Abstract
We describe Goal-Based Scenario Builder, a prototype
model-based authoring tool for multimedia educational
software, intended for teachers and curriculum designers.
Keywords:
Educational software, multimedia,
authoring tools, model-based interface tools.
Introduction AND MOTIVATION
"In [the] truly modern School, the computer will be as
much part of all learning as the pencil and book have been
in the past."[7]. For educational software to make a major
impact on education, we need tools that put teachers in
charge of creating their own software, just as they prepare
paper-based lesson plans today. In addition, to help
improve the quality of education, these tools
need to be based on sound educational principles. Goal-
Based Scenario (GBS) Builder is a prototype authoring
tool for multimedia educational software, intended for
teachers and curriculum designers. GBS Builder provides
strong guidance using an underlying educational model of
situated learning, called Goal-Based Scenarios [9]. Using
the model insures that the resulting software is based on
sound educational principles, while freeing educators to
concentrate on curriculum content.
The novel contribution of GBS Builder to HCI is to extend
model-based generation of interfaces as it applies to
educational software: (1) GBS Builder is a specialized,
rather than generic, model-based authoring tool; and (2)
GBS Builder generates the interface interactively with the
designer, rather than automatically.
BACKGROUND AND RELATED WORK
Much educational software is either supplied
commercially, giving teachers no control over content or
teaching style, or built using commercially available
authoring tools (e.g. Hypercard(tm) or Macromedia
Director(tm)) giving teachers no guidance, and resulting in
software that may be educationally unprincipled. The
commercial authoring tools also require too much attention
to low-level detail [3]. Recent research on model-based
automatic generation of interfaces (e.g. [2,4,5,8]) intends
to alleviate some of the limitations of standard authoring
tools by allowing designers to work with a high-level
model of an application and then automatically generate
the interface.
In applying model-based generation to creating educational
software, the key question is: Will these tools generate
software that helps teach based on sound educational
principles? The answer is no for at least two reasons.
First, most model-based generation tools aim for
generality, resulting in interfaces with limited, generic
styles (usually form- or menu-based). This is because they
often use layout rules that map generic domain object types
to graphical widgets (e.g. a choice becomes a check box
widget [2]). By contrast, educational software may require
complex, engaging and highly artistic interfaces, tailored to
the specific underlying educational model.
Secondly, easing interface construction by automation
gives the designer little control over design decisions [5]
and assumes simple, clear-cut mapping between domain
objects and interface widgets. By contrast, instructional
software may be complex, and have numerous design
choices, requiring the human in the loop to design the
software successfully. This suggests a need to generate the
interface interactively, allowing designers control over
intermediate design decisions, while shielding them from
low-level detail.
GOAL-BASED SCENARIO BUILDER
The central tenet of Goal-Based Scenarios is that skills are
best learned in the context of performing a task [9]. In
educational software based on the GBS model, the student
assumes a simulated role, performs a task, and as a
byproduct, acquires the intended skills and concepts.
Currently, GBS Builder embodies a specialized model of
Goal-Based Scenarios, where the mission of the student is
specialized to investigation tasks (as opposed to design or
simulation tasks, say).
We illustrate the key features of Goal-Based Scenario
Builder in terms of an example of a prototype GBS
authored with the tool by a pair of middle school teachers.
The teachers wanted to build a GBS to teach chemistry
skills and concepts, and decided to create an "Arson
Investigator" GBS, where the student, as a forensic
chemist, helps investigate the cause of a fire.
GBS Builder is a specialized, rather than generic,
authoring tool. GBS Builder embodies a specific
educational model, an interface tailored to that model, and
an exemplar piece of software of that model. The
model provides an overview of what the
software is intended to achieve. The investigative GBS
model is divided into five phases: Problem: providing the
student with their role and task; Do: performing the
investigation; Decide: forming conclusions; Communicate:
conveying a decision; and Wrap-Up: seeing the
consequences of the investigation and decision. The
interface is designed to help the student play
their simulated role and perform their task through viewing
movies, running animations, and asking questions of
videotaped experts when needed. To provide an overview
of how the interface relates to the model, the tool provides
a map of how each graphical screen fits into a phase and its
subphases. The tool also provides design rationale to
explain how each screen component fits into the whole.
The exemplar piece of the software makes all
this concrete. The exemplar currently in the tool is a
"Sickle Cell Counselor" GBS [1], where the student
assumes the role of a genetics counselor, advising clients
about the risks of the disease for their offspring.
The teachers authored "Arson Investigator" by copying
and editing sections of "Sickle Cell Counselor" (e.g.,
instead of taking blood to determine the presence of a
Sickle Cell gene, the chemist takes a drape sample to test
for the presence of an accelerant). They were aware,
through the map, of how the interface was linked to the
model (e.g., sampling a piece of drape was linked to the
`extract sample' subphase of the `Do' phase).
GBS Builder generates the interface interactively
with the designer, rather than automatically. The
teachers built each phase of the GBS by filling out each
screen and component with appropriate content (e.g. a
movie of how a drape sample is collected) using the model
and exemplar as guide. This is in contrast to working
solely with the model and then generating the interface
automatically after the fact.
EVALUATION AND CONCLUSION
We first performed a formative evaluation [6] of the tool in
order to improve it as part of iterative design. The tool was
evaluated by 20 graduate students in computer science and
education, building 10 prototype GBS's. These users felt
restricted in copying and editing an exemplar, so we
provided a facility to let them alter their GBS by adding or
deleting any screen or component they wished. As a
result, only 5 of the 10 GBS's kept the gist of an
investigative GBS. One explanation is that it was more
difficult to be self-disciplined in staying to the gist of the
model than to have the tool enforce that discipline
automatically. One successful example is Aztec P.I., in
which the student, as a museum curator, is assigned to
verify the authenticity of an Aztec artifact. The lessons
learned from the first evaluation were that although
copying and editing an exemplar is too restrictive, allowing
unconstrained changes to the interface can turn a GBS into
an arbitrary program (and possibly educationally
unprincipled).
Next, we performed early user testing with a group of nine
elementary and middle school teachers from the Wilmette
school district in suburban Chicago. These teachers used
the tool to build, in pairs, four prototype GBS's in two
months. This time we did not provide the teachers with a
facility to make unconstrained changes to their GBS. As a
result, all four GBS's kept to the model of investigative
GBS. Through focus groups, surveys, and informal
interaction, we learned that although the teachers found the
guidance of an exemplar extremely helpful, they also felt
constrained by it. That led us to redesign the tool to allow
minor cosmetic changes which would not change the gist
of the GBS, and provide more, and varied, exemplars. We
also learned that editing the interface while providing a
link to the model does not always provide a sufficient
overview. We are now providing some high-level design
facilities separate from the interface (e.g. a causal map of
the Investigation).
Goal-Based Scenario Builder is one step towards putting
educators back in charge of education, in order to improve
education.
Acknowledgments
Roger Schank inspired this work. We thank: Bareiss,
Kass, Riesbeck and ILS colleagues for guidance; the GBS
Builder team, notably Feist, Dubach, Williamson, Thorne,
Korcuska, Zielinski, Mostovoy, Legere; Holum &
Guralnick for the Wilmette Seminar; Towle & Smith for
Aztec P.I., Neumann & Clauson for Arson Investigator;
Gomez and Goldstein for comments on earlier drafts.
References
1. Bell, B.L., Bareiss, R., and Beckwith, R. Sickle Cell
Counselor: A Prototype Goal-Based Scenario for
Instruction in a Museum Environment. Journal of the
Learning Sciences, 3(4), (1993-94), 347-386.
2. de Barr, D.J.M.J., Foley, J.D., and Mullet, K. E.
Coupling Application Design and User Interface Design.
In Proc. CHI '92 (Monterey, CA, May, 1992), ACM
Press, pp.259-266.
3. Houde, S. and Sellman, R. In Search of Design
Principles for Programming Environments. In Proc. CHI
'94 (Boston, April, 1994), ACM press, pp. 424-429.
4. Janssen, C., Weisbecker, A. and Ziegler, J. Generating
User Interfaces from Data Models and Dialog Net
Specifications. In Proc. INTERCHI '93 (Amsterdam,
April, 1993), ACM Press, pp.418-423.
5. Luo, P., Szekely, P. and Neches, R. Management of
Interface Design in Humanoid. In Proc. INTERCHI '93
(Amsterdam, April, 1993), ACM Press, pp.107-114.
6. Nielsen, J. Usability Engineering. AP Professional, 1993.
7. Papert, S. The Childern's Machine. Basic Books, 1993.
8. Puerta, A.R., Eriksson, H., Gennari, J.H. and Musen,
M.A. Model-Based Automated Generation of User
Interfaces. In Proc. AAAI ‘94 (Seattle, July 31- August
1, 1994), AAAI Press, pp. 471-477.
9. Schank, R.C., Fano, A., Bell, B.L. , and Jona, M.Y. The
Design of Goal Based Scenarios. The Journal of the
Learning Sciences, 3:4, 1993-94, 305-345.