×
Log in to StudySoup
Get Full Access to Brown U - Study Guide - Midterm
Join StudySoup for FREE
Get Full Access to Brown U - Study Guide - Midterm

Already have an account? Login here
×
Reset your password

BROWN U / OTHER / CLPS 100 / Are some cs-us combinations learned faster than other combinations?

Are some cs-us combinations learned faster than other combinations?

Are some cs-us combinations learned faster than other combinations?

Description

School: Brown University
Department: OTHER
Course: Learning and Conditioning
Professor: Professor
Term: Spring 2020
Tags: rescorla-wagner, Instrumental Methods, and operant conditioning
Cost: 50
Name: CLPS 100 Midterm #2 study guide
Description: This study guide covers the material that will be in exam #2. It includes the Rescorla-Wagner model, Instrumental Conditioning, Matching Law, Delay Discounting, and more.
Uploaded: 03/28/2020
6 Pages 107 Views 8 Unlocks
Reviews


CLPS 100 Study Guide: Exam #2 


Are some cs-us combinations learned faster than other combinations?



The Rescorla-Wagner Model 

● ΔV β(λ ) cs = α − V SUM 

○ ΔV cs: change in the associative strength of the CS/US link

○ α : the rate parameter for the CS → more salient = faster learning about CS

■ Always between 0 and 1

○ β : the rate parameter for the US → more salient = faster learning about US

■ Always between 0 and 1

○ λ : the maximum learning that the US will support


Does the us affect which part of the compound cs they are learning from?



We also discuss several other topics like What is the medication of deviance?

■ No constraint; can be any number

■ Larger values for more intense US We also discuss several other topics like How do older subjects compensate?
If you want to learn more check out What do parasite need to grow and reproduce?

○ V : the sum of the associative strengths of all the CS present in the trial SUM 

■ This enables the model to predict blocking & conditioned inhibition

○ λ − V : expected outcome versus actual outcome SUM 

■ Indicates “error” or “surprise”

■ Larger value → more surprise → more learning 

■ Negatively accelerated learning curve: increments in learning decreases as trials progress


Does the intensity of the reinforcer affect which element of a compound stimulus controls the response?



● Features of the R-W model

○ US must be surprising in order for learning to occur

○ CS parameters (rates of learning) are fixed → salience of the stimulus doesn’t change ○ Associative strength of CS can only change if CS is present → we can only learn about what is present in a trial If you want to learn more check out How stimulus generalizing and discrimination affect learning?

○ All CSs move in the same direction on a given trial Don't forget about the age old question of How should the researcher show the maximal change in action potential firing over time?

■ Positive discrepancy → all CSs will increment

■ Negative discrepancy → all CSs will decrement

● Novel predictions brought about by the RW model

○ Role of contingency effects 

■ Positive contingency: P(US | CS) > P(US | no CS) → excitation & conditioning ■ Zero contingency: P(US | CS) = P(US |no CS) → no conditioning

■ Negative contingency: P(US | no CS) > P(US | CS) → inhibition

■ Experiments to support this novel prediction: Rescorla (1968) & Durlach (1983) ○ Overexpectation 

■ Two stimuli (A & B) are conditioned separately to predict the same US

■ Since both predict the same US, presenting them together creates an

overexpectation of the US (animals will predict a larger US, e.g. more food pellets) ■ When the animal receives the regular (not larger) US, the associative strength (V A and V ) of both CS (A & B) decrease B If you want to learn more check out What consumers consider business success?

■ Experiment to support this novel prediction: Kremer (1978)

○ Extinction of conditioned inhibition → thisisthe only prediction that turns out to be false! ■ A conditioned inhibitor will extinguish when presented alone

■ During non-reinforced presentations of the CI alone, it will generate an increment in its associative strength bc the animal was not expecting anything anyways

■ The associative value of a CI starts off negative but when it’s presented on its own, a positive association begins to form → FALSE!

■ However, it does protect a CS from extinction: if a CS is presented with a CI and there is no US presented, the associative strength of the CS will NOT decrease

because the organism attributes the absence of the US to the CI, not the CS

Foundations of Instrumental Conditioning 

● Thorndike’s Law of Effect

○ If a response is followed by a satisfying event, the association between the stimulus (S) present and the response (R) is strengthened

○ If a response is followed by an annoying event, the association between the stimulus (S) present and the response (R) is weakened

● They are not learning about the consequences of the response, they are learning because of the consequences

○ The consequence (reward/punishment) is not part of the learning, but it is important for the learning process

○ S-R relationship not S-O

● Relationships between Behavior and Consequences

○ Positive reinforcement: presenting an appetitive stimulus when a behavior is performed ■ Cognitive Psych label: positive reinforcement

○ Negative Reinforcement: removing an aversive stimulus when a behavior is performed ■ Cognitive Psych label: escape/avoidance

■ Constructional Aggression Treatment (CAT): we want to reinforce a relaxed behavior in an organism. On each trial, we reward the animal by backing away if they do not respond. On each successive trial, we move the threshold closer.

○ Positive Punishment: presenting an aversive stimulus when a behavior is performed ■ Cognitive Psych label: punishment

○ Negative Punishment: removing an appetitive stimulus when a behavior is performed ■ Cognitive Psych label: omission training

○ Reinforcement is used when we want to increase a behavior, punishment is used when we want to decrease a behavior

○ Reinforcement is generally more useful than punishment

● Shaping: systematically and differentially reinforcing successive approximations to a terminal behavior

○ Step 1: pair the stimulus with the treat

○ Step 2: change the criteria needed to receive the treat → get the behaviors closer and closer to the desired final behavior

○ Step 3: we will see early behaviors begin to disappear in favor of the more advanced behaviors that are closer to the desired behavior

Learning vs Performance 

● Thorndike: reinforcement is crucial for learning the relationship between the S-R ● Tolman: reinforcement encourages you to use your knowledge to perform a behavior, but it is not crucial to learn

○ Latent learning experiment: showed that rats could learn about a maze without being reinforced

○ Animals are capable of learning without being reinforced; they are encouraged to perform when there is some type of reinforcement present

● Hull: believed the learning that happens is S-R but in order to get a behavior to be expressed you have to motivate your subjects

○ Reinforcement is drive reduction

○ E = D x H

■ E = performance

■ D = drive

■ H = habit

● Tolman believed rats learn a cognitive map, Hull believed rats learned an S-R relationship (motor actions)

○ T maze experiment carried out to test both hypotheses

■ A reinforcer is placed on the right side of the “T” maze

■ After some trials, the rats are put on the opposite side of the maze

■ Hull: If they are simply learning motor actions they should still turn right

■ Tolman: if they are learning a cognitive map, they will use the visual cues to arrive at the correct location

○ Turns out rats can learn in multiple ways: responses & visual cues

Early Challenges to S-R Theory 

● Tinklepaugh (1928, 1932) → expectancy violation

○ Animals have expectations: they don’t like to expect something good & receive something worse

○ E.g.: Monkeys shown banana but given lettuce → do not savor the lettuce as much as if they were originally shown the lettuce in the first place

● Crespi (1942) → contrast effects

○ Elation effect: increases in reward results in increases in behavior

○ Depression effect: decreases in reward results in decreases in behavior

○ E.g.: varied the # of pellets each rat received in phase 1, then in phase 2 all rats received the same amount

■ Those who had previously received more pellets in phase 1 ran slower in phase 2 ■ Those who had previously received less pellets in phase 1 ran faster in phase 2 ● Colwill & Rescorla (1985) → learning the relationship between behavior & reward ○ Animals are not simply learning habits, they learn which behaviors produce which rewards ○ E.g.: rats given 2 outcomes for 2 different responses (lever press=food & pulling chain=sugar water); then one outcome is made less attractive (sugar water → sick) ○ When given a choice, animals will perform the behavior that results in the attractive outcome (press lever for food because they know the chain=water and water=sick) ○ Animals learn the relationship between behavior and outcomes

Selective Pavlovian Conditioning 

● Are some CS-US combinations learned faster than other combinations?

○ Organisms are biologically predisposed to learn some things faster than others ○ It is easier to use the CS as a signal for one class of USs than another class of USs ● Experiment by Garcia & Koelling (1966)

Phase 1

Phase 2 (test)

Group 1: Sacc + LN → illness

½ Sacc and ½ LN + Water

Group 2: Sacc + LN → shock

½ Sacc and ½ LN + Water

● Does the US affect which part of the compound CS they are learning from?

○ Group 1: associates the flavor (instead of the light/noise) with the illness

○ Group 2: associates the light/noise (instead of the flavor) with the shock

● Organisms are genetically wired to learn some associations quicker than others

Selective Instrumental Conditioning 

● Does the intensity of the reinforcer affect which element of a compound stimulus controls the response?

● Experiment by LoLordo (1973)

○ Stimuli: visual (red light) and auditory (tone); response: lever press

○ Appetitive Group: positive reinforcement → pressing the lever gives them food ■ Predict: they will associate the light (sight) with the food → more lever press when the light is present

○ Avoidance Group: negative reinforcement → pressing the lever avoids a shock ■ Predict: they will associate the tone (sound) with the shock → more lever press when the tone is present

● Test trial: present the light & the tone by themselves

○ Appetitive Group: learned the association between the light & the food

■ Respond more to the light

○ Avoidance Group: learned the association between the tone & the shock

■ Responds more to the tone

● Conclusion: there is selective learning → one stimulus is a better reinforcer for a stimulus than another

○ Animals can learn about some CS-US relationships faster than others

Selective Learning Hypotheses 

● Selective Sensitization Hypothesis

○ Some USs increase weariness about some CSs relative to others

■ E.g.: when they get sick → augmented responses to taste

● Head-Start Hypothesis

○ Some CSs enter with pre-existing connections to some USs; biological preparedness ■ E.g.: you already have an association between sickness & taste before any training is done

● Accelerated Learning Hypothesis

○ There are no pre-existing connections between CSs and USs, but you learn some associations faster than others

○ The increments in associative strength are larger for some CS-US associations ● Rescorla’s (2008) Experiment to test these hypotheses

T1

Test (t2)

Group Related

sucrose water → injection (made sick) Clicker/tone → shock

Sucrose + clicker/tone

Group Non-related

sucrose water → shock

clicker/tone → injection (made sick)

Sucrose + clicker/tone

● Selective sensitization hypothesis: We shouldn’t see a difference between the two groups because they are equally sensitized.

● Head-start hypothesis: We should see no difference between the groups because we are presenting a compound stimulus during the test phase.

● Acceleration hypothesis: we should see a difference. Related group should have more learning than unrelated group.

○ They should have learned more about sugar/illness and clicker/tone/shock than about sugar/shock and clicker/tone/illness.

● Test: As they are drinking the water they are also hearing the clicker/tone

○ Related group: drinks a lot less of the sugar water than the unrelated group ○ This supports the third hypothesis (accelerated) → good evidence for selective learning

The Matching Law 

● If there are other reinforcers present, they will reduce the rate of response to the stimulus ● Increasing the reinforcement to stimulus A relative to stimulus B will increase responding to stimulus A relative to stimulus B

RA 

● = RA+RB 

rA 

r +r A b 

○ RA = rate of response of A

○ rA = payoff (reinforcement) rate for response A

● Implications

○ Rate of responding depends on its own reinforcement & and reinforcement of other behaviors

Delay Discounting 

● Most people will choose a smaller reward now rather than waiting for a larger reward later ○ The present value of a reward declines as time passes

● v = V (1 + kD) 

○ v→ objective value of the reward

○ V → subjective value

○ k→ the degree of discounting

■ Small k: very little discounting; delayed reinforcers are still highly valued ■ Large k: a lot of discounting; delayed reinforcers are not valued much

○ D → delay in the reward

Zoo Training - Pamela Jones Talk 

● What type of training is used on zoo animals?

○ Classical conditioning

■ Teach animals to understand cause & effect between two items

○ Operant conditioning

■ Mostly use positive reinforcement

■ Shaping behaviors

■ Teach animals to operate on the environment

○ Habituation

■ Getting them used to being around people → petting

● Positive training

○ Marking behaviors with a clicker & reinforcing it with a treat

● SPIDER Model

○ S- setting goals → what is the behavior you want to train them to do?

○ P- planning → create a shaping plan by breaking behaviors down to manageable steps ○ I - implementing → make adjustments as you move along with the training plan

○ D - documenting → not what is happening during the training

○ E - evaluating → what could have been done differently?

○ R - readjusting → readjust the parts of the plan that didn’t work out ● Reasons for training

○ Assist with the husbandry of the animals

○ Assists with the medical care of the animals

○ Offers an opportunity for the animals to have choice and control in their daily lives ○ Physical stimulation

○ Enhance visitor experience

Page Expired
5off
It looks like your free minutes have expired! Lucky for you we have all the content you need, just sign up here