Open Source Research, A Quiet Revolution

OWEN DENSMORE
Complexity Workshop

Abstract
The Open Source revolution is not limited to operating systems, file sharing and web
servers. It is flourishing in other areas as well. This paper introduces the reader to a
quiet revolution occurring in the research community, where open source research
includes not only algorithms and shared techniques. It also includes hardware design
and open community collaboration as well. One research community in particular, the
complex adaptive systems researchers, is introduced here in detail.

Introduction
Imagine a world wide project to land on an asteroid. Imagine this done with no barriers and total
participation by interested parties. All information about the “rocket science” is freely available with no
proprietary or secret technologies. The details of the pilot's console are online, and have even been turned
into this year's hottest game. Every fastener used within the system is standard, many available at high
end hobbyist supply stores or at Silicon Valley's famous Fry’s Electronics. The “buckyball,” nanotube
straps used for securing pairs of asteroids in mining operations were created by Cal Tech's
undergraduate advanced studies group and are available in kit form. The entire project has been made
the topic of MIT's1 graduate engineering school, and all course materials are freely available on their
website.

As odd and unapproachable as this may appear to us today, it is slowly becoming fact. Researchers world
wide are lowering the barriers to collaboration and communication. Not only is software being “open
sourced” but so is hardware and scientific research itself. Many researchers are going beyond passively
being open; they are actively seeking participation and creating communities to refine the research.

This discussion will not attempt an exhaustive coverage of open source research as the topic is simply too
huge. Instead I'll give several in-depth examples from research, primarily from the field of complexity
science. For now, we'll define complexity science as the exploration of systems consisting of individual
interacting components acting with only local knowledge in a spatial (2D or 3D) environment over time.
The purpose of complexity science is to understand the emergent properties of such a system as it
evolves. Examples of these systems include economies, societies, biological systems, the Web and
Internet, peer network systems, and even autonomous robots. They typically do not yield to closed form
mathematical solutions.

Evolution of Cooperation
One of the early examples of open source research came from the social sciences. In the early 1980s,
Robert Axelrod was puzzled by a seeming anomaly in a non-zero sum game seen in a game theory
problem demonstrated in the Prisoners Dilemma. This game's name derives from an imagined scenario in
which two criminals are arrested without sufficient evidence against them. The two players can choose
between two moves—“cooperate” or “defect”—in an attempt to minimize the length of their prison
sentences. Each player gains when both cooperate. However, if only one of them cooperates, the player
that does not cooperate will serve a shorter sentence. If both players choose not to cooperate and thus
“rat” on each other, they will both serve the longest sentence. This outcome is what game theory grimly
predicts will happen.

Because the dilemma stems from not knowing the other player's plan, Axelrod decided to resolve this by
“iterating” the game: having it played many times, thus introducing memory and learning. He held two
famous tournaments where participants (both experts and hobbyists) were invited to submit simple
programs playing against each other. The results were surprising. The simplest program won in both
tournaments. It was called TIT-FOR-TAT, which called for initial cooperation and then always repeated
the opponent's prior move: defecting if the opponent had defected, cooperating if the opponent had
cooperated.

This is an extraordinary result! For example, it can be proven that TIT-FOR-TAT never beats its opponent,
but only ties at best. Yet because it promotes high scores for itself and its opponent, it comes out on top
during prolonged play with many different strategies. This discovery became one of the cornerstones of
current Computational Social Science, the study of the Evolution of Cooperation.

From the beginning, Axelrod and his collaborators decided to not only make the tournament public, but
also to promote open source research by publishing the algorithms and extensive details of the actual
tournament play. This was done initially in book form,2 and later on FTP and websites.3 From this effort
arose dozens of sites and papers using these freely available details to further refine the initial work.

Papers
Scientific papers, generally published in professional or academic journals, are the keystone for research.
In the 1980s, many papers started to be made available in electronic form, generally formatted as text
(ASCII), PostScript, or later PDF (Portable Document Format) and Word (doc). Although some of these
formats were initially proprietary, all have since become standard and available from multiple vendors.
This may be the single most important component of open source research—the easy and free availability
of research papers in standard formats.

Initially these papers were exchanged among colleagues via email and FTP. Later, as the Web emerged,
individuals and then institutions began creating sites, collecting their papers in standard formats.
Generally these are made freely available to the public, although a few do still persist in requiring either
subscriptions or institutional affiliation. However, even these are beginning to become available via
public library interlibrary loan, or at worst, making individual papers available for a fee.

Perhaps the most impressive collection today is the CiteSeer Scientific Literature Digital Library index.4
This is a very large collection of scientific literature with complete citation linkage among the papers. This
linkage is managed by Autonomous Citation Indexing (ACI), which also automatically generates the
paper in several formats: PostScript and PDF being the most used. Well over 100,000 documents are
included in the index. Steve Lawrence also provides convincing evidence of the increasing access to well-
indexed, freely available research papers in his book Online or Invisible.5

A similarly ambitious index is the arXiv archive,6 which evolved from the former xxx.lanl.gov site. It is a
fully automated electronic archive for research papers within physics, mathematics, nonlinear sciences,
computational linguistics, and neuroscience.

The Directory of Open Access Journals 7 is more directly focused on freely available journals. There are
now 1,149 journals in the Directory, with 314 journals searchable at the article level. Currently 55,343
articles are included in the service. This is a broader index than CiteSeer, covering many nonscientific
disciplines.

An interesting related phenomenon is the emergence of online journals. An example in the complexity
science field is the Journal of Artificial Societies and Social Simulation.8 This is a quarterly, Web-only, peer-
reviewed journal in its seventh year focused on articles discussing Agent Based Modeling.

Sadly, some of the most important journals still make their articles available only through subscription.
The most egregious example is the Association for Computing Machinery (ACM), 9 one of the pioneers in
computer science, which makes its articles available to non-subscribers for $10.00 USD. The worst
example of this is ACM’s Turing Award series,10 in which the recipient of the award delivers a lecture
that is then published in the journal only available to subscribers. These are cornerstone computer science
discussions that are simply not available to most interested readers.

Some journals have a more reasonable subscription access limitation: the articles are made generally
available after a brief period. For example, the Proceedings of the National Academy of Sciences11 makes
its articles freely available six months after publication.

The difficulty with expensive and restricted access to research journals is being widely discussed amongst
respected practitioners. John Baez,12 the researcher who has written or taught about such subjects as loop
quantum gravity, quantum geometry and spin foams, is not to be taken lightly! His recent detailed
discussion entitled “What We Can Do About Science Journals”13 is a great read from the viewpoint of the
mathematics researcher.

Docking
One of the more interesting outcomes of freely available research papers is “docking” experimental
results. Docking means performing similar experiments in different environments, thereby verifying and
expanding the results of the initial work.

This has lead to an openness not usually associated with academic environments. A fascinating example
is from a pair of researchers at Brookings Institute, Robert Axtell14 and Joshua Epstein,15 authors of a
seminal study on artificial societies in the book Growing Artificial Societies: Social Science from the Bottom
Up16. In this book, they presented SugarScape, an artificial world in which inhabitants are easily
programmed to explore a wide variety of behaviors. Just as Axelrod's books and papers published the
Prisoner's Dilemma algorithms in detail, Axtell and Epstein openly provided the details for their
SugarScape world.17 Indeed, a companion CD-ROM to the book included the entire simulation
environment so that readers could repeat the experiments as they read the book.

Not unexpectedly, Axtell, Epstein, and Axelrod joined forces to write a paper on the docking process,
“Aligning Simulation Models: A Case Study and Results.”18 This study not only discussed the difficulties
and value of docking experiments, it also quantified the costs and made recommendations on creating a
general practice of doing so. The paper includes complete work logs detailing to the minute the effort
involved in the study. It was discussed more generally later in book form in an article entitled
“Advancing the Art of Simulation in the Social Sciences.”19

Other docking efforts have taken place, such the works of North and Macal;20 and Xu, Gao, and Madey.21
Perhaps the most detailed is Replication of Eight “Social Science” Simulation Models22 found at the
Center for the Study of Complex Systems website. These efforts have lead to the first “Model to Model”
workshop23 in which modelers met to discuss collaborating to strengthen the validity of experimental
results.

Books
A surprising number of books are beginning to be freely available in electronic form. All of Stephen
Wolfram's A New Kind of Science24 has recently become available online,25 although requiring free
registration. The book is already an amazingly encyclopedic work, but the online version actually
augments each page with additional material. It also enhances the book by providing a full search facility.
Finally, by making the thousands of images within the book available, researchers can exchange notes on
specific parts of the books along with supporting images.

A very popular undergraduate text, Introduction to Probability by Grinstead and Snell, has similarly
become available online.26 They also provide considerable additional material and a complete set of
programs are made available in four formats: Mathematica, Maple, Basic, and Java Applets. In addition,
the answers to the odd-numbered problems are available, and instructors can get the even ones with an
email to the authors. It’s striking to read their discussion about how the GNU version of the book will
improve over the existing paper version, and about their gratitude to the American Mathematical Society,
publisher of the paper version.

Perhaps the most respected introduction to computer science is Abelson and Sussman's Structure and
Interpretation of Computer Programs.27 It also is freely available in electronic form. In this case, the electronic
version evolves the book into an entire course. For self study, the book's website28 includes sample
programming assignments, all the source code used in the book, and assistance in obtaining Scheme, the
language used by the book.

It appears that making books available in electronic form is likely to increase rather than decrease the
sales of the paper version. Personally, I always obtain both forms, using the digital version to expand on
and enrich the paper version.

As an example of using the electronic form to enhance the paper book, consider another amazing
instance, David MacKay's Information Theory, Inference and Learning Algorithms,29 which is becoming a
classic in the complexity community. A person in our reading group presented the “Efficient Monte Carlo
Methods” chapter, which stimulated discussions on other sections and motiviated many of the group’s
members to order the book after witnessing its effectiveness. In addition, I've begun translating the book's
examples into the J programming system. The electronic version of the book naturally includes
enhancements to the paper version. For instance, many of the examples are available as Octave source
code, downloadable from the website. Also, additional future chapters are first made available
electronically.

A similar, book-like offering occurs with subject reviews made by respected researchers. These are
typically candidates for sections in books, but are published initially in reviewed journals. Two examples
within complexity science are Mark Newman's overview of graph theory in “The Structure and Function
of Complex Networks,”30 and Cosma Shalizi's overview of complexity science in “Methods and
Techniques of Complex Systems Science.”31 These large (58 pages and 48 pages respectively) reviews are
invaluable to new researchers in the field.

The idea of freely available books may seem unimportant to the casual observer. But there are many ways
their availability accelerates research and enhances collaboration. For example, my ability to send a
couple of pages of MacKay's book to a European collaborator saved a month of copying and mailing
material. And the ability to read the book while the paper version is in transit speeds up projects
considerably. In the MacKay example, I was able to cut and paste an image used to study noisy
communication channels, writing my own versions of the redundancy algorithms using the actual
example data in the book! Obviously, the time savings are huge.

Courseware
With papers and books being freely available on the Net, a natural next step is for courses to use them.
One of the earliest examples is the Open University32 in England, started in 1969. This amazing amalgam
of home study, multimedia learning (tapes, web, radio and TV broadcasts), local tutors, mailed course
packets, and even telephone conferences is Britain's largest university, with 200,000 current students and
over two million participants in its lifetime. At this university, a graduate typically earns a degree in six
years while working, at a total cost of roughly £4,000. In addition, the university’s structure and openness
makes higher education much more available to those with disabilities (nearly 8,000 such students). And,
despite its unique approach to higher learning, the Open University even has support for research and
funding for research projects.
Perhaps the best publicized example in the United States is MIT's OpenCourseWare.33 While not a
degree-granting facility, it is an ambitious project to make all of MIT's course materials available freely
online. For example, Abelson and Sussman's introduction to computer science referred to previously has
a section within the OpenCourseWare site.34

Tools
A powerful tool used by the complexity science community is Agent Based Modeling (ABM). As
mentioned above, complexity science studies the (generally nonlinear) interactions of many objects
(agents). They interact both spatially (neighbors and with motion) and repetitively, iterated over time.
One of the first Agent Based Modeling systems built was Swarm,35 developed by Chris Langton and
colleagues at the Santa Fe Institute. Several are now in existence, many written in Java so as to be portable
across different types of computers. Two prominent examples today are RePast36 and NetLogo,37 both
freely available, Java-based, platform independent systems. They are easy to install and have very active
developer communities sharing ideas via mail lists. Naturally there are numerous tools in addition to
ABMthat are freely available and invaluable to research. 38

Hardware
People often assume that freely available implementations of research projects are limited to software.
Amazingly, this is not the case. Many digital electronic projects are following the open source trend. And,
because the digital formats are standard for creating digital boards, one can often receive the digital
layout documents, send them to an inexpensive web fabrication facility, and obtain the hardware within
a few days! One of the more well known examples is from Berkeley's TinyOS and related hardware
project TOS.39 Downloading a package brings detailed information in the readme file, including a pointer
to a file which literally can be sent to a web-fab house that will return the board within a week. A similar
project, RoboMote,40 is being built and deployed by USC's robotics lab. 41 Their website includes board
layout schematics and bill of materials.

Summary and Futures

This paper has presented a very narrow review of open source research activities, limited to recent
experiences in complexity science. We have shown that a growing number of research papers, broad
topical reviews, book chapters, and even entire books are becoming freely available in standard digital
formats on the Web. We also explored active collaboration efforts such as the docking of experiments and
making considerable supplementary material openly available, including the initial papers and
descriptions of experiments.. Some of these efforts become community efforts involving a group of
experts honing initial experimental findings such as the iterated Prisoner's Dilemma and the SugarScape
artificial societies. Beyond this, entire courses and degree programs are being made openly available, and
even plans for hardware devices are being exchanged with fabrication-ready machining files.

This article could have mentioned many more examples and certainly practitioners in every field are
striving to make their systems more and more open and accessible over time. Our mythic asteroid
landing is not far away!

There are still many barriers, however. Business plans for journal publishing should evolve to use
resources differently, much as software manufacturers are doing. For example, they may give away their
resources but sell their document research expertise, bundling groups of papers into sector reports and
evaluations. An intermediate strategy would be for such businesses to join together in a brokering
arrangement, providing access to broad sets of subscriptions at a much reduced price. The Creative
Commons organization42 is succeeding in its effort to find the middle way. As it states on its website:
“Thus, a single goal unites Creative Commons' current and future projects: to build a layer of
 reasonable, flexible copyright in the face of increasingly restrictive default rules.”

Significantly, one of their projects is the Creative Science Commons.43 This will help unify the various
individual efforts mentioned here, and give them the legal structure to manage the tension between
control and free access.

Innovative approaches and new business models such as the ones described in this chapter are needed to
facilitate open research so that we can build upon, rather than recreate information. Faced with declining
subscriptions for its traditional publication models, publishers and other intellectual property holders
would benefit from embracing the principles of standardization in which competing companies cooperate
and share information to create better products and larger markets. This type of business model
migration is normal for the expansion phase of rapidly developing technologies such as computing and
communications and is needed for the publishing industry to flourish under this new economy. It is a
migration from Intellectual Property to Intellectual Capitol, from secrets to know-how.

Notes

1
MIT: Massachusetts Institute of Technology.
2
Axelrod, Robert. The Evolution of Cooperation (New York: Basic Books), 1984.
3
Examples include: Complexity of Cooperation.
http://www.cscs.umich.edu/Software/ComplexCoop.html and Robert Axelrod’s research papers.
http://www-personal.umich.edu/~axe/research_papers.html. Note “supporting material” sections
which include additional material used within the paper.
4
CiteSeer Scientific Literature Digital Library index. http://citeseer.ist.psu.edu/.
5
Lawrence, Steve. “Online or Invisible,” Nature 411, no. 6837 (2001): 521.
6
Cornell University arXiv e-Print archive. http://arxiv.org/.
7
Directory of Open Access Journals. http://www.doaj.org/.
8
Journal of Articial Societies and Social Stimulation. http://jasss.soc.surrey.ac.uk/JASSS.html.
9
Association for Computing Machinery. http://www.acm.org/.
10
Association for Computing Machinery A. M. Turing Awards of the Association for Computing
Machinery. http://www.acm.org/awards/waward.html.
11
National Academy of Sciences. Proceedings. http://pnas.org/.
12
John Baez’s official Web site. http://www.math.ucr.edu/home/baez/.
13
“What We Can Do About Science Journals.” http://www.math.ucr.edu/home/baez/journals.html.
14
Robert Axtell’s official Web site. http://www.brook.edu/scholars/raxtell.htm.
15
Joshua Epstein’s official Web site. http://www.brook.edu/scholars/jepstein.htm.
16
Epstein, Joshua and Robert Axtell. Growing Artificial Societies: Social Science From the Bottom Up
(Washington D.C.: Brookings Institution Press, 1996).
17
Growing Artificial Societies. http://www.brook.edu/press/books/ARTIFSOC.htm
18
Axtell, Robert, et al., “Aligning Simulation Models: A Case Study and Results,” Computational and
Mathematical Organization Theory 1 (1996): 123-141.
19
Axelrod, Robert. “Advancing the Art of Simulation in the Social Sciences,” in Conte, Rosario, Rainer
Hegselmann and Pietro Terna (eds.). Stimulating Social Phenomena (Berlin: Springer, 1997).
20
North, Michael and Charles M. Macal. “The Beer Dock: Aligning Three and a Half Implementations of the Beer
Game,” SwarmFest (Seattle, WA: Swarm Development Group, 2202).
21
Xu, Jim, Yongqin Gao and Gregory Madey. “A Docking Experiment: Swarm and Repast for Social
Network Modeling,” (paper presented at the Seventh Annual Swarm Researchers, Notre Dame, IN., April
2003) Swarm 2003.
22
Replication of eight “Social Science” simulation models. http://www.pscs.umich.edu/Software/CAR-
replications.html.
23
Model to Model Workshop. http://cfpm.org/m2m.
24
Wolfram, Stephen. A New Kind of Science (Champaign, IL: Wolfram Media, 2002).
25
A New King of Science. http://www.wolframscience.com/nksonline/toc.html.
26
Grimstead, Charles and J. Laurie Snell. Introduction to Probability, 2nd. rev. ed. (Providence, RI: American
Mathematical Society, 1997); http://www.dartmouth.edu/~chance/teaching_aids/articles.html.
27
Abelson, Harold, Gerald Jay Sussman,.and Julie Sussman. Structure and Interpretation of Computer
Programs, 2nd ed. (Cambridge, Mass.: MIT Press, 1996).
28
Structure and Interpretation of Computer Programs. http://mitpress.mit.edu/sicp/.
29
McKay, David J. C. Information Theory, Inference and Learning Algorithms (Cambridge, U.K.: Cambridge
University Press, 2003).
30
Newman, Mark E. J. “The Structure and Function of Complex Networks,” SIAM Review 45 (2003): 167-
256.
31
Shalizi, Cosma Rohilla. “Methods and Techniques of Complex Systems Science: An Overview” in
Deisbroeck, Thomas, J. Yasha Kresh and Thomas B. Kepler (eds.). Complex Systems Science In Biomedicine
(New York: Kluwer, forthcoming).
32
Open University. http://www.open.ac.uk/.
33
MIT OpenCourseWare. http://ocw.mit.edu/index.html.
34
Abelson, Sussman and Sussman. Structure and Interpretation of Computer Structures course site.
http://ocw.mit.edu/OcwWeb/Electrical-Engineering-and-Computer-Science/6-001Structure-and-
Interpretation-of-Computer-ProgramsFall2002/CourseHome/index.htm.
35
Swarm. http://wiki.swarm.org/wiki/Main_Page.
36
University of Chicago's Social Science Research RePast. http://repast.sourceforge.net/.
37
NetLogo. http://ccl.northwestern.edu/netlogo.
38
A list of the freely available tools used by myself and colleagues in the last 3 months include:
• GnuPlot: the standard scientific plotting system, available on all platforms,
http://www.gnuplot.info/.
• Octave: enhances GnuPlot by adding a sophisticated Matrix algebra system using a language
compatible with the commercial MatLab system, http://www.octave.org.
• R: is a statistical computing and graphics language largely compatible with S, its commercial
counterpart, http://www.r-project.org.
• J: a general and complete mathematical environment descending from APL,
http://www.jsoftware.com/index.htm.
• GraphViz: a graph (nodes and edges) drawing system,
http://www.research.att.com/sw/tools/graphviz.
• VisAD: a Java Component Library for interactive analysis and visualization of numerical data,
http://www.ssec.wisc.edu/~billh/visad.html.
• GeoTools: , Geo Tools, http://www.geotools.org/. OpenMap, http://openmap.bbn.com/ and
Jump: Java-based Graphical Information Systems (GIS), http://www.vividsolutions.com/jump/.
• ImageJ: a Java image processing and analysis application and library, http://rsb.info.nih.gov/ij/.
39
TinyOS and Hardware. http://www.tinyos.net/scoop/special/hardware.
40
Dantu, Karthik, et al., Robomote: Enabling Mobility in Sensor Networks Technical Report CRES-04-006 (Los
Angeles: Center for Robotics and Embedded Systems, Viterbi School of Engineering, University of
Southern California, 2004).
41
RoboMote. http://www-robotics.usc.edu/projects/robomote/.
42
Creative Commons. http://creativecommons.org/.
43
Creative Science Commons Site. http://creativecommons.org/projects/science.