Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Adaptive
language
learning
Challenges & opportunities
Piet DESMET & Mieke VANDEWAETERE
CALICO 2015
ADAPTIVE
INSTRUCTION
CHALLENGES &
OPPORTUNITIES
WHAT
HOW
WHY
USE CASES
ADAPTIVE INSTRUCTIONI
WHAT IS ADAPTIVE INSTRUCTION?
“the method by which learners
are offered tailored instruction and support,
personalized to ...
WHY ADAPTIVE INSTRUCTION?
Rationale
Learners differ
Cognitive
prior knowledge, metacognition, beliefs, goals, etc.
Affecti...
WHY ADAPTIVE INSTRUCTION?
Goal
To design supportive learning environments that account for individual
differences between ...
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
Vandewaetere, Desmet & Clarebout 2011 / Vandewaetere & Clarebout, 2012
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
Learner
cognition
Learner
behavior
Learner
affect
Vandewaetere & Clarebout, 2012
ADAPT TO WHAT?
Learner characteristics
Cognition
 field (in)dependency, prior knowledge, learning style, information
skil...
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
static
dual pathway
dynamic
Vandewaetere & Clarebout, 2012
Static: prior to/after interaction
Eg. Student’s starting level is defined by teacher or pretest
Eg. Update of student’s l...
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
sequence
form
content
Vandewaetere & Clarebout, 2012
Adaptive curriculum sequencing
- From computerized adaptive testing
- Based on IRT:
- a measurement theory where the proba...
Adaptive form & content representation
- Adaptive content presentation of learning objects
 example: varying degrees of s...
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
learner-controlled
shared control
program-controlled
Vandewaetere & Clarebout, 2012
Program or system-controlled
System reasoner decides what content is offered.
Evaluation:
 Lack of choice lowers motivati...
Learner-controlled
Learner selects what content is offered.
Evaluation:
 Not all learners can deal with choice: “the art ...
Shared control
System preselects, learner chooses from preselection.
Best of both worlds
ADAPT HOW?3d
HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
Vandewaetere & Clarebout, 2012
 Device: Mobile vs desktop adaptive learning
 Short duration – long duration
 Short term – long term
 Environment char...
ADAPTIVE
INSTRUCTION
CHALLENGES &
OPPORTUNITIES
WHAT
HOW
WHY
USE CASES
Adaptive item selection based on combination of judgment and
data (What?) [Wauters, Desmet, & Van Den Noortgate, 2012]
- I...
Illusion of adaptivity might be as effective as adaptivity
(to what?) [Vandewaetere, Clarebout, Desmet, 2011]
 Adaptive i...
Learner control as a means to provide adaptive instruction
(how?) [Vandewaetere & Clarebout, 2011]
 LC has to be perceive...
Learner control as a means to provide adaptive instruction
[Vandewaetere, Clarebout, Desmet, 2011]
USE CASE 3II
Learner control as a means to provide adaptive instruction
USE CASE 3II
Instruction of Learner control as a means to provide adaptive
instruction
 Direct effect of instruction of LC on percepti...
ADAPTIVE
INSTRUCTION
CHALLENGES &
OPPORTUNITIES
WHAT
HOW
WHY
USE CASES
Adaptive feedback/support (adaptive representation) &LC
(What? & How?))
• Shed light on the effect of giving learners cont...
Adaptive media enhancement (adaptive representation)
(What?)
OPPORTUNITIESIII
 Different types of enhanced audio-visual input may serve different learning goals (e.g.,
improve comprehension, stimulat...
CHALLENGESIII
Big data, big opportunities
From manually entering data to online massive storage
From self-reporting data t...
Hype cycle for education (Gartner, 2013)
Contact
Piet Desmet
Piet.Desmet@kuleuven.be
http://wwwling.arts.kuleuven.ac.be/franling_e/pdesmet
be.linkedin.com/in/pietd...
Upcoming SlideShare
Loading in …5
×

2015 05-22 presentatie-calico_desmet & vandewaetere - def

508 views

Published on

Presentation at CALICO 2015

Published in: Education
  • How very interesting; I wish I could have made the trip! Thanks for sharing this.

    I like this presentation -very useful insights! Thanks for sharing it
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here

2015 05-22 presentatie-calico_desmet & vandewaetere - def

  1. 1. Adaptive language learning Challenges & opportunities Piet DESMET & Mieke VANDEWAETERE CALICO 2015
  2. 2. ADAPTIVE INSTRUCTION CHALLENGES & OPPORTUNITIES WHAT HOW WHY USE CASES
  3. 3. ADAPTIVE INSTRUCTIONI
  4. 4. WHAT IS ADAPTIVE INSTRUCTION? “the method by which learners are offered tailored instruction and support, personalized to the individual cognitive, affective and behavioral profile of the learner.” 1
  5. 5. WHY ADAPTIVE INSTRUCTION? Rationale Learners differ Cognitive prior knowledge, metacognition, beliefs, goals, etc. Affective motivation, fear, anxiety, etc. Behavioral need for help and feedback, gaming the system, etc. 2
  6. 6. WHY ADAPTIVE INSTRUCTION? Goal To design supportive learning environments that account for individual differences between learners (Shute & Zapata-Rivera, 2008) To enhance performance & learning (Shute & Towle, 2003)  individualized instruction is superior to the one-size-fits-all approach (Cohen, Kulik, & Kulik, 1982; Kadiyala & Crynes, 1998; Kulik, Kulik, & Bangert-Drowns, 1990) 2
  7. 7. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3
  8. 8. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 Vandewaetere, Desmet & Clarebout 2011 / Vandewaetere & Clarebout, 2012
  9. 9. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 Learner cognition Learner behavior Learner affect Vandewaetere & Clarebout, 2012
  10. 10. ADAPT TO WHAT? Learner characteristics Cognition  field (in)dependency, prior knowledge, learning style, information skills, working memory capacity, etc Affect  motivation, self-efficacy, frustration; relief, etc Behavior  need for help, number of attempts, duration, etc. 3a
  11. 11. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 static dual pathway dynamic Vandewaetere & Clarebout, 2012
  12. 12. Static: prior to/after interaction Eg. Student’s starting level is defined by teacher or pretest Eg. Update of student’s level after a series of completed tasks/exercises – after logging out. Dynamic: during interaction Eg. Update of student’s level after every completed task/exercise Eg. Update of student’s parameters after every interaction with system (hint use -> update in learner model) Dual pathway Eg. Combination: starting level is defined by teacher (pre-defined student model) – during interaction student model is updated WHEN TO ADAPT?3b
  13. 13. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 sequence form content Vandewaetere & Clarebout, 2012
  14. 14. Adaptive curriculum sequencing - From computerized adaptive testing - Based on IRT: - a measurement theory where the probability of a correct answer depends on person characteristics and characteristics of the items.  the difficulty of the items is adapted to the demonstrated level of knowledge  This presumes: • Difficulty level of exercises is known • Skills/knowledge level of student can be tracked and sketched reliably • An item selection algorithm that offers the most suitable exercise (wrt to difficulty) to the learner at a certain time in the learning process, taking account of the learners’ knowledge level. ADAPT WHAT?3c
  15. 15. Adaptive form & content representation - Adaptive content presentation of learning objects  example: varying degrees of support - with or without embedded support (eg. hints) - with several degrees of feedback (eg. from 1|0 to faultspecific) - with or without annotation of co-learners - Adaptive form presentation of learning objects  example: multimodality adjusted according to context - only text when slow connection - no audio in noisy environment - video enhanced with annotations of co-learners when fast connection and much time ADAPT WHAT?3c
  16. 16. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 learner-controlled shared control program-controlled Vandewaetere & Clarebout, 2012
  17. 17. Program or system-controlled System reasoner decides what content is offered. Evaluation:  Lack of choice lowers motivation and fosters dependence (Hannafin & Rieber, 1989)  More risk to become dependent of pre-structuralized instruction (Elen, 2000)  Sparse interaction between learner & environment  High investment and development costs (eg. ITS) ADAPT HOW?3d
  18. 18. Learner-controlled Learner selects what content is offered. Evaluation:  Not all learners can deal with choice: “the art of choosing”  Not for novice, low-motivated learners  Increases learners’ involvement, responsibility and self-regulation strategies ADAPT HOW?3d
  19. 19. Shared control System preselects, learner chooses from preselection. Best of both worlds ADAPT HOW?3d
  20. 20. HOW TO PROVIDE ADAPTIVE INSTRUCTION?3 Vandewaetere & Clarebout, 2012
  21. 21.  Device: Mobile vs desktop adaptive learning  Short duration – long duration  Short term – long term  Environment characteristics (eg. noise on train)  Certification or not  Quality of internet connection  Etc. ADAPT WHEN? CONTEXT3e
  22. 22. ADAPTIVE INSTRUCTION CHALLENGES & OPPORTUNITIES WHAT HOW WHY USE CASES
  23. 23. Adaptive item selection based on combination of judgment and data (What?) [Wauters, Desmet, & Van Den Noortgate, 2012] - IRT: estimation of item difficulty taking into account a learner’s ability.  Computationally intensive – a lot of data required - Proportion correct - ELO-rating system (Brinkhuis & Maris, 2010) - Learners’ judgment “How difficult was the presented item to you?” - One-to-many comparison by learners - Expert ratings “How difficult do you think will this item be for your students” This six techniques all provide reasonably accurate estimates of the difficulty of an item, even with small sample sizes Wauters K., Desmet P., Van Den Noortgate W. 2012. Item Difficulty Estimation: an Auspicious Collaboration Between Data and Judgment. Computers and Education. Pergamon Press nr.58 , pp. 1183-1193 , ISSN 0360-1315 USE CASE 1 – ADAPTIVE ITEM SEQUENCINGII
  24. 24. Illusion of adaptivity might be as effective as adaptivity (to what?) [Vandewaetere, Clarebout, Desmet, 2011]  Adaptive instruction is motivating  Illusion of adaptive instruction is also motivating  Learners’ perceptions are important in the relation adaptive instruction – motivation. USE CASE 2 – ADAPTIVITY & MOTIVATIONII adaptive instruction ↗ learning outcomes perception beliefs motivation Vandewaetere, M., Desmet, P., Clarebout, G. (2011). The contribution of learner characteristics in the development of computer-based adaptive learning environments. Computers in Human Behavior, 27, 118-130.
  25. 25. Learner control as a means to provide adaptive instruction (how?) [Vandewaetere & Clarebout, 2011]  LC has to be perceived by learners:  additional instruction of LC strengthens the perception of control  higher perception of control is related to higher learning outcomes and motivation  Study in language learning – N=165, age 18-20  English tenses – 3 conditions: NC, LC, LC with additional instruction of control USE CASE 3: ADAPTIVITY & LEARNER CONTROLII Vandewaetere, M., Clarebout, G. (2011). Can instruction as such affect learning? The case of learner control. Computers and Education, 57(4), 2322-2332.
  26. 26. Learner control as a means to provide adaptive instruction [Vandewaetere, Clarebout, Desmet, 2011] USE CASE 3II
  27. 27. Learner control as a means to provide adaptive instruction USE CASE 3II
  28. 28. Instruction of Learner control as a means to provide adaptive instruction  Direct effect of instruction of LC on perceptions, which in turn were related to motivation  Additional instruction of control: higher satisfaction with control, higher interest/enjoyment, higher perceived competence and higher interest in learning. USE CASE 3II
  29. 29. ADAPTIVE INSTRUCTION CHALLENGES & OPPORTUNITIES WHAT HOW WHY USE CASES
  30. 30. Adaptive feedback/support (adaptive representation) &LC (What? & How?)) • Shed light on the effect of giving learners control on feedback levels. Different behavior and use of feedback for learners having low and high prior knowledge, and for learners with low and high self-regulated learning skills and motivation. Also, we expect the selection of feedback to be different when learners ask feedback on difficult versus easy items. • Hypothesis 1.1: There is a main effect of learners’ prior knowledge on the type of requested feedback. Learners with low prior knowledge will ask more frequently for detailed feedback as compared to learners with higher prior knowledge. • Hypothesis 1.2: There is a main effect of item difficulty level on the type of requested feedback. After completing a difficult item, learners will ask more frequently for detailed feedback as compared to completing an easy item. • Hypothesis 1.3: There is an interaction effect between item difficulty and prior knowledge. Learners with higher prior knowledge will request less times detailed feedback after completing a difficult item than learners having lower prior knowledge. • four types of feedback: • Faultspecific (wrong/correct + if wrong: correct answer + specific error feedback) • General (wrong/correct + if wrong: correct answer + general attention remark) • Correct answer (wrong/correct + if wrong: correct answer) • Binary (wrong/correct) OPPORTUNITIES: SOME EXAMPLESIII
  31. 31. Adaptive media enhancement (adaptive representation) (What?) OPPORTUNITIESIII
  32. 32.  Different types of enhanced audio-visual input may serve different learning goals (e.g., improve comprehension, stimulate learning of formulaic language). Yet, the visualisation can also be adapted to learners’ profile. • H 1: Different types of enhancement are chosen in function of the learner profile (proficiency level, motivation, etc.). • H 2: Different types of enhancement can be situated on an implicational scale going from less complex to more complex viewing experiences.  If the aforementioned hypotheses are confirmed, adaptive viewing paths (such as the one suggested above) can be established in function of different learner profiles and objectives. L1 subtitles L2 subtitles with L1 gloss L2 subtitles L2 keywords
  33. 33. CHALLENGESIII Big data, big opportunities From manually entering data to online massive storage From self-reporting data to behavioral data From single measurements to longitudinal measurements From inaccessible to everywhere Research domains: -Learning analytics -Educational data mining
  34. 34. Hype cycle for education (Gartner, 2013)
  35. 35. Contact Piet Desmet Piet.Desmet@kuleuven.be http://wwwling.arts.kuleuven.ac.be/franling_e/pdesmet be.linkedin.com/in/pietdesmet @PietDesmet ITEC www.kuleuven-kulak.be/itec

×