Paper ID #7305
Translating Educational Theory Into Educational Software: A Case Study of
the Adaptive Map Project
Jacob P. Moore, Virginia Tech
Jacob Moore is a PhD Candidate in the Department of Engineering Education at Virginia Tech.
Michel Paul Pascale
Michel Pascale is a student at Virginia Polytechnic Institute and State University, and was one of the
developers of the Adaptive Map project.
Dr. Christopher B. Williams, Virginia Tech
Chris North, Virginia Tech
c American Society for Engineering Education, 2013
TRANSLATING EDUCATIONAL THEORY INTO EDUCATIONAL
SOFTWARE: A CASE STUDY OF THE ADAPTIVE MAP PROJECT
Abstract
In this paper, the authors describe the development of an instructional software, where
developers engaged in the process of translating educational theory into a cyber-learning tool,
and the challenges encountered in evaluating its usability and effectiveness of the tool as a
learning aid. Specifically, the authors reflect on their experience in creating the “Adaptive Map”
– an instructional software designed to help students gain conceptual understanding of large
stores of content information. This concept map -based system explicitly shows how discrete
concepts are linked to the whole of the course with a large node-link diagram. This explicit
mapping of expert knowledge structures has been shown to promote conceptual understanding in
students. Because concept maps become visually cluttered and unusable when they get too large
though, an interactive visualization tool was developed to maintain the advantages of concept
maps as learning tools while managing the visual clutter in maps that cover entire courses or
even an entire curriculum. In this paper, the authors discuss the process they took in integrating
the educational literature with the information visualization literature to understand how to best
make an information visualization that addresses educational goals. Results from a heuristic
analysis using Munzner’s four level validation framework for an information visualization are
also presented.
1. Background and Motivation
Technology has the potential to aid instruction, but the simple act of using technology to deliver
instruction does not improve the instruction being delivered [1]. In order to have a positive
impact on student learning, instructional technology developers must draw on what is known
about how people learn and then use technology to improve the quality of the instructional
materials. This often involves collaboration between researchers with backgrounds in education
and those with backgrounds in software, or other technologies. This paper serves as a case study
of one such instructional software development process, the development of the Adaptive Map
digital textbook, so that future collaborations might gain insight into developing instructional
technologies.
1.1 The Theory Behind the Adaptive Map
The Adaptive Map is an attempt to use the flexibility of modern software to improve the design
digital textbooks in ways that was not previously possible. The Adaptive Map tool is a
collection of content pages that collectively serve as a textbook for a course. In its current
iteration, the Adaptive Map (available at www.adaptivemap.me.vt.edu ) has been developed with
content in engineering statics. Statics was chosen for an initial content area because it is a
foundational course for a variety of engineering majors, thus maximizing the potential impact of
the tool.
The central design feature of the Adaptive Map tool is a concept map based navigation system.
Research studies have previously found that by using expert generated concept maps to show
how different topics in lesson are related, students are better able to understand and retain the
content information presented [2–6]. Expert-generated concept maps serve as advance organizers
[7]
and improve understanding by mobilizing relevant prior knowledge the new content can be
cognitively connected too [8].
However, concept maps have problems with scalability. When concept maps become too large
and complex, users encounter “map shock” [9]. Map shock is a cognitive and affective reaction
to large scale concept maps where learners become overwhelmed and disengage from processing
the concept map. This diminishes the effectiveness of large expert generated concept maps as
learning tools. Concept maps are therefore effective for organizing small amounts of content
information in a way that students can better understand it, but they cannot currently be used to
organize large quantities of information. This limits the effectiveness of concept maps as a way
to enhance the effectiveness of course textbooks.
The symptoms of map shock match the more broadly defined phenomenon of cognitive
overload, as described in cognitive load theory [10, 11]. In order to avoid map shock in learners,
the visuals presented must be managed in a way that helps learners manage their cognitive load.
In order to better manage the cognitive load of users, the authors used information visualization
techniques to create the Adaptive Map’s interactive navigation system for large scale concept
maps. This instructional tool was built to test the hypothesis that information visualization
techniques could be used to increase the effectiveness of large scale concept maps as advance
organizers by managing a user’s cognitive load and eliminating map shock.
1.2 Overview of the Adaptive Map Tool
The Adaptive Map software features semantic zooming techniques to enable the user to explore
the large-scale concept map. With this technique the scope of the material being covered
determines the level of detail presented in the visualization. By having the software limit the
amount of information presented at any one time, the software is managing the cognitive load
imposed on the learner.
The tool opens by presenting users with an overview of all the information covered at a high
level of abstraction. At this overview level the topics are grouped into clusters of highly related
ideas, similar to chapters in a traditional book. Each cluster is represented as a node in a graph,
and the links between the nodes represent direct relationships between the topics in each of the
clusters. These links are directed and generally flow from more basic prerequisite clusters at the
top of the screen to more advanced post-requisite concepts at the bottom of the screen. The
links’ line thickness is directly related to the number of direct connections between topics within
those two clusters. An image of the overview can be seen in Figure 1.
Figure 1: Adaptive Map Overview
The user at any level can pan by either clicking on cluster nodes to center them on the screen, or
by clicking, holding and dragging the background. The user can also zoom in or out using scroll
wheel, or by using the + and – buttons on the screen.
If the user zooms in to (or clicks on) any of the clusters in the overview, the cluster node will
break apart into topic nodes and give the user more details on that cluster. Each cluster node
contains several topics, where a topic was defined as the smallest independently teachable
lesson. Information on how the topics and clusters were identified can be found in previous
literature [12]. At this level, the topics are represented by individual nodes in the concept map and
the relationships between the topics are represented by links. A sample screenshot of the “Static
Equilibrium” cluster is shown in Figure 2.
Figure 2: Adaptive Map Cluster View
In Figure 2, the user is focused on the static equilibrium topic, as indicated by the yellow border
around the node and the description of the topic in the node. The background color and all the
topics within the static equilibrium cluster are red to match the color of the static equilibrium
cluster node in the overview. Any topic that is directly related to the focus topic from other
clusters is also drawn in. In this case, the two force member topic is directly related to the static
equilibrium topic, though the two force member topic is part of another cluster. These crosscluster relationships are amalgamated into the links seen in the overview. More details on the
topics or the relationships can be found by hovering over the nodes or link in these views.
If one zooms in further to a topic nodes, one will view the content page associated with that
topic. These topic pages resemble the content in more traditional textbooks, with explanations,
images and worked example problems. A screenshot of part of the “Static Equilibrium” Topic
Page is shown in Figure 3.
Figure 3: Adaptive Map Topic Page
Through these three levels of zoom, learners can explore the topics and the relationship between
topics contained in the adaptive map. The interface helps users maintain a sense of context
within the overall concept map, and helps match the level of detail displayed to learner’s current
level of focus (from a course wide overview, to a chapter overview, to a single topic focus).
2. Software Development Process
2.1 Developing the Software Goals
The software development process began with the formation of software design goals. These
design goals were based on an extensive review of the educational literature relating to concept
maps and cognitive load theory. In addition computer science literature related to information
visualization, particularly the information visualization literature pertaining to graphs, was
consulted. The software design goals and citations of the related literature are listed below.
1. The tool will act as a digital content repository with a concept map based navigation
scheme. The tool must be able to display detailed information as well as concept maps of
the embedded information at several levels of detail and abstraction, and it must allow the
user to easily navigate horizontally (from one topic to another) and vertically (from one
level of detail to another).
2. Usability of the visualization as both a learning and navigation tool is the primary goal of
the software. The information should be displayed in a way that does not cognitively
overload [10, 11] the user.
3. The tool will automatically generate the visualizations based on metadata from the
content developer. The content developer will develop content pages with metadata to
inform the software what other topics a given idea is directly related to, what the nature
of those relationships are, and to what groups the topic belongs. The software should
interpret this information and determine how to best visualize the information
automatically based on the expert-generated metadata.
4. The tool should be as modular as possible. Software and content should be kept separate
so that future content can be developed without understanding the inner workings of the
visualization software. This should lower the barrier to adoption for future content
developers.
5. The software should provide means for the user to adjust the visual presentation settings.
Since many of the ideal parameters for a concept map are learner dependent [13], the user
should be able to change the settings to match his or her needs.
6. The concept map displays should provide good “symmetry” and “predictability.” The
concept map displays should display node-link structures with similar relationships in
similar ways. The displays should also be consistent temporally - if the user leaves a
view and then later comes back to it, it should look the same as it did the first time. Good
symmetry and predictability are shown to enhance comprehension of concept maps and
therefore induce lower cognitive loads [13–15]. Also, small changes to the content should
not radically alter the output. Both of these goals can be accomplished by having
consistent and structured rules for how the concept map visual is built.
7. The concept map should minimize link lengths and link crossings. This will provide
good proximity for directly linked nodes and reduce the mental effort needed to interpret
the links [13, 16-17].
8. The concept map should provide good “continuation”. Concept maps with directional
links that generally flow in a single direction are easier to understand than concept maps
that have links with random or outward directions of flow [13]. This direction of flow
should be vertically oriented for further ease of interpretation [2].
9. The nodes and links should utilize color and shape as indications of different features of
the nodes and links. This should be kept fairly simple, however, and the degree of color
and shape variation should be able to be adjusted by the user.
10. The tool should help the user maintain a sense of context while viewing details, as this is
difficult to do within large scale concept maps [18, 19]. To accomplish this goal, smooth
zooming and panning, multiple window setups, distorted views, or some combination of
these methods will be employed.
11. Upon startup, the tool should display an overview of all the content available to the user.
Introducing a visualization via an overview has been shown to enhance the sense of
context for later displayed information [20].
2.2 Developing the Visualization Software
The software design team then began to develop the software to support these software design
goals. As a first decision, the design team decided to develop the Adaptive Map tool as a Java
applet, as this approach offered many advantages. First, the native cross-platform compatibility
of Java allowed the team to distribute the program to students on all major platforms, without
having to write much platform-specific code. Second, prior experiences with Java and the
widespread availability of free web tutorials about programming with Java allowed all team
members to participate in development process quickly without learning the details of more
uncommonly used software. Finally, using a Java applet allowed the team to host the program on
one central server rather than having users install and have to update software on their personal
computers. This provided all users access via web browser and the centralized location allowed
for simple and efficient updates to both the software and the content.
Once the team chose the programming language, the team selected a tool to use for the visual
aspect of the program. The tool needed to be able to animate many objects at the same time in a
smooth and efficient manner in order maintain a sense of context (software design goal #10) Of
the possibilities that were considered, the “Zoomable Visual Transformation Machine” (ZVTM)
toolkit [21] was chosen. This toolkit provided a simple programming interface that allowed the
team to create and animate objects without much effort. It also provided the source code for all
of its objects, which allowed for easy modification that better allowed the design team to address
a variety of software design goals.
To properly implement design goals 3, 6, 7 and 8, which deal with graph layout, the development
team researched available graphing software. Many free graph layout algorithms have already
been developed and optimized by outside researchers, and it was deemed best to not try to
figuratively “reinvent the wheel” for this particular task. The first graphing toolkit the team
tested was the “Hierarchical Layout Plugin”, developed by Robert Sheridan for the Institute for
Systems Biology [22]. This plugin worked well for getting a basic layout of nodes, but did not
adequately minimize link crossings (software design goal #7). Graphviz [23] was ultimately
selected as it provided a better graphing solution than the Hierarchical Layout plugin, with less
link crossings overall. Graphviz also provided the ability to pre-generate the graph layout
solutions beforehand so that overall layout consistency could be maintained (software design
goal #6). This created an extra step for the content developer when changing the content, but
this drawback was balanced by the decreased load on user’s machines. These pre-generated maps
could be downloaded once upon loading the Adaptive Map and saved, rather than having user’s
machines continually regenerate the new maps as the user was exploring the concept maps.
Finally, it was decided that the content should be stored in the form of HTML and XML pages.
HTML and XML were chosen as ways to encode the content pages and the concept map layout
because of their simplicity. This ensured that content developers not familiar with programming
would be able to translate their concept into a virtual book.
The HTML pages were designed to offer instruction on specific topics and provide relevant
worked examples. Developing HTML is a fairly simple exercise and many tools are already
available to help aid the construction of HTML pages. XML was used to encode the concept
map structure, how all the topic pages were linked together. XML was chosen because of its
simplicity, and because it allowed the content to be read by the Java visualization software while
still keeping the content and software separate (software design goal #4).
2.2 Creating Content for the Adaptive Map
In order to develop the concept map itself, content developers would start by creating a course
wide concept map of the subject area. There are many software packages that aid in creating
concept maps, but the developers for this project simply used Microsoft PowerPoint. A process
for creating large-scale concept maps is outlined in previous literature [12]. After a concept map
is created it is converted into an XML document so that the software can interpret the concepts
and links between concepts. This XML
document captures the structure of the
concept map (what the nodes are and
what other nodes they are linked to),
while leaving the physical layout of this
concept map to be determined by the
visualization software. To further aid
with content management, a simple Java
interface was developed to allow content
developers to more easily create and
manage the large XML documents
needed to represent the large concept
Figure 4: The XML Editor
maps (Figure 4).
After the main features of the program were in place, the team tested many optional features that
the user could adjust (software design goal 5). Some users found the ability to drag different
nodes around useful, while others found that it got in the way. This made it a good candidate for
a user option. Some users that dragged nodes found that the nodes could often get messier than
they had expected. Thus, the team implemented a feature that returned all nodes to their original
positions. Others found the size of the font to be too small, so the team added buttons that allows
users to increase or decrease the size of the font. When the program transitioned from
“Overview” mode to “Cluster View”, some users found themselves zooming out after each
transition, while others enjoyed using the default zoom. A slider-bar was added to allow users to
set a default post-transition zoom level. All of these optional features made the program more
flexible, and allowed users to customize the Adaptive Map to meet their particular preferences.
3. Software Evaluation
The ultimate goal of the textbook is to promote conceptual understanding across a variety of
topics in engineering statics. To evaluate this, the researchers are using data collection sessions
in which engineering statics students use the Adaptive Map tool to study a prescribed topic area
and then explain that topic area to the researchers. Transcripts of these sessions are being
analyzed to get a measure of conceptual understanding. During these sessions, students are also
self-reporting levels of cognitive load experiences while preparing for the explanations. Testing
of this type is ongoing, but a way to more quickly identify design flaws for correction was
desired.
In order for the Adaptive Map tool to be useful as a learning tool, it must first be an effective
information visualization. Students must be able to draw meaning from the concrete
representations of abstract concepts and relationships. In order to evaluate the Adaptive Map
tool as an information visualization, a heuristic analysis of the visualization was performed. For
this heuristic analysis, the researchers used Munzner’s four level information visualization
framework [24].
3.1 Munzner’s Four Level Model for Visualization Design and Validation
Munzner’s model breaks the information visualization development process down into four
stages, with each stage being dependent upon the previous stage [24]. To validate information
visualizations using Munzner’s model, a developer needs to examine his or her design with
respect to a number of threat’s to validity at each level. If the design fails to address the threats
to validity at any of the four levels, the information visualization will not be an effective tool.
Table 1 outlines Munzner’s four level model for visualization design and validation.
Table 1: Munzner’s Level Model for Visualization Design and Validation
Level
Problem Level
Primary Question
Is the visualization attempting aid a
legitimate user task?
Data Abstraction
Level
Is the visualization presenting the
appropriate data to perform the user
task outlined in the problem level?
Data Encoding
Level
Do the visualization strategies and
interaction techniques effectively
communicate the data outlined in the
data abstraction level?
Algorithm Level
Do the software algorithms
effectively implement the
visualization strategies and
interaction techniques outlined in the
in the data encoding level?
Threats to validity
The visualization is aiding a task the user does
not need to perform.
The visualization is aiding a task the user does
not need help with.
The data necessary to complete the user task is
not presented.
The data presented to the user is irrelevant to the
user task.
The data is presented in a way that is difficult to
understand.
The data is presented in a way that makes it
difficult for the user to navigate and explore the
data in a meaningful way.
The software algorithms cannot effectively
execute the visualization strategies and
interaction techniques.
3.2 Evaluating the Adaptive Map Using Munzner’s Framework
The Adaptive Map tool was examined through the lens of Munzner’s framework by both the
developers and by a number of evaluators external to the project. The external evaluators
consisted of 22 students enrolled in and information visualization class, these students were
asked to evaluate the Adaptive Map tool according to Munzner’s model as part of a class
assignment. Because most students in the class were not familiar with the content in engineering
statics, two additional external participants who were familiar the statics content (one
professional and one student) were brought into a class to interact with the tool to aid in the
evaluation. An outline of the heuristic analysis and a summary of the suggestions from external
evaluators is outlined below.
3.2.1 The Problem Level
At the problem level, users of the Adaptive Map tool were encouraged to use the tool as a
replacement to their textbook. Common user tasks consist of looking up information to solve
homework problems, studying for exams, and catching up with the information from missed
classes. In all of these cases, the students are attempting to learn the content information in the
textbook. Concept maps in particular aid in the formation of conceptual understanding [8], a type
of learning that usually lags behind the acquisition of procedural knowledge in mechanics
courses such as engineering statics [25–27].
In general, the external evaluators agreed that the task of learning the material during out of class
hours was a valid task for statics students, the presumed users of the Adaptive Map tool.
Because the tool is filling the role of an existing tool, that of a textbook, the user tasks were well
defined and familiar to the users already. Additionally, several external evaluators agreed that
gaining an understanding of the concepts in a course was more difficult than learning and using
procedures.
3.2.2 The Data Abstraction Level
At the data abstraction level, the developer had to determine if the types of data being presented
were relevant to the identified user tasks. The Adaptive Map presents two primary types of data
to the user, concepts and relationships. Each of the concepts and relationships has many
associated characteristics, but the core data is the concepts and relationships between those
concepts. Concepts and relationships were specifically chosen because this is how people store
knowledge in cognitive schemas in long term memory [28]. Concept maps were originally
developed as a way to externalize these cognitive schemas in learners for evaluation [8], and the
close correlation between the data presented and the way humans remember information speaks
to the validity of the data presented. Overall, the external evaluators found that concepts and
relationships were relevant to learning, but there was also a call for more data types that were not
present.
The first data type that evaluators found was missing was a single linear ordering of the
concepts. Though the directed nature of the links provided some guidance through the content,
there was desire from a large number of evaluators for a linear ordering to the concepts. Without
a “first page”, evaluators were left wondering where to start. This is an artificial data, stemming
not from the way the mind works or from expert understanding but from experiences with
previous textbooks. A number of different orders are valid for learning the topics, but as novices
with a newfound choice on their hands (where to start and what to learn next), the evaluators
were understandably uneasy. Preliminary results from the classroom based evaluation do not
show similar feelings however, most likely because the course schedule provided the information
on where to start and what to learn next. Since the linear ordering would interfere with the
presentation of the relationships between ideas, the suggestion to add a linear ordering is not
immediately being pursued. Developers have taken this into consideration however as a possible
barrier to adoption.
The second data type that evaluators found was missing was user history data. The evaluators
wanted a system to record what pages they visited, what had already been covered in class, and a
system to record notes on the topics. Though some data was available through the back button,
there was a call for more of this data to be recorded and available. Preliminary results from the
classroom evaluation indicate that actual students desire to have this information as well. The
desire for this information may stem from a performance orientation to learning [29], where
students measure their success by checking off what they have done, what they understand, and
how they compare to their classmates. A system of tracking user history and user performance
on integrated quizzes will be added to future versions of the adaptive map tool.
3.2.3 The Data Encoding Level
At the data encoding level, the primary visualization strategy was to use a graph, in the form of a
concept map, to represent the data of concepts and relationships. To navigate and interact with
the graph, a semantic zooming strategy was employed to manage cognitive load. Graphs are the
most obvious and usually the best way to encode sets of data consisting of items and
relationships between those items. The external evaluators generally agreed that the concept
map structure was the best way to display the concept and relationships and that the semantic
zooming technique was a valid way to manage cognitive load, but they did have two prominent
suggestions to improve the data encoding design.
First, the evaluators had trouble learning what all the encodings meant (node color, node
gradient, link thickness, etc.). There was some textual explanation of what these encodings
represented, but the prevalence of this confusion indicated that the explanations were not
sufficient. In order to address the issue, videos were created to introduce new users to the tool
and to better explain what each of the encodings meant. Preliminary results from the classroom
assessment corroborate this initial confusion with the data encodings, but after a few instances of
using the tool, students did seem to be able to comprehend all the data encodings with ease.
The second recurring suggestion the evaluators offered was to improve the overview encoding.
There were still concerns about cognitive overload due to the many overlapping links that were
present in the overview. Preliminary results from the students in the classroom evaluation
corroborate this claim, indicating that the overview was the most difficult view to make sense of.
More design work will need to be done to determine an alternative method for representing the
data in the overview.
3.2.3 Algorithm Level
The data encoding strategy is carried out via a Java applet supported by graph layout and
visualization softwares. Evaluators found the software to be sufficiently fast and efficient in
carrying out the design encoding strategy. The evaluators did find software bugs however, when
refreshing the page and when clicking during a transition animation. These bugs could interfere
with the encoding design under some conditions. These bugs are being addressed to improve the
quality of the Adaptive Map tool.
4. Closure and Future Work
The software development process for the Adaptive Map tool ensured that design choices for this
instructional technology stemmed from the literature on how people process information and
learn (conceptual understanding, concept map, cognitive load theory and information
visualization literature in this case), not what the technology was capable of. By using a learning
centered approach to design, the technology was served as a way to expand the design
possibilities of the instructional tool.
Results from a heuristic analysis with external evaluators served as a way to identify and address
issues with the instructional technology quickly and efficiently before the tool is tested in the
classroom. This analysis has served as a precursor to the forthcoming in class evaluation with
authentic users (students in a statics class).
Analysis of the effectiveness of using information visualization techniques to improve
instructional technology is ongoing. Work needs to be done to more carefully examine the
Adaptive Map tool’s effect on conceptual understanding in students, and furthermore to examine
how many of the design features of the adaptive map relate to student perceptions of the content
and student learning. However, initial results indicate that there is a strong potential for
collaboration between information visualization and education.
5. Acknowledgments
This material is based upon work supported by the National Science Foundation under Grant No.
NSF TUES-1044790. Any opinions, findings, and conclusions or recommendations expressed in
this material are those of the authors and do not necessarily reflect the views of the National
Science Foundation. The authors would also like to acknowledge Lauren Gibboney, Joseph
Luke, James McIntyre, John Nein, and Joshua Rush for their work developing the Adaptive Map
tool.
6. References
[1] T. L. Russell, The No Significant Difference Phenomenon. North Carolina State University, 1999.
[2] D. F. Da sereau, Node-Li k Mappi g Pri iples for Visualizi g K o ledge a d I for atio , i
Knowledge and Information Visualization, vol. 3426, S.-O. Tergan and T. Keller, Eds. Berlin,
Heidelberg: Springer Berlin Heidelberg, 2005, pp. 61–81.
[3] G. W. Ellis, A. ‘ud itsk , a d B. “il erstei , Usi g o ept aps to e ha e u dersta di g i
Engineering Edu atio , International Journal of Engineering Education, vol. 20, pp. 1012–1021,
2004.
[4] M. W. A. “alata, Co ept Maps as Orga izers i a I trodu tor U i ersit Le el Biolog Course,
PhD., University of Virginia, Charlottesville, VA, 1999.
[5] R.-“. “ha , A “tud of Lear i g Perfor a e of E-learning Materials Design With Knowledge
Maps, Computers & Education, vol. 54, no. 1, pp. 253–264, 2010.
[6] M. Willer a a d ‘. A. Ma Harg, The Co ept Map as a Ad a e Orga izer, J. Res. Sci. Teach.,
vol. 28, no. 8, pp. 705–711, Oct. 1991.
[7] D. P. Ausubel, Educational Psychology; a Cognitive View. New York, NY: Holt, Rinehart and Winston,
1968.
[8] J. D. No ak a d A. J. Cañas, The Theor U derl i g Co ept Maps a d Ho to Co stru t a d Use
The , Technical Report Cmap Tools 2006-01 Rev 01-2008, 2008.
[9] J. Bla ke ship a d D. Da sereau, The Effe t of A i ated Node-Link Displays on Information
‘e all, The J. of Expt. Educ., vol. 68, no. 4, pp. 293–308, 2000.
[10] J. “ eller, Cog iti e Load Duri g Pro le “ol i g: Effe ts o Lear i g, Cognitive Science, vol. 12,
no. 2, pp. 257–285, Apr. 1988.
[11] F. Paas, A. ‘e kl, a d J. “ eller, Cog iti e Load Theor : I stru tio al I pli atio s of the
Interaction Between Infor atio “tru tures a d Cog iti e Ar hite ture, Instructional Science, vol.
32, no. 1/2, pp. 1–8, Jan. 2004.
[12] J. Moore, ‘. “. Pier e, a d C. B. Willia s, To ards a Adapti e Co ept Map : Creati g a E pertGenerated Concept Map of an Engineering Stati s Curri ulu , prese ted at the A“EE A ual
Conference, San Antonio, TX, 2012.
[13] D. A. Wieg a , D. F. Da sereau, E. C. M Cagg, K. L. ‘e e , a d U. Pitre, Effe ts of K o ledge
Map Chara teristi s o I for atio Pro essi g, Contemporary Educational Psychology, vol. 17, no.
2, pp. 136–155, Apr. 1992.
[14] “. C. North, I re e tal La out i D aDAG, i Graph Drawing, vol. 1027, F. J. Brandenburg, Ed.
Berlin/Heidelberg: Springer-Verlag, 1996, pp. 409–418.
[15] I. Herman, M. Delest, and G. Melanço , Tree Visualisatio a d Na igatio Clues for I for atio
Visualisatio , Computer Graphics Forum, vol. 17, no. 2, pp. 153–165, Jun. 1998.
[16] D. “. Walla e a d “. C. West, The Effe t of K o ledge Maps That I orporate Gestalt Pri iples o
Learning., Journal of Experimental Education, vol. 67, no. 1, pp. 5–16, 1998.
[17] H. Pur hase, Whi h Aestheti Has the Greatest Effe t o Hu a U dersta di g?, i Graph
Drawing, vol. 1353, G. DiBattista, Ed. Berlin/Heidelberg: Springer-Verlag, 1997, pp. 248–261.
[18] I. Her a , G. Mela ço , a d M. “. Marshall, Graph Visualizatio a d Na igatio i I for atio
Visualizatio : A “ur e , IEEE Transactions on Visualization and Computer Graphics, vol. 6, no. 1, pp.
24–43, 2000.
[19] R. Spence, Information Visualization: Design for Interaction, 2nd ed. Harlow, England: Pearson,
2007.
[20] C. North, I for atio Visualizatio , i Handbook of Human Factors and Ergonomics, 3rd ed., New
York, NY: John Wiley & Sons Inc, 2005.
[21] E. Pietriga, A Toolkit For Addressi g HCI Issues i Visual La guage E iro e ts, i 2005 IEEE
Symposium on Visual Languages and Human-Centric Computing, 2005, pp. 145 – 152.
[22] M. S. Cline, M. Smoot, E. Cerami, A. Kuchinsky, N. Landys, C. Workman, R. Christmas, I. AvilaCampilo, M. Creech, B. Gross, K. Hanspers, R. Isserlin, R. Kelley, S. Killcoyne, S. Lotia, S. Maere, J.
Morris, K. Ono, V. Pavlovic, A. R. Pico, A. Vailaya, P.-L. Wang, A. Adler, B. R. Conklin, L. Hood, M.
Kuiper, C. Sander, I. Schmulevich, B. Schwikowski, G. J. Warner, T. Ideker, and G. D. Bader,
I tegratio of iologi al et orks a d ge e e pressio data usi g C tos ape, Nature Protocols,
vol. 2, no. 10, pp. 2366–2382, 2007.
[23] J. Ellso , E. Ga s er, L. Koutsofios, “. C. North, a d G. Woodhull, Graph iz— Open Source Graph
Dra i g Tools, i Graph Drawing, P. Mutzel, M. Jünger, and S. Leipert, Eds. Springer Berlin
Heidelberg, 2002, pp. 483–484.
[24] T. Mu z er, A Nested Model for Visualizatio Desig a d Validatio , IEEE Transactions on
Visualization and Computer Graphics, vol. 15, no. 6, pp. 921–928, Nov. 2009.
[25] J. Cle e t, “tude ts Pre o eptio s i I trodu tor Me ha i s, American Journal of Physics, vol.
50, no. 1, pp. 66–71, 1982.
[26] L. C. M Der ott, ‘esear h o o eptual u dersta di g i
e ha i s, Physics Today, vol. 37, no.
7, p. 24, 1984.
[27] P. “. “teif, A Arti ulatio of the Co epts a d “kills Whi h U derlie E gi eeri g “tati s, i
Proceedings of the 34th Annual Frontiers in Education Conference, Savannah, GA, 2004, p. F1F–5.
[28] J. E. Ormrod, Human Learning, 5th ed. Upper Saddle River, N.J: Pearson/Merrill Prentice Hall, 2008.
[29] C. A es a d J. Ar her, A hie e e t goals i the lassroo : “tude ts lear i g strategies and
oti atio pro esses, Journal of Educational Psychology, vol. 80, no. 3, pp. 260–267, 1988.
