rummel example based learning naples v3 · ruhr%universitätbochum example.based!learning!...
TRANSCRIPT
Ruhr-‐Universität Bochum
Example-‐Based Learning
May 16th, 2014 Prof. Dr. Nikol Rummel InsGtute of EducaGonal Research Ruhr-‐Universität Bochum, Germany
Ruhr-‐Universität Bochum
Instances of Learning by Example?
2
Ruhr-‐Universität Bochum
Instances of Learning by Example
3
Math course book: example of how to solve equaGon for ‚y‘
Watching YouTube tutorial on how to play a song on guitar
Observing host using cutlery in fancy restaurant
Reading manual to figure out how the new camera works
Learning how to Cha-‐Cha in dancing lesson
Watching mother make pancakes
Ruhr-‐Universität Bochum
Lines of Research -‐ Two Paradigms
Social-‐Cogni+ve Theory (Bandura)
Cogni+ve Load Theory (Sweller)
Learning from observing a model‘s problem-‐solving behavior
Learning from studying worked examples
4
Both perspec=ves share: construcGng appropriate cogniGve representaGons in order to
show the behavior later Difference:
nature of aZenGon processes
Ruhr-‐Universität Bochum
ObservaGonal Learning (OL) -‐ IntroducGon • Learner observes model performing task • Model shows how problem is solved • Model can behave naturally or didacGcally
• PresentaGon: FtF, video, screen recording, animaGon
• Danger: Transient informaGon in modeling examples!
5
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Research Example: Rocky Experiment (Bandura 1965)
• Children (age: 3-‐6) watched video of an adult male (Rocky) and his behavior towards a Bobo-‐Doll
• the model showed – novel aggressive behavior – novel verbal aggression
Exp. Phase
6
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Research Example: Rocky Experiment (Bandura 1965)
• Children (age: 3-‐6) watched video of an adult male (Rocky) and his behavior towards a Bobo-‐Doll
• the model showed – novel aggressive behavior – novel verbal aggression
• Model experiences consequences – Cond. 1: reward (drinks, sweets, praise) – Cond. 2: punishment (scolding, threat) – Cond. 3: no consequences
• Children play in the same room à spontaneous imitaGon • Children are asked to reproduced observed behavior; receive reward
à reinforced imitaGon of behavior
Exp. Phase
Test Phase I
Test Phase II
7
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Research Example: Rocky Experiment (Bandura 1965)
8
Ruhr-‐Universität Bochum
• Test Phase 1: spontaneous imitaGon of observed behavior; – model punished: less imitaGon
– model rewarded: more imitaGon à observing consequences for others has effect
• Test phase 2: reinforced imitaGon – imitaGon same in all condiGons
à observing a model leads to learning
9
ObservaGonal Learning -‐ Research Example: Rocky Experiment (Bandura 1965)
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Learning Process (Bandura)
1. AZenGon 2. RetenGon 3. ReproducGon 4. MoGvaGon
Bandura ,1986
10
AcquisiGon
Performance
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Research Example: (Rummel, Spada & Hauser, 2009)
• Students observe scenes of a well-‐structured, successful
collaboraGon • model seing resembles the collaboraGon seing
• delivered as mulGmedia presentaGon on the computer screen • audio recordings of dialog supplemented by animated slide-‐
clips
Ruhr-‐Universität Bochum
-‐ Research Example: (Rummel, Spada & Hauser, 2009)
!" Scenes of the model collaboraGon are presented via audio
Symbolic images of the collaboraGng model partners
Shared text editor with joint model soluGon.
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Research Example: (Rummel, Spada & Hauser, 2009)
• Observing model leads to beZer collaboraGon and beZer
joint soluGon in subsequent own problem-‐solving as compared to – ScripGng first collaboraGon – Unsupported first collaboraGon
• prompGng of self-‐explanaGon acGviGes further enhanced effect of model
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Design Principles • Model-‐observer similarity • Mastery modeling vs. coping model • PersonalizaGon and voice principle • Image principle • AZenGon guidance
Van Gog & Rummel, 2010
Renkl, 2014
14
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ EffecGveness: advantages & boundaries • No trial learning vs. complexity of behaviour
15
Ruhr-‐Universität Bochum
ObservaGonal Learning -‐ Categorizing Influences of Model (Bandura)
16
Response Facilita=on
• Behavior already acquired • Shown unconsciously
Inhibi=on • Behavior has already been acquired à observing consequences: effect on frequency
Acquisi=on of new behaviour by observa=on
• New behavior, parts are already available (configuraGon)
Social Promp=ng • Observing (socially accepted) alternaGve behavior being rewarded leads to alternaGon of own behavior
Changing Emo=onal State • EmoGonal state is transferred from model to observer (acquisiGon by repeGGon)
Ruhr-‐Universität Bochum
Worked Examples (WE) -‐ IntroducGon • Contain problem and worked-‐out soluGon steps leading to
final soluGon • Demonstrate didacGcally how to solve a problem • Focus learner‘s aZenGon on problem states, operators, and
applicaGon of operators • PresentaGon: wriZen account (usually)
17
Ruhr-‐Universität Bochum
Worked Examples -‐ Example: Problem
Renkl et al. 1998: 100
18
Ruhr-‐Universität Bochum
Worked Examples -‐ Example: Problem
19
Ruhr-‐Universität Bochum
Worked Examples -‐ Worked-‐Example Effect • PosiGve effects of studying worked examples
compared to problem-‐solving • Self-‐explanaGon acGviGes are paramount to
success
• Robust effect: learning from WE compared to learning with ITS
Atkinson, Derry, Renkl & Wortham, 2000
20
Sweller, Van Merrienboer & Paas, 1998
Van Gog & Rummel, 2010
McLaren, Lim & Koedinger, 2008
Ruhr-‐Universität Bochum
Worked Examples -‐ Worked-‐Example Effect • WE more effecGve parGcularly for novices • Problem solving of novices:
weak strategies à high cogniGve load à less learning
• With worked-‐out steps: student can spend cogniGve capacity on understanding soluGon
Sweller et al. 1998
21
Ruhr-‐Universität Bochum
Worked Examples -‐ Research Example (Sweller & Cooper 1985)
Experiment 3: Algebra • Students (year 9 High School) learning 4 categories
of Algebra problems: solving for a
• WE condiGon: example – problem pairs • PS condiGon: problems to solve
• 4 Problems to solve
• WE spent less Gme in acquisGon phase • WE required less Gme to solve test problems +
made less math. errors
Acquis.Phase
Test Phase
22
Ruhr-‐Universität Bochum
Worked Examples -‐ Design Principles • Avoid split aZenGon • Avoid redundancy
• SegmenGng/ explicit subgoals
• Prompt self-‐explanaGons • Example-‐problem pairs • CompleGon Problems • MulGple examples • Erroneous examples
For Overview see: van Gog & Rummel 2010
23
Ruhr-‐Universität Bochum
Worked Examples -‐ EffecGveness: advantages & boundaries • LiZle cogniGve load vs. need for self-‐explanaGon • EffecGve vs. not at any stage of the learning process • Efficient vs. liZle transfer of schemas
24
Ruhr-‐Universität Bochum
IntegraGon of PerspecGves -‐ CommonaliGes and differences of WE and OL
• Discuss in your groups and try to sort the examples in the google file: – Which examples are instances of OL, which are instances of WE?
– Why? What are commonaliGes and differences of OL and WE?
25
Ruhr-‐Universität Bochum
26
Instances of example-‐based learning Worked Example or Observa=onal Learning?
Math course book: how to solve equaGon for ‚y‘ Worked Example
Observing host using cutlery in fancy restaurant Observa=onal Learning
Learning how to Cha-‐Cha in dancing lesson Observa=onal Learning
Watching YouTube tutorial on how to play a song on guitar
Observa=onal Learning/Worked Example
Reading manual to figure out how the new camera works
Worked Example
Watching mother make pancakes Observa=onal Learning
Ruhr-‐Universität Bochum
IntegraGon of PerspecGves -‐ CommonaliGes of WE and OL
• Providing example cases for learning is effecGve and efficient
• ParGcularly in iniGal stages of learning • Goal: Schema-‐AcquisiGon • Important:
• ConnecGng examples to underlying principles • FacilitaGng example comparison • FacilitaGng learner acGviGes to connect example(s) and principles
27
Ruhr-‐Universität Bochum
IntegraGon of PerspecGves -‐ Phases of Skill AcquisiGon 1. Principle encoding:
• acquiring declaraGve knowledge about domain
2. Relying on analogs • to solve problem via analog (AR)
3. Forming declaraGve rules: • learners form “if-‐then”: how to act to solve
the problem 4. fine tuning: automaGon and flexibilizaGon.
• chunking and automaGng soluGon steps • adapGng skill to new problems
Renkl 2014
28
Ruhr-‐Universität Bochum
29
Ruhr-‐Universität Bochum
Literature • Anderson, J. R. (1993). Rules of the mind. Hillsdale, NJ [u.a.]: Erlbaum.
• Bandura, A. (1965). Influence of models reinforcement conGngencies on the acquisiGon of imitaGve
response. Journal of Personality and Social Psychology, 1, 589-‐595.
• Bandura, A. (1986). Social foundaGons of thought and acGon: A social cogniGve theory. Englewood Cliffs,NJ: PrenGce Hall.
• Renkl, A. (2014), Toward an InstrucGonally Oriented Theory of Example-‐Based Learning. CogniGve Science, 38: 1–37. doi: 10.1111/cogs.12086
• Renkl, A., Stark, R., Gruber, H. & Mandl, H. (1998): Learning from Worked-‐Out Examples: The Effects of Examples Variability and Elicited Self-‐ExplanaGons. Contemporary EducaGonal Psychology 23, 90-‐108.
• Sweller, J., Cooper, G. (1985). The Use of Worked Examples as a SubsGtute for Problem Solving in Learning Algebra. CogniGon and InstrucGon, 1985, 2 (1) 59-‐89.
• Sweller, J., van Merriënboer, Jeroen J.G. & Paas, Fred G.W. C. (1998). CogniGve Architecture and InstrucGonal Design. Educa+onal Psychology Review, 10 (3), 251–296.
• van Gog, T. & Rummel, N. (2010). Example-‐Based Learning: IntegraGng CogniGve and Social-‐CogniGve Research PerspecGves. Educa+onal Psychology Review, 22 (2), 155–174.
30
Ruhr-‐Universität Bochum
Other Sources • Slide 07:
– hZp://www.moguul.de/papil/wp-‐content/uploads/2008/02/bobo-‐doll-‐experiment.jpg [05/03/2014]
• Slide 14,25, 32: – hZp://3.bp.blogspot.com/-‐aWbxYuniqH0/UM3B3CitwXI/AAAAAAAAAqI/iiwZ82CtjDM/s1600/quesGon2.jpg [05/04/2014]
31