no longer surprising because it is fully predicted by the CS, no further learning occurs.

*Rescorla-Wagner Mathematical Model (1972) = ΔVi+1=αβ(λ-Vi) Vi= associative strength at the end of previous pairing (previous trial). Vi+1=associative strength after the current pairing. λ=maximum associative strength (US provokes a certain amount of associative strength). α=salience of the CS (varies from 0-1 e.g. dim light alpha close to 0). β=salience of the US (varies from 0-1 e.g. tasty dog food beta close to 1). ΔVi+1=K(λ-Vi) = (associative strength vs conditioning trials). K=salience of the CS and the US; varies from 0-1. *Rescorla-Wagner Model states that the changes in the associative strength of the CS during a trail will depend directly on (a) the salience of the CS (alpha), (b) the strength of the US (beta), and (c) the difference between the maximum associative strength value the US can support (λ) and its current value (Vi), (λ-Vi). -λ-Vi is known as the ‘prediction error’ or the ‘surprisingness’ of the US: difference between the expected and observed values. -The prediction error describes how much bigger the US is than what was predicted by the CS. -The change in associative strength on a given trial is proportional to the prediction error. -When there is no more prediction error, there is no more conditioning. -When a graph shows a line that increases fast and then starts to become horizontal, we know that prediction error is going down. *Acquisition: -CS is initially neutral =>Vo=0% (as there is no previous trial for trial 1), with moderate salience =>α=0.5%. -Salience of the US assumed maximum =>β=1. -λ is set to 100 (CR scaled as a percentage) => λ=100%. -Therefore initial parameters= Vo=0%, α=0.5%, β=1, and λ=100%. -Therefore equation to plot graph for 1st trial= ΔV1 (trial 1)= 1 x 0.5 (100-0)= 50% =V1= Vo + ΔV1= 50 + 0 = 50% -V now equals 50 as it is equivalent to the associative strength of the previous trial. -Therefore equation to plot graph for 2nd trial= ΔV2 (trial 2)= 1 x 0.5 (100-50)= 25% =V2= V1 + ΔV2= 50 + 25= 75%. -Then for trial 3, Vi will equal 75.

-R &W predict that acquisition should progress as normal during phase 2. *Rescorla & Wagner’s Additional Insight into Blocking: -Changes of associative strength of a given CS on a given trial depend not only on the current associative strength of that CS, but on the associative strengths of all CS present on that trial.

-R-W incorporate this idea mathematically by replacing V with ΣV (“sum of Vs”)= ΔVA=α(A)β(λ-ΣV). - ΔV(A) is the change in associative strength of a CS(A). -When multiple CS’s are present, ΣV is the sum of the associative strengths of all the CS’s present on that trial. -ΣV represents the associative strength of all the stimuli (e.g. both light and tone) that occur just before the US and also represents the degree to which the US is predicted by all the preceding stimuli. -Therefore compound conditioning initial parameters= VT0=0%; VL0=0%; αT= αL =0.2; β=1; λ=100%. *Associative Strength: -The concept of association is central to the RW model and all associative models (including attentiona...