(11) Diversity and Complexity of Ecosystems: Exploring Balance and Imbalance in Nature
 Quicktime Movie ( 360MB / .MOV ) |
Neo Martinez neo@peacelab.net Pacific Ecoinformatics & Comp. Ecol. Lab (copyright holder) Berkeley, CA, USA Ilmi Yoon ilmiyoon@gmail.com San Francisco State University Computer Science San Francisco, CA, USA |
Judges Quotes:
"The animation clearly reveals cyclic variations & suggests interactions among the species; the two views seem to make it easier to quantify & follow through history those variations. I'd be interested in seeing what analysis modes "brushing and linking" between those views would support."
W. Bradford Paley (Designer)
Digital Image Design Inc
"This visualization illustrates the delicate balance in an ecosystem. It allows a user to experiment with the model and track the changes in the food web. It has great educational value because of its interactive nature."
Lada Adamic (Information Science)
University of Michigan
"This is a very clear demonstration of the food chain and how changes to specific organisms influence others throughout the web. This has high educational value as students can see and understand quickly the connections between different life forms, and that the study of individual species requires an understanding of the broader ecosystem."
Daniel Zeller (Visual Artist) New York
"Simulation combined with visualization provides deeper insight into dynamics of food-webs."
Vladimir Batagelj (Computer Science)
University of Ljubljana, Slovenia
Description: This work briefly (~4 minutes) introduces the fundamental ecological concept of a food web while illustrating our latest understanding of the structure and dynamics of these complex networks. The structural data visualized include a food web created by the "niche model" (Williams and Martinez 2000). Dynamic time series data are created by a nonlinear, high dimensional, bioenergic model explored in many of my lab's recent publications (e.g., Williams and Martinez 2004). Visual techniques include 3D Java Open GL program-fed data from highly optimized numerical Java code that efficiently and simultaneously solves up to hundreds of linked ordinary differential equations. The main program used in this visualization employs some of the latest computer science technologies including semantic web technologies for distributed knowledgebases of both empirical and simulated data describing unlimited numbers of species and food webs. The main insights gained from this visualization are an enhanced understanding of food webs as complex dynamic networks along with the fundamental and more rigorous notion of the interdependence of all life including humans.
Scientific Value: The scientific value of this work is both for the introduction of fundamental network concepts in ecology and also as a powerful research tool for computional ecology. We use these visualizations to develop optimized algorithms for arranging network nodes in 3D and 2D so that visual crossing of links is minimized and compelling open source platform is provided for ourselves and others to develop more and improved functionality of the software. Our current research uses the software to simulate the structure and dynamics of thousands of networks that are the base of much of the research being published by our lab. As interest continues to grow in ecological network structure and high dimensional nonlinear dynamics, we will provide our field improved interfaces and knowldegebases so that ecologists without programming skills can effectively conduct sophisticated research in computational ecology.
Educational Value: Beyond the research described above, the educational value of these network visualizations includes providing viewers with a much more compelling and rigorous concept of food-web networks than was available before our visualizations. Simply looking at empirical networks have enabled viewers to accurately guess that the average path length for these networks is about two. However, before 2002 when we published that statistical finding, such short path lengths were an important surprise to researchers. While introductory students gain and retain a clearer concept of food webs, researchers such as ourselves have used these visualizations to detect subtle differences between food-web models and data which directs additional research. For example, they illustrated that the best models including our own underestimate the number of herbivorous species in food webs, something we did not see statistically until compelled so by pronounced discrepancies between network visualization of the models and those of empirical data.
References to Publications:
(12) Moving with Meaning: Visualizing Interaction Data as Dynamic Networks
 Quicktime Movie ( 22.4MB / .MOV ) |
Skye Bender-deMoll |
Judges Quotes:
"Moving with Meaning will give me a new way to daydream during meetings and conferences; now I'll be drawing diagrams in my head while I should be listening!"
Allen Caroll (Cartographer)
National Geographic, Washington D.C.
" 'Moving with Meaning' illuminates how human interaction is like a square dance, with a two steps in, two steps out, dosey do your partner structure and elegance."
David M. J. Lazer (Social Science)
Harvard University
"This entry was extremely well executed. The network model they demonstrated was clearly explained, easy to follow, and elegantly displayed. The movie presented interesting examples of the applications of the model, with not just what was being shown, but how this information is being used (something very few others managed to do)."
Elizabeth Kerr (Science & Technology)
Apple Computer
Description: McFarland has collected a very large dataset of interactions / turns of speech among participants in high school classroom settings. To understand the pace, flow, and structure of the interaction rituals, we needed a tool that would let us get an overview of the networks at various time scales and aggregations.
Network animations are produced by using the SoNIA software to bin the interaction data streams into 2.5 minute slices and apply a custom version of the Kamada-Kawai layout algorithm to position the nodes in the resulting weighted distance networks. The software then interpolates the positions between the time slices and renders the network as a movie in which node movement largely reflects structural change in the network. The visualizations have helped us understand that network processes and effects often have specific time-scales associated with them. They also allow us to better grasp the distinctions between the classrooms when performing the statistical analysis. The same visualization techniques work well when applied to simulations of network change and diffusion - highlighting the crucial effects of edge sequence and timing on transmission.
Scientific Value: Although this specific analysis gave us insight into the structure of interaction in educational settings, the tool is now used much more broadly to produce visualizations of dynamic networks in models of disease spread, friendship networks, and political contributions. Software like this can act as a "browser" for time based relational data, putting it into forms that allow researchers to explore large quantities of data in ways that can leverage their visual intuition and pattern recognition ability.
Educational Value: Animations of network processes can provide concrete examples which help people grasp the distinctions among various network effects. They also provide a tangible reference point for scholarly communication about dynamics, allowing researchers to do something more than just wave their hands. Network movies also provide efficient ways of summarizing and conveying the salient properties of a complicated network dataset to an audience.
References to Publications:
Related Projects:
http://sonia.stanford.edu
http://www.stanford.edu/group/sonia/examples/index.html
http://sourceforge.net/projects/sonia
(13) Competition Dynamics in Chicago Public Schools
 Quicktime Movie ( 2.3MB / .MOV ) |
Eytan Bakshy Northwestern University |
Judges Quotes:
"This visualization shows quite clearly the outcomes of Chicago's school choice program. Poorly performing schools are losing their good students to better performing schools. It also manages to show many things at once in a cohesive picture - the affluence of the neighborhoods where the schools are located, the schools' performance, the students' performance, and the flow of students from their assigned schools to schools of their choosing. One could evaluate the policy outcomes from this visualization. Now that's a sign of a good visualization!"
Lada Adamic (Information Science)
University of Michigan
"A network visualization with a genuine social purpose. Understanding the structural patterns of education inequity in terms of a network, and also how this plays out geographically, provides people with much needed insight into otherwise hard to comprehend statistical tables. I can see such visualization as becoming potent tools to leverage for political change in the future."
Martin Dodge (Geography)
University of Manchester, UK
"This animation clearly tells a story that might otherwise remain hidden in data columns."
Ulrik Brandes (Graph Theory)
University of Konstanz, Germany
Description: This is a visualization of student enrollment patterns from the Chicago Public Schools (CPS) “open enrollment” program – a program that gives outgoing 8th grade students a large amount of flexibility in choosing a public high school to attend within the school district. Open enrollment is based on the idea that giving households more choice creates the competitive incentives necessary for system-wide improvement. Opponents to school choice, however, point out that a potential downside of such programs is that only the best students take advantage of choice, further harming the students that get left behind.
This visualization analyzes the extent to which student enrollment
patterns reflect the hopes and fears of the various stakeholders in school
choice reform. More specifically, we create a visualization to track 9
cohorts of 8th grade students in CPS from 1995-2004. Schools in the
visualization are represented as nodes in a network and placed according
to their geographic coordinates. The flows of students are represented
by links between the nodes. The network is overlaid on a contour plot of
social status census data from Chicago in 2000.
The visualization highlights several important patterns of choice in
Chicago. With respect to inducing competition, we see that students
assigned to schools with low standardized test scores flee those schools for
higher achieving schools. We also see that students assigned to poorly
performing schools flee them in bursts when new schools are created
nearby. With respect to the distributional impact, examining the link
colors shows that the higher achieving students within an assigned school
are consistently the ones to choose to go elsewhere.
Scientific Value: Many current public education reforms propose giving
parents and students the ability to choose schools that go beyond the
traditional assigned neighborhood school. Proponents of choice-based
reforms claim that giving parents the ability to choose the school their
children attend provides both access to better schooling to the
disadvantaged populations as well as the incentives necessary for school
reform. On the other hand, opponents worry that school choice will have
adverse distributional consequences, with only the best students from the
most involved households actually exercising choice. To the extent that
the presence of higher achieving students helps current schools, the
students left behind – likely the students in need of most help – will
be even worse off than before.
Underpinning both sides of the argument are assumptions about changes
in enrollment patterns brought about by school choice; or more
precisely, assumptions about how individual decisions give rise to
district-wide patterns. By documenting and animating these emergent patterns for
one of the largest districts in the country, we can begin to glimpse
into both the underlying processes that give rise to the enrollment
patterns, as well as examine whether those aggregate patterns are consistent
with the policy stories.
Educational Value: Whether school choice “works” and for whom, is a
very unclear issue from both research and policy perspectives. Indeed,
deciphering the impact of current small pilot projects, or analyzing
data that takes advantage of variation between school districts and
states, often requires sophisticated research techniques not readily
accessible to a large range of stakeholders involved with this issue. One
contribution of this work is that we conceptualize a school district in
terms that a variety of stakeholders can relate – as network of schools
with annually changing flows of students between them. Our ultimate goal
of this work is to provide a platform for which researchers,
policymakers, and households can better communicate and ultimately make more
informed decisions in improving the quality of education for thousands of
students.
References to Publications: N/A
Related Projects:
http://ccl.northwestern.edu/edpolicysim/ed-policy-sim-intro.htm
(14) The Evolution of Malaysian Exports
 High Res Image ( 35.8MB / .TIFF ) |
Cesar Hidalgo |
Judges Quotes:
"There is a lot of innovative thinking on how to use metrics and information in a network perspective that is condensed in this visualization. The final result is highly explicative and can be repeated and generalized to other countries."
Alessandro Vespignani (Internet Research & Epidemics)
School of Informatics, Indiana University
"This is another instance of a visualization that beautifully illustrates a point while tracking the time evolution of a network, the point being that a country exporting one type of product is more likely to expand its portfolio of exports to 'similar' products in terms of production methodology and resources."
Lada Adamic (Information Science)
University of Michigan
Description: We needed a way to visualize the proximity matrix
which is constructed through an outcome based way to study the
similarity between products.
We used export data from the Feenstra trade flows available from NBER.
Because of the structure of the space we consider a sample of the
matrix constituted by the maximum spanning tree and all links with a
proximity higher than a given threshold. This represents less than 1% of the
links but ends up being an accurate representation of the structure of
the full matrix. A striking feature is that the space is modular and
has a core-periphery structure which affects the development of nations.
Scientific Value: Countries move along the product space by spreading to
nearby nodes. In the figure the products which Malaysia exports are
denoted with black squares and it is clear that between 1985 and the year
2000 Malaysia spread through the electronics sector, which is the light
blue cluster located at the bottom of the network. The product space
represents a way to study the evolution of a country's export basket and
explains why some countries develop while other ones get stuck in their
diversification process.
Educational Value: The product space teaches us that when a country
wants to develop a new industrial product its ability to do so depends
strongly on the products that it has already developed. This allows
tailoring specific industrial policies given the location of a country in the
product space. The location of rich and poor countries is extremely
different showing that the same development strategies should not work for
both of them.
References to Publications: N/A
Related Projects:
http://www.nd.edu/~chidalgo
(15) Persistence of Social Ties
 High Res Image ( 28.2MB / .TIFF ) |
Cesar Hidalgo |
Description: There are currently no off-the-rack methods to
study and visualize several social networks panels. We propose persistence,
the probability of observing a tie when looking at a panel of a given
duration, as a way to summarize the temporal stability of a tie. The
submitted figure shows the all nodes and ties up to three degrees away
from a randomly chosen agent in the mobile phone network. The picture
illustrates the coupling between topology, socio-demographics and the
stability of ties. The image summarizes (1) the structure of the mobile
phone network (2) the stability of its ties [which is a temporal attribute]
(3) the gender of the agents and (4) their age.
Scientific Value: None given
Educational Value: None given
References to Publications: N/A
Related Projects:
http://www.nd.edu/~chidalgo
(16) The Global Network of Terrorism: Dynamic Trends from 1969-1997
 High Res Image ( 3.8MB / .TIFF) |
Adam Perer |
Judges Quotes:
"Networks are best read if they are not only 'technically accurate' and visually attractive but when they employ a type of rendering that creates a landscape. That creates a bridge for the uninitiated audience to cross into the field of expertise. Dataland travels have now become so enjoyable; they may soon appear as special fare destinations at a travel agency near you. Perer's visuals make that trip into the land of terror networks absurdly attractive. Having intellectual entertainment and visual pleasure with terrorism analysis is perhaps one way to diffuse the very essence of terror - by analyzing it without being terrified. And in the end it leads to a hopefully more rational dealing with it, which is the opposite of what terrorism is trying to instill."
Ingo Günther (Journalism & Art)
Tokyo National University for Fine Arts & Music, Japan
Description: Our data was generated from a database of over
70,000 terrorist attacks ranging from 1969 through 1997. We created a
social network from this database based upon co-occurence of attacks (e.g.
if two terrorist groups attacked in the same country in a given year, a
link was drawn). In this submission, two visualizations are used. The
first is stacked timeline visualization that shows the evolution of the
network. The second are standard node-link diagrams using a
force-directed algorithm. Since node-link diagrams tend to become unreadable, we
use filters to only highlight important nodes (high degree), and use
convex hulls to point out communities.
Scientific Value: The social scientists collecting this massive database
of terrorist attacks had not been using visualizations to analyze their
data. However, after becoming partners, they were able to confirm
certain hypotheses, think of new ways to use the data, and find
coding errors from the data collection phase. The ability to quickly
interact, search, filter, and visualize the data has proven to be valuable.
Educational Value: By combining typically messy node-link diagrams with other more comprehensible visualizations, users can be guided toward insights. The ability to zoom and filter allows users to untangle the messy network visualizations according to attributes they care about. Although these visualizations were colored and sized according to a simple network metric (degree), these visualizations are also very expressive if using social network metrics such as betweennness centrality, power centrality, and so on.
References to Publications:
(17) Ideological Alliances on the Supreme Court: Visualizing Co-Voting Data
 High Res Image ( 4MB / .TIFF ) |
Peter Hook |
Judges Quotes:
" Rigorous - Slightly muddled but intelligently laid out."
Peter Christensen (Art Curator)
Museum of Modern Art, New York
Description: This work sheds light on an important branch of the
United States Government—The Supreme Court. Its intended audience is
law and political science students and anyone wishing to learn more
about the Court. The network visualized is the co-voting frequencies of
the nine justices of the Supreme Court over 50 years. The data was
collected from the annual review of the Court’s Term in the Harvard Law
Review and visualized using Pajek. Dynamic elements portrayed include the
frequently changing composition of the Court and the changing
ideological alliances that result from these changes. The main network
visualization is a spatial distribution of all Justices over the 50 year time
span based on their co-voting frequencies. There is an implied element
of time moving from left to right. Those serving longer time periods
are drawn more to the center while still remaining adjacent to their
ideological compatriots. The largest image serves as a base map or
common reference frame on which to layer additional information. For
instance, each of the specific courts can be visualized by graying out
the Justices that did not serve on that particular court. In an
interactive environment, this could be the front end to a digital library or
an online learning module in which students explore the specifics of a
particular Court (major cases, major themes, major divisions and
alliances, the human stories behind the cases, etc.). This concept is
illustrated by the three visualizations on the bottom right.
Scientific Value: The value of this visualization is pedagogy. There is
a need for commercial study aids that contextualize the work of the
Supreme Court and make it more real to students. The visualizations
provide a cognitive scaffold on which students can hang a more rich
understanding of the Court based on more detailed study of Supreme Court
cases.
Educational Value: The Justices are represented by their photo icons and
distributed in space based on how often they vote with their
colleagues. This spatial layout is particularly effective in illustrating
ideological divides and alliances. In one example, the swing vote status of
Justices O’Connor and Kennedy is made clear using network layouts.
Developed as a rich interactive learning environment, these images hint at
the potential for a rich online departure point from which to learn
more about the Court. Students learn who the members are, when they
served, and with whom they are most likely to agree and disagree. Layered
with more case specifics and topic maps, the images provide powerful
study aids to those learning about the Court.
References to Publications:
(18) Epidemic Pathways
 High Res Image ( 51MB / .TIFF ) |
Vittoria Colizza |
Description: Map representation of epidemic pathways. The maps
are built on the simulation results of a data-driven stochastic
computational model aimed at describing the SARS spread across the globe by
airliners and analyzing the resulting epidemic pattern with respect to the
complex features of the airline network. The model explicitly
incorporates the complete International Air Transport Association database
including airline connections between the airports of the world coupled with
travel fluxes and census data for 3,100 urban areas located in 220
countries. The virus can be transmitted among individuals living in the
same city and can propagate to a different city as carried by air
travelers. The highly complex nature of the airline infrastructure is shown in
the main panel of the image. The simulated spatio-temporal evolution of
the disease is represented through a color code which indicates the
week of onset of the epidemic in a given country, since the start
of the outbreak in Hong Kong, chosen in agreement with available
historical data. The resulting disease evolution is found to be in good
agreement with the historical data and very robust with respect to the
intrinsic uncertainty in the virus transmission and in the travelers’
behavior. Detailed statistical analysis monitoring the path followed by the
virus shows the emergence of epidemic pathways, i.e. preferential
channels along which the epidemic will more likely spread, selected out of
the huge number of possible paths the disease could take by following
airline connections. Based on models’ results, it is therefore possible
to identify these pathways by defining a probability associated to a
given path of infection. Epidemic pathways are represented in the maps
with arrows whose thickness relates to the probability associated to a
given virus propagation path.
Scientific Value: The SARS outbreak in 2003 has clearly demonstrated how
the highly interconnected nature of our world can be a major
disadvantage against the large-scale propagation of emerging infectious diseases.
In particular, the air transportation infrastructure plays a major role
in shrinking distances around the globe, by connecting far apart
regions and allowing infectious travelers to potentially spread the disease
to different geographic areas in a relatively short time. In this
context detailed computational approaches can be used as an effective risk
assessment tool and the visualization of obtained results can provide
useful insights in the understanding of the geographical and temporal
evolution of an infectious disease. The visualization itself can lead to
new insights. Mapping the most probable routes of virus propagation,
coupled with dynamical data on the invasion process, provides fundamental
information with a crucial impact in the design of non-medical intervention measures such as travel restrictions in similar
propagation events.
Educational Value: The opportunities and power of Geographic Information
Systems (GIS) are expanding rapidly in various research fields and the
detail which was previously only approached by GIS professionals is now
available on our personal computers. Mapping techniques integrated with
multiple network layers and longitudinal data converge in a truly
multidisciplinary language fundamental for the understanding of epidemic
processes where the spatial aspect is a crucial ingredient. Given the
availability of new technologies, the integration of dynamical spatial
thinking into college curricula could provide added value to traditional
epidemiological courses and also the ability to explain key aspects of
epidemic evolution by simply using a unique image.
References to Publications:
(19) Visualizing Dynamics in Markets
 Quicktime Movie ( 50MB / .MOV ) |
Michael Blume |
Description: The context of the work is a monitoring application
to track fraudsters in markets. Markets are often subject of
manipulation since many people want to increase their wealth or power in an easy
way and therefore use every given chance to improve their portfolios.
Therefore market operators have to supervise and monitor market activity
in realtime.
The dataset used for this contribution is taken from the sport
prediction market STOCCER where people were trading their expectations about
the results of the Soccer World Championships 2006. The graph is
constructed the following way:
- User accounts are represented as nodes having as an attribute the
total traded volume (price * quantity for each transaction).
- A transaction, where A is selling to B, is represented as a directed
edge from node A to node B. Edges have two attributes: timestamp of the
last transaction (edge serve as aggregation for one or several
transactions) and the total volume (price * quantity) flowing through the edge.
The visualization contributed to the competition is only one element of
a larger monitoring and detection application. The goal for the
contributed view is to get a clue of the corresponding actors causing certain
market effects observed in other views (e.g. price/index charts and
stacked volume charts).
Insights gained by this view of the application are which group of
traders caused certain price movements and which groups of traders are
together online and dealing with each other. This is particularly
interesting if already some fraud cases are identified. Previous work of Cortes
and Pregibon in 2001 in the field of telecommunication found that
fraudsters have a smaller average distance to other fraudsters. The
application allows the user to mark nodes as fraudsters, which can be
displayed highlighted in the graph.
Scientific Value: The problem of common graph displays is that after
reaching a critical size the graph is not very easy to monitor because of
the visual complexity (many elements) and the tendency to visual
clutter (depending on the graph and layout algorithm). But clutter hinders
the user in finding the active group of traders and their relationship.
In the best case we would like to end up with clusters of people who
often traded together, though their nodes may not be connected in that
very moment.
To achieve the clustering, we use the logarithmically transformed
volume attribute of the nodes as the mass for the gravity-based spring-layout algorithm. This means that nodes, which have traded a lot more have a
higher mass and thus move slower than nodes with a low trading
volume. To prevent the visual clutter we have to get rid of inactive nodes
and older edges. Therefore we filter the edges with a sliding window
technique using the last transaction timestamp attribute. Edges outside
the window will disappear. Due to the missing attraction force the node
will move slowly away from the group and thus reduce clutter.
Educational Value: This contribution is driven by two ideas. First,
markets have besides the commonly considered price and information
aggregation features a social aspect: who traded with whom! To make use of the
highly dynamic market graph, which is changing continuously over time,
we need a meaningful dynamically filterable visualization. The
contribution may only be seen as a first step, since many more features can be
thought of (e.g. opening the filter for nodes selected by the user, to
show all transaction partners and not only recent, changing the edge
color depending on which shares are traded at the moment, changing node
size depending on age, wealth or activity, and so on). The ordinary
graph distance filters are perhaps only a beginning for more creative
filters in the field of dynamic networks.
References to Publications:
N/A
Related Projects:
http://www.stoccer.de
(20) An Emergent Mosaic of Wikipedian Activity
 High Res Image ( 79MB / .TIFF ) |
Bruce Herr |
Judges Quotes:
" Bruce and Todd have shown a volcanic landscape with lava pools, geysers and crusted-over areas. So far the best representation out there to show what moves mankind's minds. A true mindscape of the public."
Ingo Günther (Journalism & Art)
Tokyo National University for Fine Arts & Music, Japan
" The idea is brilliant and the technique used for the visualization conveys a wealth of information."
Alessandro Vespignani (Internet Research & Epidemics)
School of Informatics, Indiana University
" The mosaic stunningly illustrates the broad spectrum of what I would call the diffuse focus of the masses. Its value is in its all-encompassing overview, and that it allows one to explore and compare this focus. It would be interesting to see how it changes over time, my faith in humankind would be restored to someday see that Albert Einstein and Muhammad generated more interest than Britney Spears."
Daniel Zeller (Visual Artist)
New York
"I liked this entry for being very unique, topical and fun to explore. They were able to display a great deal of information, multi-layered information, with a flat image. Something that stands out among the entries. This would be fascinating to see animated over a time course as well!"
Elizabeth Kerr (Science & Technology)
Apple Computer
"A very nicely realized mapping that is both visually interesting and conceptually valuable. Providing such 'high-level' overviews of large, dynamic social information spaces as a single image is really useful to answer people's basic question 'what does Wikipedia look like?'. In much the same way as Marc Smith's Usenet overview treemaps captured the sense of the landscape of the conversation, Herr and Holloway's shows us the shape of open source knowledge creation. The layout of such large volumes of data seems effective, but also pleasing to the eye. And it left me wondering what would happen if you could spatially morph this onto a conventional world map to show the geography of authorship of this newly emerging Wiki world."
Martin Dodge (Geography)
University of Manchester, UK
Description: This visualization provides a macro view of activity across all of Wikipedia. This visualization was created by first computing the similarity of all pairs of Wikipedia articles using co-citation analysis. These similarities were then used to generate coordinates using the VxOrd layout algorithm. As a novel means of labeling this 'map', we divided the space into a grid. The nodes in each tile of the grid were ranked according to in-degree and the first image of the highest ranked page containing an image became the background of the tile. We then overlaid the laid-out network on top of the images (with transparency). Further, we size and color coded the network based on current editing behavior. The larger darker nodes represent articles that are being worked on furiously, while the smaller yellow nodes have much less edit activity. Edit history was pulled and parsed from dumps provided by the WikiMedia foundation.
This visualization provides a means for active Wikipedians and passive users to see activity going on in Wikipedia at a global scale. Wikipedians can see what is currently being worked on and either police them or join in. The mosaic itself serves to help everyone get a feel for what areas are covered in Wikipedia and to provide striking landmarks for quickly finding related articles.
Scientific Value: Such a macro view of activity in Wikipedia may indicate where researchers and educators should consider looking for bias or fraud. This map could be used as a basis for further study of trends in Wikipedia. Commercially and scientifically, the techniques used could be applied to other large scale datasets with large amounts of related image data, providing a global view and a way to show trends at such a scale.
Educational Value: This poster may serve well in middle or high school level classrooms as a picture mosaic of human knowledge and current societal interests. It might be used to spark debate about why certain topics are emphasized over others, as well as to motivate expanding under-represented topics. Further, it could serve as a way to build awareness of network science in all levels of education and to the masses.
References to Publications: George S. Davidson, Bruce Hendrickson, David K. Johnson, Charles E. Meyers and Brian N. Wylie (1998) Knowledge mining with VxInsight: Discovery through interaction. Journal of Intelligent Information Systems, 11(3): p. 259-285.
Related Projects: N/A