Metacognition Flashcards
Inaccurate predictions caused by
- Take perspective of one’s “future” self being tested
- faulty mental model of how memory works
- informative patterns of errors and associations
Monitoring
Keeping track of own learning and how well you’re learning
e.g. Judgements of Learning (JoLs) - ask people to make judgements as encoding new information and base it on/compare it to future performance
Middle stages of metamemory/metacognition
Acquisition, retention, retrieval
Metamemory
In the real world, we frequently under uncertain conditions e.g. when we try to evaluate our own (imperfect) learning or predict future performance
Control
Making changes to your study schedule as a result of monitoring outcomes/beliefs
What you do based on the outcome of these monitoring processes e.g. feel like you look at something and feel like you won’t remember it later on - make changes to schedule based on that introspection
The ability to evaluate own learning accurately
General method to study this - lab - study cue-target word pairs and provide a JOL for each pair
How likely will you recall this information later on from 1-7?
related pairs: doctor-nurse
unrelated pairs: coffee-nurse
(includes distractor task)
All tests provide target in response to cues
Are JoLs strong predictors of future performance?
Immediate JoLs show overconfidence
Are metacognitive judgements directly affected by strength of memory trace?
NO - when asked to evaluate what we know - we make an inference based on a range of cues
Judgements based on CUES not strength of original memory
who did Cue Utilisation Framework?
KORIAT (97)
What did Koriat say?
Higher accuracy if cues are consistent with factors that improve memory
We rely on multiple cues
Cues
INTRINSIC
EXTRINSIC
MNEMONIC
Intrinsic cues
A word’s inherent pre-existing ease of learning - relatability, imageability - STRONG INFLUENCE ON JOLs
More sensitive to item relatedness
Extrinsic cues
Conditions of learning or learner’s encoding operations
WEAKER INFLUENCE ON JOLs
we ignore encoding conditions/study time/processing levels manipulations/spacing
Mnemonic cues
Internal subjective idiosyncratic signals of learning
Intrinsic and extrinsic cues
Support analytical conscious inferences reflecting a person’s naive theories about learning
Mnemonic cues
HEURISTICS - non-analytical processing
Cue utilisation
Learners UNDERESTIMATE value of extrinsic cues. We predict that “easy” material will be easier to remember than “hard” material regardless of amount of times practiced/exposed to it
underestimate practice/repetition effects
Cue utilisation e.g. Carroll (97)
Participants studied:
- related word pairs (“easy” pairs) - 2 recalls
- unrelated word pairs (“hard” pairs) - 8 recalls
Participants believed that they’d remember more easy pairs (their JoL) disregarding the power of extrinsic cues of study - no. of recalls
Final test performance was opposite to JoLs
Koriat (1997) - How do people monitor their knowledge during acquisition?
Cue utilisation approach
3 types of cues for judgements of learning
- intrinsic (item difficulty)
- extrinsic (recall, environment stimulus duration)
- mnemonic
Koriat (1997)
In 4 experiments, using paired-associate learning, item difficulty (intrinsic) exerted similar effects on JoL and recall.
Factors; list repetition, item repetition within a list and stimulus duration (extrinsic) affected JoLs LESS strongly than recalls
- underestimate extrinsic cues
Effects of practive
Saw a shift from reliance on intrinsic factors towards greater reliance on mnemonic-based heuristics
Koriat and Bjork (2006) - MNEMONIC CUES
tried out procedures to alleviate FORESIGHT BIAS
enhance learners sensitivity to mnemonic cues - resort to THEORY-BASED judgements
JoLs underestimation in recall increase that occurs with repeated study (extrinsic cue) - what is this called?
underconfidence-with-practice effect (UWP)
What does delaying JoLs do?
REDUCES debiasing effect of UWP
Koriat study (1997) - effects of practice
One group were given one word list and tested twice
Other group were given one word list 1, tested once, then given list 2 and tested on list 2
Results: UNDERCONFIDENCE-WITH-PRACTICE
same condition (list 1) ppts who had JoL of 10% - recalled 30%
If you give ppts multiple chances to study material - they underestimate their abilities
Underestimate abilities when manipulate/increase these extrinsic cues
- number of study opportunities
- duration of study time
Predicting one’s own forgetting - how well do you think you’ll do on an immediate test a DAY later and a WEEK later?
Results - striking difference between actual and predicted after day/week
People are OVERCONFIDENT in prediction
folk predicting low JoLs are under-confident (on average)
high JoLs are overconfident
Predicting one’s own forgetting
Overconfident - insensitive to RETENTION INTERVAL
Kornell & Bjork (2009)
Overestimating remembering - overconfident about retention and underestimate influence of time on forgetting
Underestimating learning - underconfidence-with-practice effects - effects of repetition and practice are underestimated
Serra & Metcalfe (2009) - METACOGNITIVE ILLUSIONS
Effective implementation of metacognition
3 components of metacognition: Knowledge, Monitoring, Control
Metacognitive illusions
Systematic ERRORS in metacognitive monitoring
Why is metacognitive accuracy generally low?
Inferences based on cues/beliefs NOT heuristics/theoretical fact
Heuristic definition
Enabling person to discover or learn something for themselves
Metacognitive illusions
- Familiarity heuristic - perceptual fluency
- Retrieval fluency heuristic
- Current knowledge heuristic - hindsight bias
- Association heuristic - foresight bias
Familiarity heuristic
We mistake familiarity/perceptual fluency for knowledge/comprehension
DECEPTIVE
- superficial
- does not predict retrievability accurately
- leads to people overestimating DEGREE OF MASTERY
Familiarity study
Overall familiarity with a domain does not predict future performance accuracy
POORER JOL PREDICTIONS FROM EXPERTS IN A DISCIPLINE - ASSUME KNOWLEDGE COS FAMILIAR - more likely to misattribute familiarity to comprehension
Ex 2 - Boosting word’s accessibility via priming increases familiarity - increases error rates
Subliminal presentation of target before test - increases accessibility - higher false alarm rate
Retrieval fluency
We believe fast/easy retrieval of information is knowledge
DECEPTIVE
Info easy to retrieve when you are given a cue can be more difficult to retrieve on FREE RECALL test
Effort, not retrieval fluency, predicts performance
increased awareness will decrease this illusion
Retrieval fluency study
Benjamin et al (1998)
Ex1 - compare retrieval from semantic memory and episodic memory - base our judgements on the ease of retrieval from . semantic memory, underestimating episodic distinctiveness
Ex1 study: ppts answered questions about world events then made JoLs about later then FREE RECALL of all answers Easy events (well known info stored in semantic memory)- fast memory search, so we believe this info will be easy to retrieve
Weak and less accessible episodic memory trace = poorer recall
Ex1 results - OPPOSITE TO PREDICTIONS - fast responses at study (indic. high retrieval fluency) –> poor recall at test, ppts mistakenly relied on ease of retrieval from semantic memory to predict performance
Hindsight bias (“know it all”)
Fischoff (1975) - read stories of historical events and predict how likely the outcome of these events were in the past
results: OVERestimate past likelihood based on current knowledge
Jacob & Kelley (1987) - predict how difficult an anagram will be for OTHER PEOPLE
Predictions made whilst looking at answers or not
Results: Anagram rated easier when answers were available
Foresight bias (Association)
Curse of knowledge - illusions of competence
When we study, we make predictions about the future performance based on LOCAL STUDY conditions
Foresight bias (Association)
Curse of knowledge - illusions of competence
When we study, we make predictions about the future performance based on LOCAL STUDY conditions
Bjork & Koriat (2005) - study cue-target pairs (paired-associate learning) - items with different theoretical and observation relatedness (Ex 1) and different forward and backward associations (Ex 2)
Results: JoLs don’t distinguish between the two - people are over confident - actual recall vs. predicted for BACKWARDS associations = lower
Immediate judgements - not accurate
Immediate JoLs are non-diagnostic (not accurate)
Judgements are made for info still in WORKING MEMORY thus very memorable
Delayed judgements - more accurate
Delayed JoLs are more diagnostic
They often show UNDERconfidence - studies say they . are based on RETRIEVAL attempts
Monitoring –> Control
Control: learner must manage learning by making number of assessments/decisions
Knowing what you know (JoLs)
Knowing what to study and re-study
Knowing when to stop studying
Study: what we believe we know influences how we study (Metcalfe & Finn. 2008)
Self regulation: to optimise learning, study selectively - eliminate what you have learned and re-study “harder” material
Study-time allocation - DISCREPANCY REDUCTION MODEL
Norm of study = desired level of learning (learning criterion)
When the goal is to learn as much as possible, you study the hard items to minimise the discrepancy between your learning criterion and actual learning level
(neg. correlation between JoLs and study time allocations - if you have higher JoL, you believe you know it well and thus allocate less time to re-study it
REGION OF PROXIMAL LEARNING framework
You pick the items that you almost mastered –> highest payoff - stop studying when you feel like you are not making new progress (the stop rule)
What about time pressure?
Short study time available - focus on “easy” items - items in region of PROXIMAL learning
Longer study time available - focus on “harder” items that’ll take longer to master
Testing
Retrieval = learning
Information that is recalled is subsequently remembered better (but sometimes at the expense of non-recalled information; RIF). Thus, testing can modify the memory trace itself (Bjork, 1975).
But testing also has metacognitive benefits: it provides you with information about what you don’t know and reduces the foresight bias (Soderstorm & Bjork, 2014).
If you experience a retrieval failure, you can then modify your study strategies.
I.e., testing can drive subsequent learning (Roediger & Karpicke, 2008).
Spacing effects
Spaced study episodes - better retention
Lag effect - longer lags - better retention
Increasing delays = . better learning
Retrieval failures provide important feedback
up to a point where longer intervals may mean more forgetting
Spacing accounts
Accounts:
- Encoding variability: without spacing, the study context at S1 and S2 is very similar, so you consistently use the same retrieval cues to retrieve a specific memory. Each additional spaced study session increases the distinctiveness of the material and the cues.
- MIX UP CUES - Diminished processing: without spacing, you engage in less extensive processing at S2 than S1 (labour-in-vain effect: you haven’t forgotten what you learned at S1 yet, so additional studying doesn’t help).
LABOUR IN VAIN - Lack of diagnostic retrieval attempts: without spacing, you don’t engage in retrieval attempts that give insight into study success. You master a small set of cues and fall prey to the illusion of learning.
Are students able to optimise learning?
Study with postal workers: Workers preferred massed schedule of practice
but spaced schedule more beneficial/showed strongest learning - another metacognitive illusion showing faulty mental model of human memory
When re-studying and when given a choice about when to re-study, do students choose to space the material?
SPACING > MASSED STUDY
- When re-studying, 75% of participants choose to mass rather than space
- Participants predict better performance after massed than spaced study – possibly because massed study gives you the ILLUSION that processing gets easier over time
- Participants also prefer shorter to longer lags
- But with practice, participants begin to predict better performance after spacing
Why do students prefer massed practice?
Massed practice gives you feeling of learning more quickly than spacing and thus more confidence in learning
Immediate tests - massed practice schedule leads to better memory performance
Delayed test - spaced practice schedule leads to better memory performance
Desirable difficulties
- Varying the (cognitive, environmental) conditions of learning
predictable conditions contextualized learning
variable conditions more robust learning - Spacing
spacing enhances long-term retention - Interleaving topics
interleaving enhances learning of higher-level information
(similarities and differences increases transfer), requires re-
loading information in memory, and this benefits long-term
retention - Using tests as study events
a) the generation effect = generating an answer yourself results in
better retention than being presented with an answer
b) tests are retrieval practice, memory modifiers
Under these conditions, learners do not have a subjective sense of progress during training (they do not have illusions of competence). Yet, these conditions result in better long-term retention.
Participants “are apparently unaware that the more difficult or involved the process of retrieval, provided it succeeds, the greater its impact on subsequent recall” (Bjork, 1999, p.451)