Week 5 - Learning from Errors

Question 1

Differentiate between high and low stakes errors (learning from errors)

Answer 1

• High stakes is where no-one wants mistakes if there are detrimental consequences
• Low stakes is where the consequences don’t really matter.

Question 2

What is the traditional view of errors?

Answer 2

• Errors are harmful
• Avoiding errors during learning are to avoid them in a high-stakes test (Skinner and Bandura)
• Errors will become ‘stamped’ in –> more likely to be repeated
• This led to errorless learning approaches: step by step guidance, reinforce correct answers and ignore and avoid errors.

Question 3

What is an errorless learning approach?

Answer 3

• An approached used in clinical and neurological rehabilitation that provides a structured, scaffolded reward and practice - whilst rewarding correct answers. It avoids opportunities to make and rehearse mistakes.

Question 4

How does proactive interference interact with errors?

Answer 4

• Proactive interference: old memories interfering with recall of new information
• Errors, in this case, may interfere with memory for the correct answer.
• So.. therefore, errors are expected to be repeated on the final test.

Question 5

What is an alternate perspective against the errorless learning view on errors?

Answer 5

• Engaging with errors promotes active, deeper learning - as feedback is crucial for learning from errors

Question 6

What is the Generation Effect in errors?

Answer 6

• A phenomenon in which people typically remember information they have generated themselves better than information that has been presented or generated for them.

Question 7

What did Kane and Anderson do in investigating the Generation Effect in error?

Answer 7

• Participants gave sentences that were either determined (predictable answer) or undetermined (could be any answer) - and allowed them to make errors for the second.
• They then were given feedback, told what the correct answer was for the undetermined one and then tested their memory later on.
• They either were tested in a recall or recognition test
• Findings:
• Generation of the answer led to better later recall of the correct answers than reading intact sentences
• The same pattern appeared for undetermined sentences, where generations were almost always wrong at encoding.
• Even incorrect generation didn’t harm learning.

Question 8

How can you apply the generation effect to the real world?

Answer 8

• Try to generate answers yourself, rather than rereading notes
• Don’t worry about mistakes - they are not harmful for learning

Question 9

What is the Pretesting Effect?

Answer 9

The finding that attempting to answer questions about a topic before studying it improves memory of that topic – even when your initial answers are incorrect.

Question 10

Kornell looked at the Pretesting Effect in errors and learning - what did he do and find?

Answer 10

• Kornell submitted participants to a study phase where they studied weakly related words (pond–frog), followed by a cued recall test.
• There were 2 conditions for the study phase: read-only condition where they read word pairs intact (no generation) to study. The second condition: test condition they were presented with a word, and were asked to guess what the second word would be (1st word = cue, 2nd word = target).
  Researchers only look at the cases where the participant got the answer wrong and received corrective feedback.
• The cued recall test had the participants having to recall the target words.
• Findings:
• Across all experiments, taking a test - generating an error and receiving corrective feedback led to consistently better cued recall performance.

Question 11

Search Set Theory is used to explain how errors lead to improved memory - what is it? (part of pre-testing effect)

Answer 11

• Theory proposes that when you’re in a test trial and asked to guess what the target word may be - you activate related concepts through producing a guess.
• Partial activation of the correct answer (frog) makes it easier to encode when revealed
• Prediction: Pretesting benefits occur mainly for related word pairs, where the target is part of the activated search set

Question 12

‘Attentional accounts’ is a different theoretical explanation for the pre-testing effect - explain this and a study on it

Answer 12

• Pretesting enhances processing of the correct target word, when it is subsequently revealed
• So, when a guess is generated - it can make you more motivated or curious to learn the correct answer.
• Seabrooke et al
• Participants viewed faces and learned four facts associated with each face
• There were 2 trials: generate trials: participants guessed 2 facts before seeing correct answers; and studied 2 facts. Read trials: saw all 4 facts without guessing
• Some participants rated their motivation to learn each fact before it was revealed.
• Motivation ratings showed that Guessed Facts > Studied and Read facts

Question 13

What are the different theoretical explanations for the pre-testing effect?

Answer 13

• Theories in pre-testing on learning from errors are broadly classed into 2 types: semantic accounts and attentional accounts
• Search Set Theory (semantic)
• Attentional Accounts (motivation)

Question 14

What is the Hyper Correction Effect?

Answer 14

• The idea that, if an answer you were really confident on turns out wrong (big mismatch between accuracy of answer and confidence level) - you tend to remember the correct answer better than if your expectations were not violated.

Question 15

Outline the study on the Hyper Correction Effect

Answer 15

• Butterfield and Metcalfe 2001
• participants studied general knowledge questions
• They answered the questions, and also rated their confidence on how correct they thought their answer was
• Immediate feedback was given to their answer
• After 5 minutes they did a cued recall test on the general knowledge questions
• Finding: High confidence errors were more likely to be corrected than low confidence errors

Question 16

What is inductive learning?

Answer 16

A learning method in which general principles are discovered by examining specific examples, rather than being given rules upfront

Question 17

As part of semantic accounts for the pretesting effect, what are 2 parts of Search Set Theory?

Answer 17

• Elaboration: activating related concepts leads to a formation of richer, distinctive and more accessible memory traces
• Mediation: errors or other semantic associates offer an alternate retrieval route to the target.

Question 18

What are two main components of attentional accounts for the pretesting effect?

Answer 18

• Curiosity: guessing causes awareness of an information gap. Learners feel curious about the answer, so become more motivated to fill their gap in knowledge.
• Surprise: belief that a guess is correct results in surprise at feedback when it is revealed to be incorrect

Question 19

What is meant by derring - so what is the derring effect?

Answer 19

• Deliberate erring
• This is where answers (targets) were shown during error generation, so the participant was told the target word, yet still had to write possible errors.
• Errors were committed on purpose
• The derring effect is where, when you generate plausible deliberate errors to known information, learning is enhanced compared to if you did errorless studying

Question 20

Why is the derring effect similar to the pretesting effect?

Answer 20

• Both effects occur as a result of committing and correcting related errors during study

Question 21

How would semantic, and then attentional accounts look at the derring effect?

Answer 21

• Semantic accounts: error generation should benefit from semantic activation, regardless of prior knowledge of the answer. So, derring will be more effective than copying for learning word pairs (as copying does not activate the same level of semantic activation)
• Attentional accounts (op): error generation should only benefit learning when the answer is unknown (we need curiosity and surprise for the real answer). So, derring will be no more effective than copying for learning word pairs

Question 22

What is a point against error generation helping learning?

Answer 22

• The idea that error generation may potentially interfere with the real answer
• This is seen that, despite producing 1 or 3 errors being both equally higher than copying - producing 3 errors (more errors than 1) led to higher interference (but it was still rare - so unlikely to have a large effect)