Conceptual Wiki Page Simulation
A Discrete Space Agent-based Approach
Roger W. McHaney and Jonathan Mossberg
Management Information Systems, Kansas State University, 101 Calvin Hall, Manhattan, KS, U.S.A.
Keywords: Agent, Simulation, Discrete Space, Wiki, Anylogic, Agent-based Simulation.
Abstract: This paper describes the preliminary development stages of an agent-based model (ABM) used to understand
and anticipate changes to Wiki pages. A discrete space approach was used to structure the model. Letters from
words in the wiki were represented as agents which could be changed, deleted, or added based on rates derived
from wiki page histories. A C# pre-processor, called Wiki-Hist-Heist, was developed to facilitate analysis of
existing wiki page histories and provide model inputs based on detected patterns and resulting distributions.
The conceptual version of the Wiki Page ABM was built using AnyLogic. It provided a framework for user-
friendly features which allow easy changes to inputs so a variety of pages and scenarios can be modelled.
Additionally, this project illustrated the usefulness of ABM in this domain. Limitations and future study
directions are included.
1 INTRODUCTION
The overarching purpose of this research was to
develop an ABM (Heath et al., 2009) to help
understand, anticipate and ultimately manage wiki
operations more efficiently and effectively. “Wiki” is
derived from the Hawaiian word for “quick.” This
term is meant to imply a technology which facilitates
the rapid changes of web pages resulting from a range
of contributors. In a modern sense, a wiki is a web site
technology used to facilitate mass collaborative
authoring, editing, and sharing (Mader, 2008). In its
purest form, Wikis are open to all users and permit
participants to create, edit, delete, and link to pages
using a standard web browser without the need for
specialized training (McHaney et al., 2013). Wikis
serve many practical purposes. For instance, they may
offer a venue for displaying useful content or become
an organization’s information repository.
Additionally, wikis may provide convenient ways of
enabling collaborative synergies through
asynchronous interaction.
Wikis support a variety of media including
images, audio files, videos, text, hyperlinks,
embedded widgets, and other standard Web page
features. Wikis can be open access or controlled with
permissions to carefully regulate user privileges
(García, 2012). The most popular example of active
wiki technology is Wikipedia which has become the
world’s de facto virtual encyclopedia. Wikipedia was
developed by participants from across the globe who
contributed to a variety of informational subjects by:
sharing research, editing, translating, updating,
correcting, and maintaining source pages.
Wikis are not without drawbacks. For instance,
the use of a wiki requires careful oversight to ensure
contributions are consistent with a page theme. Since
most wikis are developed as venues open for
contributions, they have an inherent vulnerability to a
variety of misuses. Vandalism, spam, trolling,
commercial hijacking and other issues can emerge as
major problems (The Computer Language Company,
2011).
Depending on characteristics such as level of
oversight, size, scope, structure, and importance,
intentional vandalism and unintentional errors may
not readily be detected. This problem, in part, has
prompted the current investigation. Our research
premise becomes: Can insight into wiki page editing
patterns be gained through development and analysis
of an agent-based model (ABM)? Further, this study
describes a generic method whereby users can easily
change model inputs so their specific interests
regarding a wiki can be examined. The remainder of
this study provides a look at wikis and describes the
process used to develop a user-driven model of wiki
page changes.
393
McHaney R. and Mossberg J..
Conceptual Wiki Page Simulation - A Discrete Space Agent-based Approach.
DOI: 10.5220/0005506003930400
In Proceedings of the 5th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2015),
pages 393-400
ISBN: 978-989-758-120-5
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
2 WIKI OVERVIEW
As stated earlier, without careful oversight, changes
to Wiki pages may go unnoticed. Although most wiki
software, including the leading software system—
MediaWiki---provides tools for monitoring and
reporting changes, a long-term commitment to
maintaining oversight becomes a time commitment
that is difficult to anticipate and manage. Based on
experiences with wikis at Kansas State University, we
often asked, “How much time is needed to properly
oversee a wiki?” The webmaster or more
appropriately phrased, wiki keeper, has emerged as a
person with key roles for conducting this oversight
(McHaney, 2012). The wiki keeper must spend his or
her time reviewing changes then take corrective
action to undo any problematic entries (Sutton, 2006).
Historically, wiki keepers tend to be non-
confrontational to encourage user contribution and
interaction rather than provoke retaliatory responses.
Wiki keepers often use a soft security approach to
protect the wiki, its users, and preserve informational
integrity from injurious actions (Meatballwiki, 2011).
These defenses must tread the fine line between
offering protection and preventing legitimate users
from unnecessary constraint. Many wiki keepers
gradually ramp up action as trolling and spam
becomes more troublesome. Many wiki platforms,
such as MediaWiki, offer features that make undoing
damage easy by offering change rollback. Often,
vandals will attempt to test a wiki’s oversight
practices by making subtle changes and then gauging
the response. If no response is detected, then the wiki
might be used for posting larger amounts of
information related to the vandal’s agenda. Ignoring
vandalism generally is not an option because
contributor time is wasted and inaccurate information
can ruin the reputation of the wiki (McHaney, 2012).
2.1 Wiki Keepers
A wiki keeper’s primary role, particularly in small,
open wikis, is that of a content editor. As new content
is added by users, grammar, style, word choice, and
formatting may be inconsistent or substandard. It falls
to the wiki keeper to encourage other users to take on
editorial roles. If this does not occur, the wiki keeper
must personally make required edits. Even if others
do become editors, their work must be reviewed
periodically to ensure consistency. Depending on
contribution quality, a substantial time commitment
may be required. Higher numbers of users beget more
changes. This leads to another task: the wiki keeper
will need to view terms used as indexing tags and
update the ontology (Hai-Jew and McHaney, 2010).
It is in the best interest of a wiki to have tags with both
a “conceptual consistency” and a “syntactic
consistency.” (Hepp et al., 2007, p. 55).
2.2 Wiki Editing
In most wiki systems, including MediaWiki-based
implementations, pages are created as plain text with
basic formatting symbols or strings. Usually, wikis
track change histories of a document reflecting both
the time of change and the magnitude of the changes
in terms of a character count. Each time a collaborator
makes changes to a wiki page, the newly revised page
becomes the current version. Older versions of the
document can be reviewed, compared side-by-side
with the current or older versions, and inappropriate
edits can be "rolled back." This is convenient because
the current project relies on the use of history
documents to provide inputs for the Wiki Page ABM.
Figure 1 provides an example of a page history from
a Wikipedia page titled “Wilson Sawyer.”
3 MODEL DEVELOPMENT
The model created for this project is loosely coupled
to Wikipedia. Its page histories provide a data source
that drives model inputs. Although this article uses
Wikipedia sources, it should be noted that the model
pre-processor can derive input data from page
histories for any MediaWiki-based wiki site. Minor
modifications to the pre-processor’s code could
extend this functionality to nearly any other wiki site.
3.1 Model Pre-processor
A model pre-processor, called Wiki-Hist-Heist, was
custom developed using C# in the Visual Studio
development environment from Microsoft. The
software prompts a user to enter the name of the
Wikipedia page and then processes the history to
derive data representing frequency and magnitude of
changes to a page over time. Outputs are written to a
CSV file where the data can be further analysed with
spreadsheet functions and statistical tools. Figure 2
provides an image of the user interface for the
program.
Currently, the software includes all changes to the
history page regardless of source or nature. Future
revisions are planned to enable filtering on items
included in the data set. For instance, talk page
changes will have the option of being excluded.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
394
Figure 1: Example of wiki page history which can be pre-processed for ABM inputs.
Figure 2: Wiki-Hist-Heist user interface.
Figure 3 provides an image of the CVS file
containing the data generated by Wiki-Hist-Heist. For
this project, the contents of the output CVS files were
analysed with EasyFit Professional from MathWave
Technologies to provide change frequency and
magnitude distributions for use as model inputs.
Figure 3: CVS file contents derived from a Wikipedia
history page by Wiki-Hist-Heist.
These distributions were used to drive agent
activities within the Wiki Page simulation. Figure 4
provides a representative view of candidate
distributions derived from the character change
magnitudes per wiki edit. The output distributions
were generated and input into the ABM. The
character change magnitude data from the “John
Adams” wiki page, as shown, was found to most
closely resemble a Cauchy distribution (p=.10) in this
example.
Figure 4: Representative output from EasyFit Professional
from MathWave Technologies.
3.2 ABM Approach
An ABM worldview was used to construct the basic
units of activity for this simulation project (Bruch and
Atwell, 2013; Macal and North, 2010; Taylor, 2014).
The primary agent population, wikiletters, was
modelled using a discrete space approach where the
letter agents were represented in the form of a
rectangular grid of cells. Each cell was one letter on a
wiki page and became a wikiletter agent with unique
identities, parameters and system states. The grid was
designed to be variable in size. This allows modelling
wiki pages of different sizes. A second agent type was
also built into the model. This agent was the wiki
editor, represented by a single entity with the power
to affect changes on the wikipage. In other words, the
editor agent, driven by data from the Wiki-Hist-Heist
program, periodically interacted with the grid based
on derived interarrival frequencies, and made changes
ConceptualWikiPageSimulation-ADiscreteSpaceAgent-basedApproach
395
to the wikipage impacting a discrete number of
letters. The number of letters changed per edit was
based on the letter change magnitude distribution
derived from the Wiki-Hist-Heist output data from
wiki page histories. The editor’s actions directed
particular letter agents to become obsolete and
eventually be deleted from the page. New letters were
added or changed as agents were recycled or added.
This reflected the real world activities of a person
entering the wiki and adding, deleting or changing a
letter, word, or sentence.
It is important to note that the structure of this
model’s agent hierarchy was influenced by several
goals (Helbing and Balietti, 2011). First, the real
world system of how wikis are developed and edited
had to be distilled into its components. Second, the
model was developed to make it flexible in terms of
size. Third, the model was devised in a way that
would provide a meaningful visual component. And
finally, the model was developed to allow the addition
of more details and constraints as the system matured
and became more sophisticated.
3.3 AnyLogic
AnyLogic simulation software was used to create the
model for this project. AnyLogic contains elements
that provide support for mixed modelling including
discrete event, system dynamics, and agent-based
modeling. AnyLogic facilitates prototyping models,
detailing system design, and constructing user
interfaces. It is powerful and flexible, and offers pre-
built model constructs as well as a Java environment
for custom coding. It approaches software and model
development from an object-oriented perspective and
includes facilities for implementing models based on
UML conventions such as statecharts, inheritance,
and transition diagrams (Borshchev, 2013).
AnyLogic has been used in a variety of ABMs and
has achieved industry-wide acceptance as a robust,
flexible tool. The professional version of AnyLogic
version 7.1 was used for this project.
3.4 Wiki Letters
As stated previously, the model was created using an
ABM discrete space approach (Mustafee and
Bischoff, 2013, p. 479). Agents were graphically
displayed in a variable-sized grid with each cell
representing a particular letter. Figure 5 provides the
initial grid appearance. Letters in each cell were
derived from a function based on expected
distributions in English language writing. The
distributions used are publically available on the data-
compression website (Data-compression, 2015).
Figure 6 shows a portion of the custom Java code used
in the function as represented in AnyLogic. Figure 7
provides the custom distribution used to drive the
GetLetter function for selecting a letter for display.
Figure 5: Initial discrete space grid for the Wiki Page ABM.
Figure 6: GetLetter function used to distribute letters with
frequencies expected in typical English.
Although the letters in discrete space were not
arranged according to natural words, using a
representation of letters found in typical writing, gave
the visual display a sense of realism and provided a
more interesting user interface.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
396
Figure 7: Custom AnyLogic distribution for typical English
letter occurrence in writing.
3.5 Agent Interaction
Interaction between the wikiletter agents and the
editor agent was accomplished using a combination
of state charts and messages. The wikiletter agents
resided in a variety of states including Fresh (just
entered by an editor), ReFresh (newly changed),
Approved (stable and part of the wikipage), Obsolete
(marked for change or deletion by the editor), and
Gone (temporarily blank). The state diagram
indicated the various states as shown in Figure 8. The
letters transitioned between the states based on
messages sent by the editor agent. Figure 9 shows the
editor agent state diagram with two states: working or
resting. The transition times between working and
resting were derived from data gathered with the pre-
processor. The number of letters to be changed also
came from that source.
The editor agent sent messages which told letters
to move to the next state in their transition diagram.
This essentially drove the model and the visual
display.
Figure 8: Wikiletter agent state diagram.
Figure 9: Editor agent states.
3.6 Model Execution and Visualization
Currently the model is in its preliminary stages of
completion. A basic set of functionality has been
developed and is in place but the final user dashboard
has not been constructed. This means that letter
change frequencies and magnitudes must be manually
entered. Currently, the model provides a visual
display that indicates the number of letters being
changed and the current state of the wiki page. Figure
10 provides a view of small page with 400 characters
as an illustrative example. The final user interface
will permit easy changes to: user-specified input
distributions, wiki page sizes, editing frequencies, run
lengths, and output characteristics. These changes
will comprise Phase 2 of this project.
ConceptualWikiPageSimulation-ADiscreteSpaceAgent-basedApproach
397
Figure 10: Model during execution. Red letters are
currently available for replacement, yellow letters are
obsolete but not yet deleted and pink letters are newly
added.
4 DISCUSSION
The purpose of the Wiki Page ABM was to provide a
tool for wiki keepers and organizational information
specialists whereby their time commitments could be
better understood and managed. As described in early
sections of this article, wikis are widely used and have
become a useful technology that provide both
advantages and challenges. Among the greatest
advantages is that wikis can become long-term
information repositories developed and maintained
through collaborative efforts. Challenges, on the
other hand, require developing approaches and
policies to ensure quality, consistency, and responses
to intentional and unintentional changes that may not
be aligned with wiki goals. The current project
mitigated these challenges through providing a better
way to visualize and understand edits to the text of
wiki pages. The model also provided a better way to
anticipate wiki changes and determine human
resource requirements needed to ensure wiki quality.
Our preliminary development led us to believe
that an agent-based approach was useful in this
domain. The constructed, preliminary ABM provided
an interesting and useful way to longitudinally
examine changes made to wikis based on the reality
of its history. The Wiki-Hist-Heist pre-processor
mined data regarding events that occurred on a
specific wiki page and facilitated creation of letter
change frequency and magnitude distributions. These
distributions provided a natural way to drive the
model.
5 LIMITATIONS AND FUTURE
Although the Wiki Page ABM is in its preliminary
stages of development (See Figure 11), we believe it
offers much promise. The pre-processor written in C#
is flexible and permits data collection to be
customized according to a variety of specifications.
We anticipate using it in two ways. One is to permit
analysis of a specific wiki page. This means that
distributions unique to a page of interest can be
created and used to determine human oversight
requirements over time. A second use is to analyse a
series of wiki pages to arrive at a more universal set
of change frequency and magnitude distributions.
This information would provide general wiki staffing
requirement information and permit long term studies
to understand the impact of various strategies and
policies on managing a wiki site.
Already, use of the model and pre-processor has
provided insight into patterns of changes to wikis. We
have experimented with fine-tuning our results with
additional filtering features and by looking more
deeply at history page data. For instance, it might be
possible to predict future change activity based on
past changes to talk pages or page views.
Another planned change to the Wiki Page ABM
includes the addition of page controller, vandal, wiki
keeper, and legitimate user agents. Currently the
model utilizes letter agents and an editor agent which
help simulate words within the wiki page. To make
the visual interface more realistic and interesting, we
plan implement changes to contiguous groups of
letter agents. This will bring a greater sense of realism
to the model viewers through improved visualization.
We also plan to make the view screen resemble a wiki
page rather than a grid. Currently, the model operates
with random letters being changed by the editor agent
but they are spread throughout the grid. The desired
pattern of notification would be to represent letters as
a contiguous string that may start on one row and
continue on the next. AnyLogic offers built in
functionality to make agent communication easy but
these functions do not provide a default method for a
partially contiguous group of agent to ‘talk’. A
solution has been devised and involves adding a page
controller agent which manages the letter agents
according to ID parameters. This change is mostly
cosmetic so it has been relegated to Phase 2 of the
project. Other planned enhancements to the model in
Phase 2 include adding a user interface with menu
items that make it easily specify model input values
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
398
Figure 11: Overall flow of model in Phase 1.
and distributions, as well as provide custom run times
and replication counts. An output report will be
formatted and created to make tabulating results
easier and more accessible. Extensive validation
activities are also planned. Comparisons to prototype
discrete event simulation (DES) and system dynamics
(SD) models will also be provided (Chan et al., 2010;
Tako and Robinson, 2009).
6 CONCLUSIONS
This paper has provided information related to the
preliminary development of Wiki Page ABM. We
used a discrete space approach to structure the model
comprised of agents representing letters in the words
on a wiki page. The agents moved through states
representing whether the letters were changed,
deleted, or added based on rates derived from wiki
page histories. We discussed our custom developed
C# pre-processor, called Wiki-Hist-Heist, which
pulls information from wiki history pages to facilitate
derivation of change frequency and magnitude
distributions. These distributions provide model
inputs based on a compilation of past events. The
initial version of the Wiki Page ABM was built using
AnyLogic 7.1 Professional. It provided a framework
with user-friendly features. Overall, the initial stages
of the project have been beneficial and we plan to
continue adding enhancements that make the model
useful to wiki keepers and information managers that
need to staff and understand the behaviour of their
wiki sites better. Limitations of the current
preliminary implementation and ideas for future
study were also described.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the help of
Matthew McHaney with his insight into Java
programming and his idea to create a wiki page
controller agent to be included in Phase 2 of this
project.
REFERENCES
Borshchev, A., 2013. The Big Book of Simulation
Modeling: Multimethod Modeling with AnyLogic 6.
AnyLogic North America.
Bruch, E., Atwell, J., 2013. "Agent-based models in
empirical social research." Sociological Methods &
Research. 1-36.
Chan, W., Son, Y., Macal, C., 2010. Agent-based
simulation tutorial-simulation of emergent behavior
and differences between agent-based simulation and
discrete-event simulation. In Proceedings of the Winter
Simulation Conference. 135-150.
Data-Compression, 2012. “First Order Statistics for
Distribution Values for Letters.” Retrieved from
http://www.data-compression.com/english.html.
García, G. P., 2012. Improving Creation, Maintenance and
Contribution in Wikis with Domain Specific Languages.
Diss. University of the Basque Country.
ConceptualWikiPageSimulation-ADiscreteSpaceAgent-basedApproach
399
Hai-Jew, S., McHaney, R., 2010. ELATEwiki: Evolving an
E-Learning Faculty Wiki. In Cases on Digital
Technologies in Higher Education: Issues and
Challenges (Rocci Luppicini and A. K. Haghi, eds.). 1-
23.
Heath, B., Hill, R., Ciarallo, F., 2009. A survey of agent-
based modeling practices (January 1998 to July 2008).
Journal of Artificial Societies and Social Simulation,
12(4), 9.
Helbing D., Balietti S., 2011. How to do agent-based
simulations in the future: from modeling social
mechanisms to emergent phenomena and interactive
systems design, in Technical Report 11-06-024. Santa
Fe, NM. Santa Fe Institute.
Hepp, M., Siorpaes, K., Bachlechner, D., 2007. Harvesting
Wiki Consensus: Using Wikipedia Entries for
Knowledge Management. Special issue on Semantic
Knowledge Management, IEEE Internet Computing.
54–65.
Macal, C., North, M., 2010. Tutorial on agent-based
modelling and simulation. Journal of simulation, 4(3),
151-162.
Mader, S., 2008. Wikipatterns: A practical guide to
improving productivity and collaboration in your
organization, Wiley Pub., Indianapolis, IN.
Meatballwiki, 2011. Softsecurity, Retrieved from
http://meatballwiki.org/wiki/SoftSecurity.
McHaney, R., 2012. "The Web 2.0 Mandate for a Transition
from Webmaster to Wiki Master." Open-Source
Technologies for Maximizing the Creation,
Deployment, and Use of Digital Resources and
Information. IGI Global. 193-218.
McHaney, R, Spire, L., Boggs, R., (2014). "E-
LearningFacultyModules.org." Packaging Digital
Information for Enhanced Learning and Analysis: Data
Visualization, Spatialization, and Multidimensionality:
Data Visualization, Spatialization, and
Multidimensionality. IGI Global. 103-119.
Mustafee, N., Bischoff, E., 2013. Analysing trade-offs in
container loading: combining load plan construction
heuristics with agent based simulation. International
Transactions in Operational Research, 20(4), 471-491.
Sutton, A. (2006, September). Stop Using Wikis as
Documentation. Symphonious. Retrieved from
http://www.symphonious.net/2006/09/02/stop-using-
wikis-as-documentation/
Tako, A., Robinson, S., 2009. Comparing discrete-event
simulation and system dynamics: users’ perceptions.
Journal of the Operational Research Society. 60. 296-
313.
Taylor, S. (Ed.), 2014. Agent-based Modeling and
Simulation. Palgrave Macmillan.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
400
