Date post: | 14-Dec-2015 |
Category: |
Documents |
Upload: | maud-golden |
View: | 213 times |
Download: | 0 times |
Theoretical Relevance: Lecture 2 for the IV track of the 2007 PSLC Summer School
Robert G.M. Hausmann
(Holodeck version of Kurt VanLehn)
Literature: It is painful not to know the answer to a known question
Known questions appear at the ends of papers, reviews, etc.– At least one informed person cares about the answer.
Common (bad) ways to pose research questions– Cool software– Pop psychology– I learn this way, so…
No customers
Select a question to add information and clarity to the literature
Information value (in Shannon’s sense) – High if prior probability of the answer is very different
from the answer obtained in the experiment.– Low if experiment just confirms the expected answer.
Clarification value (real pains here)– Low if the literature is a mess, and the experiment just
adds one more fact to the mess.– High if the experiment somehow clarifies the mess.– Moderate if there is little prior literature.
Look for (and relieve) the pain in the…
Literature– Known question– Answer would add information and/or clarity
Classroom Sciences of Learning
Next
What pains the classroom?
Ask the instructor (you?) what’s most frustrating– Teaching a certain concept?– Transfer to real world?– Depth of understanding?
Ask the students…
Andes is not “selling” (can’t give it away!)
Andes teaches quantitative problem solving.
Most instructors think this is not a bottleneck.
Instead, qualitative problem solving is their concern.
Look for (and relieve) the pain in the…
Literature– Known question– Answer would add information and/or clarity
Classroom– Instructors consider the question important
Sciences of LearningNext
Where is the pain in the Learning Sciences?
Too many results No organization of the results No theory No clear implications No classic results that everyone knows No accretion Progress is more like politics than medicine
To cure the pain, Learning Science needs a theoretical framework
Not like physics– A few basic principles from which all else follows.
More like Medicine– A few basics (anatomy, physiology, genetics)– Many specializations e.g., lymphatic cancers
» Few principles; many diseases, syndromes, therapies
– A standardized, rigorous terminology– Digital libraries becoming essential
Types of theories
Computational models
“How People Learn”
principles
Shared theoretical vocabulary
Boxology
PSCL theoretical framework
Computational models
“How People Learn”
principles
Shared theoretical vocabulary
Boxology
Shared terminology
Micro-level– Knowledge component: A principle, concept, fact,
schema, strategy, meta-strategy…– Learning event: An application of a knowledge
component
Macro-level: A taxonomy of robust learning processes– Sense-making– Fluency-building
Micro level is just (good, old fashioned) cognitive psychology
Instructional activities Prior knowledge
Cognitive processes
Knowledge components
Observable outcomes
Knowledge can be
decomposed
Learning processes
can be decomposed
and taxonomized
Knowledge of the solo
student
Knowledge of the solo
student
Macro level is a taxonomy of learning process
Sense making– Coordination of multiple types/sources of learning
» Example: step plus a rule
– Interaction of the student with other agents» Agents can be peers, experts, or tutoring systems.
Fluency– Three Mechanisms:
» Strengthening» Deep-feature perception» Headroom
PSLC research clusters Coordinative learning
– How do students coordinate multiple sources of information, media, representations, strategies?
Interactive communication– How does interaction between a student and a
peer, tutor or teacher affect learning? Fluency and refinement
– How does skill become fluent?
Coordinative learning Co-training (Blum & Mitchell) Learning from multimedia (Clark & Mayer)(Tversky) Learning from analogies (Novick & Holyoak, 1991) (J.R.
Anderson, Fincham, & Douglass, 1997) (VanLehn, 1998) Learning from multiple representations & multiple
solutions (Ainsworth, 1999) Learning from agents (Lester, Converse, Stone, Kahler,
& Barlow, 1997) (Graesser et al., 2003) (Moreno, Mayer, Spires, & Lester, 2001)
Interactive communication Feedback and hint effects (J. A. Kulik & Kulik, 1988)
(McKendree, 1990) (Hume et al., 1996) (Kluger & DeNisi, 1996) *(Corbett & Anderson, 2001) (Mathan & Koedinger, 2005) (V. J. Shute, 1992)
Learning from examples, self-explanation and fading *(Collins, Brown, & Newman, 1989) (Nguyen-Xuan, Bastide, & Nicaud, 1999) (Kalyuga, Chandler, Tuovinen, & Sweller, 2001) (Renkl, Atkinson, Maier, & Staley, 2002) (Kalyuga, Ayres, Chandler, & Sweller, 2003) (Atkinson, Renkl, & Merrill, in press) *(M. T. H. Chi, 2000) (M.T.H. Chi et al., 2001) (V. Aleven & Koedinger, 2002) (Siegler, 2002) (Corbett, Wagner, lesgold, Ulrich, & Stevens, 2006)
Tutorial dialogues vs. monologues *(VanLehn et al., in press) (Vincent Aleven, Ogan, Popescu, Torrey, & Koedinger, 2004)
Learning with a peer, including collaborative learning, peer tutoring, learning by teaching (Reif & Scott, 1999) (Okita & Schwartz, 2006)
Fluency and refinement Practice effects, including spacing and part-
whole training effects *(Newell & Rosenbloom, 1981) (J.R. Anderson et al., 1997) (Pavlik & Anderson, 2005)
Macroadaptation and mastery learning effects (Bloom, 1984) *(C. Kulik, Kulik, & Bangert-Drowns, 1990) (V. J. Shute, 1992) (V.J. Shute, 1993) (Corbett, 2001) (Ainsworth & Grimshaw, 2004) (Arroyo, Beal, Murray, Walles, & Woolf, 2004)
Implicit (practice only) vs. explicit (direct) instruction. *(Berry & Broadbent, 1984) (Singley, 1990) (K. Koedinger & Anderson, 1993) (Klahr & Nigam, 2004) (VanLehn et al., 2004)
Current research projects
Empirical projects
AlgebraGeometryChemistryPhysicsFrenchChineseESLFluency & refinement 2 2 4 5 3Coordinative learning 2 3 2 1 1 2 1Interactive communication 2 1 8 1
Enabling technology 7
PSLC Theoretical framework
Glossary of theoretical terms– Micro-level – Macro-level
Analytic framework Next
Learning events over timeD
urat
ion
Fourth Third Second First Fifth
While studying an example, tries to self-explain; fails; looks in text; succeeds
While solving a problem, looks up example; recalls explanation; maps it to problem
Recalls explanation; slips; corrects
Solves without slipsSolves without slips
5 sec.
10 sec.
15 sec.
25 sec.
20 sec.
A new analytic framework, based on an analogy
A problem is to a problem space asa learning event is to a ______________
A new analytic framework, based on an analogy
A problem is to a problem space asa learning event is to a learning event space.
Key ideas A learning event space is a set of paths determined
by the instruction and the student’s prior knowledge,
but it is the student who chooses which path to follow
different paths have different outcomes:
– Deep learning
– Shallow learning
– Mis-learning
– Etc.
You get to choose the granularity
Coarse grain-size: Only observable actions – Correct vs. incorrect steps– Feedback from tutor
Finer: Reportable mental actions– Recall vs. construct
Even finer?
How to use learning event spaces
Construct a learning event space such that… it is consistent with observable actions,
and… the top level question, “Why did they learn?” becomes two easier questions:
– Path choice: Why did students tend to choose as they did?
– Path effects: Given that a student went down a path, why did that cause the observed learning/outcomes?
A simple illustration Maxine Eskenazi & Alan Juffs hypothesize
that using authentic texts will increase vocabulary acquisition in ESL.– Students read text with a few target unfamiliar
words.– Texts come either from web or from existing
primer.– Clicking on an target word displays its definition.
Why would authenticity increase learning? How?
Learning event space (one per target word)Start Ignore the word
– Exit, with little learning Infer meaning from context
– Exit, with “implicit” learning Click on the word; definition is displayed
– Read & understand the definition» Exit, with “explicit” learning
– Go to Start
Why should authentic text help?Hypotheses based on path choicesStart Ignore the word
– Exit, with little learning Infer meaning from context
– Exit, with implicit learning Click on the word; definition is displayed
– Read & understand the definition» Exit, with explicit learning
– Go to Start
Authentic text should decrease this
Authentic text should increase this
Why should authentic text help?Hypotheses based on path effectsStart Ignore the word
– Exit, with little learning Infer meaning from context
– Exit, with “implicit” learning Click on the word; definition is displayed
– Read & understand the definition» Exit, with “explicit” learning
– Go to Start
Cue validity of this path increases
No change???
No change
To summarize the theoretical framework…
Glossary– Macro-level
» Sense-making Coordinative Learning Interactive Communication
» Fluency
– Micro level: Knowledge components, learning events… Learning events space
– Decomposes “why did they learn?” into» Path choices: Which paths were chosen?» Path effects: For each path, what was learned?