Self-regulated learning: Heuristics and illusions

Robert A. BjorkUniversity of California, Los Angeles

Panel on Does Knowing What You Know Improve Study Habits and Learning?

FIFTH ANNUAL RESEARCH CONFERENCE

Institute of Education SciencesNational Harbor, Maryland

June 28-30, 2010

Components of becoming metacognitively sophisticated as a learner

Managing (optimally) the conditions of one’s own learning Spacing, variation, generation, retrieval practice, … Organizing one’s knowledge, using technology, engaging in cooperative

learning, … Judging (accurately) whether learning/comprehension that will

support later recall/transfer has been achieved Interpreting the meaning and predictive value of objective and

subjective indices of current performance Understanding that changes from the study context to the test context

will impact access to what has been learned Avoiding “foresight bias” (Koriat & Bjork, 2005)

Giving appropriate weight to the impact of retention interval and subsequent study opportunities

Avoiding “stability bias” (Kornell & Bjork, 2009)

Interpreting (and misinterpreting) objective indices of performance

Learning versus performance What we can observe is performance; What we must infer is learning; …and the former is an unreliable guide to the latter.

Conditions of instruction that make performance improve rapidly often fail to support long-term retention and transfer, whereas

Conditions of instruction that appear to create difficulties for the learner, slowing the rate of apparent learning, often optimize long-term retention and transfer

Teachers and learners alike can be fooled Teachers become susceptible to choosing poorer conditions of

instruction over better conditions; … and learners to preferring those poorer conditions

Examples

Generation Interleaving Spacing

Examples of learners being fooled

Interpreting (and misinterpreting) subjective indices of performance

Perceptual fluency or familiarity The sense of ease in processing visual or auditory

information

Retrieval fluency How readily information “comes to mind”

Fluency of induction The sense of ease in noticing the commonalities

across exemplars of a category or concept

Perceptual fluency or familiarity

Heuristic value Misattributions and illusions

Misinterpreting the cause of perceptual fluency Misinterpreting the meaning of perceptual fluency Illusions of competence

Example: Reder (1987, 1988)

(Reder, 1987, 1988)

“What is the term in golf for scoring one under par?”

Retrieval fluency

Heuristic value Misattributions and illusions

Interpreting performance as learning; illusions of competence

Egocentrism in instruction and social communication Incomplete/faulty models of ourselves as

learners/remembers Example: Benjamin, Bjork, & Schwartz (1998)

Benjamin, Bjork, & Schwartz (1998)

Phase 1: 20 (easy) general-knowledge questions E. g., “Who was the first president of the United States?” Participants asked to:

(a) hit ENTER as soon as the answer “came to mind” (latency recorded);

(b) say the answer; (c) predict the likelihood they would be able--at the end of the

experiment--to free-recall having given that answer. Phase 2: Distracting activity (spatial/map task) Phase 3: Final test

Free recall: Write down as many of the 20 answers you gave earlier as you can;

(Original questions were not shown again)

Interpreting fluency of induction(Kornell & Bjork, 2008)

• The ability to generalize concepts and categories through exposure to multiple exemplars.

“Spacing is the friend of recall but the enemy of induction.”

-Ernst Rothkopf

Lewis LewisLewisLewis Lewis Lewis

Pessani StratulatSchlorffWexler Juras Mylrea

Pessani StratulatSchlorff WexlerJuras Mylrea

Hawkins HawkinsHawkinsHawkins Hawkins Hawkins

M S S M M S S M M S S M

FeedbackTest

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Massed Spaced

Pro

port

ion C

orr

ect

Results

Actual Responses

A qualification

Subjective experience is not always misleading, it is sometimes even the best basis for judgments

Example: Jacoby and Kelley (1987)

Jacoby & Kelley (1987)

FSCAR ?????

vs.

FSCAR SCARF

Jacoby, Bjork, & Kelley (1994)

“Subjective experience, like the public media, is unavoidable, serves useful functions, and is not to be fully trusted”

Why is performance a poor guide to learning?

Performance is heavily influenced by local conditions—cues, predictability, recency—which can serve as crutches that prop up performance, but will not be there later at the time of test

Predictions of future recall are, for example, subject to a foresight bias (Koriat & Bjork, 2005, 2006)

Foresight bias: an illustration

Likelihood the second word will be given as a free associate to the first? Lamp: Light Find: Seek Sell: Buy Cheese: Cheddar Citizen: Tax

Foresight bias: an illustration

Likelihood the second word will be given as a free associate to the first? Lamp: Light (.71) Find: Seek (.03) Sell: Buy (.56) Cheese: Cheddar (.03) Citizen: Tax (.00)

Foresight bias: an illustration

Likelihood the second word will be given as a free associate to the first? Lamp: Light (.71) Forward pair Find: Seek (.03) Backward pair Sell: Buy (.56) Forward pair Cheese: Cheddar (.03) Backward pair Citizen: Tax (.00) “Purely a-posteriori” pair

0

10

20

30

40

50

60

70

80

90

100

Zero Low High

Associative Level

Actual

Predicted

Foresight bias: Dynamics

Judgments of learning are made in the presence of information that will be absent, but solicited, on a subsequent test The targets in cue-target pairs (e.g., Cheese: cheddar); or the

answers to questions (e.g, the Capital of Australia is Canberra)

We are unable to anticipate the test situation, when the cue/question alone will trigger other associations “Cheese ___?____” will trigger other strong associates, such as

“mouse,” “bread,” “wine,” etc., which will compete with “cheddar.”

“Capital of Australia?”—will trigger Sydney, Melbourne, …

Finally, the impact of intervening events: Predicting one’s own forgetting and learning

We are subject to what Kornell and Bjork (2009) have labeled a stability bias: The tendency to think that a memory representation,

once formed, will remain stable

This bias leads to both Overestimating remembering (i.e., underestimating

forgetting); and Underestimating learning.

Predicting one’s own forgetting (Koriat, Bjork, Sheffer, & Bar, 2004)

Experiment 1 (of 9): 60 paired associates

30 related; 30 unrelated

Participants judge, pair by pair, the likelihood they will remember that pair on a later cued-recall test

Retention interval to the final test (between-subjects): Immediately after the study phase; One day; One week

Koriat, Bjork, Sheffer, & Bar (2004)

Experiment 7:Predicting one’s own learning

24 paired associates 12 related (Hill-Valley) 12 unrelated (Clemency-Idiom)

Number of study/test cycles (within-subjects): ST STST STSTST STSTSTST

During the first study trial, participants judged, pair by pair, the likelihood they would remember that pair on either the first, second, third, or fourth cued-recall test cycle

Within-subjects Response panel insured that participants predicted for the designated

test

0

10

20

30

40

50

60

70

80

90

100

1 2 3 4

Trial

Pro

port

ion C

orr

ect Actual Easy

Actual Hard

Predicted Hard

Predicted Easy

0

10

20

30

40

50

60

70

80

90

100

1 2 3 4

Trial

Proportion C

orrect

Actual

Predicted

Concluding comment:

People believe, in general, that forgetting happens over time and that studying

fosters learning, That is, they have a theory of forgetting and a theory of

learning but they do not appear to believe that access to a given

item in memory will be lost over a retention interval or increased by further study.

The end, probably

Final comment, if there is time, on our subjective experience as teachers

Egocentrism in social communication

Newton (1990) as a parable of teaching;

Piaget (1962) quote

Piaget (1962)

“Every beginning instructor discovers sooner or later that his first lectures were incomprehensible because he was talking to himself, so to say, mindful only of his point of view. He realizes only gradually and with difficulty that it is not easy to place one’s self in the shoes of students who do not yet know about the subject matter of the course.”

The end (for sure)

Koriat et al. (2004) conclusions

“… participants can access their knowledge about forgetting, but only when theory-based predictions are made, and then only when the notion of forgetting is accentuated —either by manipulating retention interval within individuals, or by framing recall predictions in terms of forgetting rather than remembering.”

“Once the notion of forgetting is activated, people can then take into account what they know about the specified retention interval in making their recall predictions. It is indeed quite instructive that they do not do that spontaneously—even when the specified retention interval is a year (Experiment 4c)!”

Koriat, Bjork, Sheffer, & Bar (2004)Predicting One's Own Forgetting: The Role of Experience-Based and Theory-Based Processes.

“We examined the hypothesis that judgments of learning (JOL), if governed by processing fluency during encoding, should be insensitive to anticipated retention interval.”

“The initial impetus … was the dual-basis view of metacognitive judgments (Brown & Siegler, 1993; Jacoby & Kelley, 1987; Kelley & Jacoby, 1996; Koriat & Levy-Sadot, 1999; Strack, 1992).”

Subjective experience: “Various mnemonic cues contribute directly to produce an immediate feeling of knowing that can serve as the basis of judgments. Thus, for example, encoding and retrieval fluency may foster a feeling of competence that can serve as [an experience-based basis for judgments of learning].”

Domain-specific knowledge retrieved from memory. “Theory-based judgments, in contrast, rely on the deliberate use of specific beliefs and information to form an educated guess about one's own knowledge. Thus, JOLs may utilize such rules as “memory performance should be better on recognition than on recall memory test.”

Experiment 8:Predicting one’s own learning

24 difficult (unrelated) paired associates (e.g., Clemency-Idiom) Removing item differences should decrease participants’ tendency to

base their judgments on intrinsic between-item differences Number of study/test cycles (within-subjects):

ST STST STSTST STSTSTST

Participants judge, pair by pair, the likelihood they will remember that pair on the first, second, third, or fourth cued-recall test cycle

Within-subjects Response panel insured that participants predicted for the designated

test