Dr. Brian Nelson, Arizona State University, presentation on “Design for Learning and Assessment in Virtual Worlds” as part of our Cognitive Systems Institute Speaker Series.

1. Brian C. Nelson
Arizona State University
August 2016
Design for Learning and Assessment in Virtual Worlds

2. Why Virtual Worlds?
 Early take: teaching/training is about information
transmission and uptake by individuals
 More recent view: learning is primarily situated, social
activity of collaborating to make sense of and apply
content and concepts in specific contexts
 Many commercial virtual world-based games are
based on social networks and collaborative problem
solving…in and outside of the game environment
 Virtual world-based games are really good at
collecting data about player activities

3. Why Virtual Worlds (2)?
 Virtual Worlds and other digital media fill every
moment of a learner’s life…until they enter the
classroom
 To students in a media-rich world, the classroom can
feel like a museum
 Virtual worlds engage learners beyond a novelty
effect
 Virtual worlds may support demonstrations of
learning different/beyond that supported by
traditional assessments

4. Educational Virtual Worlds: A study in
Contrasts
1. Well -designed virtual worlds are good for learning
 Many educational virtual worlds are poorly designed
2. Virtual worlds can support innovative assessment of
“21st century skills”
 Many educational virtual worlds use traditional measures
and approaches to assess learning
3. Virtual worlds can support innovative thinking and
multiple ways of knowing
 Policy and cultural issues result in virtual worlds that guide
learners toward homogenous thinking and simple answers

5. My Goals Related to these Issues
1. Virtual Worlds are
poorly designed
2. Virtual Worlds use ill-
suited assessments
1. Theory-based design
for learning in Virtual
Worlds
2. Embed meaningful
assessments
Issue Goal

6. My Goals Related to these Issues
1. Virtual worlds are
poorly designed
2. Virtual worlds use ill-
suited assessments
1. Theory-based design
for learning
2. Embed meaningful
assessments
Issue Goal

7. Challenge: Education researchers focus
(understandably) on curricular and pedagogical
issues
Need: Study design of virtual worlds as
implemented in educational settings
Better Design for Learning in Virtual
Worlds

8. Complexity in Virtual Worlds
 Visual and interaction complexity boosts
immersion/embodiment but may hinder learning in
school settings
 large body of literature on design principles from a
cognitive processing perspective
 Summary: cut the cognitive fat
 Yet...successful virtual worlds are frequently highly
complex, but players can cope with that complexity and
learn
 Pragmatic approach: investigate the use of cognitive
processing based design to balance complexity, short
and long-term engagement, and efficiency

9. Chris Dede, Diane Ketelhut, Ed Dieterle, Jody Clarke-
Midura, Cassie Bowman, and a whole bunch more people!
River City

10. River City
 An Multi-player world to teach scientific inquiry and
content skills to middle school students
 Students work in teams to discover why people in
River City are getting sick
 Students gather data, form and test hypotheses
 More than 20,000 students have taken part

11. River City Origins
 National policy focus on real-world science
practices
 Push to include science inquiry into the classroom
 But…realistic inquiry is difficult to teach and
difficult to learn in the classroom
 Challenge: Create collaborative, situated inquiry
experiences that engage more students in science,
particularly those underrepresented in STEM fields

12. River City Interface: a mess

13. Cognitive Design
 Keep learner focus in the 3-d environment
 Reduce reading through “natural narration”
 Use visual and audio signalling techniques to focus
attention
 Apply spatial contiguity principle
 Design for “essential complexity”

14. Individual investigations of context-based science problems in a virtual world
Example 1: Simlandia

15. Voice vs. Text Chat collaboration
Ben Erlandson (former ASU PhD student) led study
people learn better when words are presented as audio
narration rather than as on-screen text (Mayer, 2005)
 helps reduce a “split attention” effect
Do students completing a science inquiry curriculum in a game using voice chat for
collaborative communication...
self-report lower levels of cognitive load
show better performance on a science learning measure
...than students collaborating via text chat?

16. Results
Cognitive load
 Voice-based Chat: lower levels of perceived cognitive
load
Learning
 almost identical performance for both groups
Why?
 Everyone did well on the pre-test
 Assessment-performance mismatch

17. Diane Ketelhut, Catherine Schifter, Younsu Kim, Uma
Natarajan, Kent Slack, Angela Shelton, (and many more
folks)
SAVE Science

18. SAVE Science
 situated assessment using virtual worlds for
science content and inquiry
 Virtual world-based Game to assess learning
of classroom curriculum in science
 Collect data evolving levels of understanding
 Enable students who don’t do well with
standardized tests to better show
understanding

19. SAVE Science Design Studies (so far)
 Visual Signaling
 Avatar Personalization
 Spatial Contiguity

20. Visual Signaling in Virtual Worlds
 Virtual worlds work well for learning, especially
over long(er) periods of engagement
 Virtual worlds are initially confusing, especially
for novice student gamers
 Low efficiency poses challenges to in-school
implementations
 Low efficiency challenges assessment reliability
and validity
 Visual signaling is used to guide players to
relevant objects and locations

21. Sheep Trouble Module

22. Sheep Trouble Module
 https://www.youtube.com/watch?v=uurufkuXu3s
 Assess students’ knowledge of beginning
speciation/adaptation as well as aspects of
scientific inquiry.
 New and old flocks of sheep
 Determine why recently imported sheep are
getting sick and dying
 Apply understanding of speciation and
adaptation

23. Investigating the sheep

24. Measuring sheep

25. Visual Signaling: Cognition
 Visual Signaling: using visual cues (such as arrows) to
direct learner attention to relevant information on the
screen or page (or virtual world)
 Visual Signaling may lower extraneous cognitive load
and/or increase germane load…letting learners
focus on tasks rather than on interface (Merrienboer,
2008)
 Mixed record in past studies: often found to reduce
self-reported cognitive load, but not always coupled
with improved learning (Morozov, 2009; Chen &
Fauzy, 2008)

26. Signaling Questions
Can the use of visual signaling techniques reduce
perceived extraneous cognitive load in a short game-
based assessment?
Can use of signaling increase the efficiency in a game-
based assessment?

27. Signaling Study
 193 7th graders
 Sheep Trouble: Assessment of Beginning Speciation
 Random assignment: signaling/no signaling
 Lower overall perceived cognitive load (p<.05)
 Increased interactions with sheep (p<.01)
 More measurements taken (p<.001)
 More records entered in notebook (p<.001)

28. Implications and Questions
 Use signaling!
 Why did signaling have a ‘sticky’ effect on
interacting with objects?
 Would the value of signaling for efficiency be
greater in a high search environment? (One with
more visual objects on the screen at once?)

29. My Goals Related to these Issues
1. Virtual Worlds are
poorly designed
2. Virtual Worlds use ill-
suited assessments
1. Theory-based design
for learning
2. Embed meaningful
assessments
Issue Goal

30. Emerging research on:
 Data-mining
 Statistical methods for making sense of learner
actions
Less research on:
 Design of tasks and “Work Products” of assessment
supported by highly immersive virtual worlds
Virtual world-based Assessment

31. Evidence Centered Design for Assessment in
Virtual Worlds
…and/or the Presentation Model aspect of ECD
(Robert Mislevy)

32. Assessment Tasks and Work products
 Assessment in Game-based learning environments:
many researchers and designers focusing mainly on
‘black box’ analysis of data output
 Virtual worlds are designed spaces
 Need full spectrum design for more meaningful
data

33. Assessment tasks in Virtual Worlds
 Virtual world-based tasks support multiple evidence
channels in isolation and in combination
 Provide complex and interwoven collection of work
tools for assessment activities

34. Example: Global Evidence Channels
Location/Movement (LM) Object Interaction (OI) Communication Activities
(CA)
Location tracking
•X location visited
•Time spent at X
•Coordinates
Movement tracking
•Direction
•Speed
•Acceleration/deceleration
•Teleporting
Movement patterns
•Order of movement
•Movement as response
•Movement strings over time
Objects:
•View
•Select
•Click
•Manipulate
•Pickup
•Release
Object Types:
•Artifacts and inventory
•Tools
•NPCs
•Humans
•“intangibles”
•Type
•Speak
•Response selection
•Emote
•In and out of character
•Human and NPC
•Goal-oriented vs. social

35. DATA-MINING BASKETBALL
TROUBLE
Shanshan Zhang and Slobodan Vucetic
Department of Computer and Information Sciences
Temple University

36. Basketball Module
 http://youtu.be/hrZVa2i-e5I
 Assess students’ knowledge of gas laws and
related properties as well as aspects of
scientific inquiry.
 Mid-winter basketball tournament
 Determine why balls at outdoor game don’t
bounce well compared to indoor setting
 Apply understanding of gas laws (air
pressure/temperature link)

37. Basketball Module

38. Study
1. Create automated grading models that predict the
number of embedded assessment questions a
student will answer correctly based on her/his
actions in the module
2. 187 students’ records analyzed
3. Analyze correlations between multiple-choice scores,
within-game behavior, and free-text answers
4. Correlation of .5 (Pearson’s p) with human graders
on predicting performance on in-game multiple
choice and short-answer questions

39. Study
 4 important and non-redundant features found:
 Distinct interactions with in-game objects
 Number of NPCs talked to
 Number of objects whose air pressure was measured
 Number of temperature measurements recorded in e-
notebook
 Key task: discover that a decrease in the temperature
of several gas systems (basketballs and balloons
filled with air) is causing their pressure to decrease.

40. Example
A graphical illustration of how our Classification Techniques are able to learn to distinguish
between Advanced, Proficient, and Basic student evaluations (x indicates incorrect prediction)
• Good predictors for grades are highlynon-linear
• Spherical boundaries approximately indicate the student groups

41. Brian C. Nelson
Brian.Nelson@asu.edu
Questions?