When I began this series of investigative reports on Second Life, one thing I was really looking forward to was to see just how much “science” was going on in Second Life.  I wanted to know is the majority of what happens inworld Social or Academic?

The reality wound up being more complex than I originally thought, being heavily influenced by the perspective of who I was speaking to.  Some people said science was everywhere, while some people said it was a nonexistant community.  After several weeks of digging around I’ve come to some conclusions, and I share them here.

(10:30am Added an addendum to the end of this story)

(2/17/2010 Another addendum)

The Definition of ‘Science’

The primary issue is one of definitions: What is ‘Science’ , and what would constitute ‘Science in Second Life’?  While I was leaning more toward high-performance computing and data analysis, some people thought of sociological studies, while others simply of science education.  All are valid in their own way, but very different and distinct systems.

Sociology & Psychology

Several psychologists and sociologists view SecondLife as a rare sandbox of human behavior.  The open nature of the system removes several of the restrictions found in online games, allowing more natural interaction between avatars, but also allows people to assume whole new personalities.  The distinctions between the user’s real persona and their chosen avatar has been the subject of many psychological studies and books.

Books like John Suler’s ‘Psychology of Cyberspace’ is a great resource that discusses the observed group dynamics and psychological behaviors monitored inside virtual worlds, and lays out possibilities for psychotherapy and clinical work inworld.  However, such world raises several ethical dilemmas:

First of all, is it ethical to attempt psychotherapy in cyberspace? If the therapist is communicating with the client through typed text (as in e-mail, chat, and message boards), all sorts of valuable information – like physical appearance, body language, and tone of voice – are missing. That easily could pose problems in making an accurate diagnosis and evaluating the treatment, which often rely on f2f behavioral cues. Without f2f cues, the therapist also may not be able to verify the identity of the client. Is the person really who he or she claims to be?

Scientific Institutes

Several scientific institutes operate inside virtual worlds, Second Life in particular, but rarely perform ‘science’ inworld.

The IEEE maintains a large region inworld, full of buildings and beautiful scenery.  However, every time I visited over the month I’ve been doing research, it was deserted.  From what I could tell, the island existed to serve only a few minor purposes:

  • Meetings were held a regular, although very wide separated, intervals to discuss new IEEE standards and regulations.
  • A library held information on all approved and pending IEEE standards, in classical library form.  (Unfortunately, the “books” were actually merely links to the PDF’s on the IEEE website, so little new information was available).
  • A fairly extensive “training” area for newbies, teaching the basics of movement, interaction, and basic UI

This same design plays out over and over throughout Second Life.  A few exceptions exist, however, like the NOAA island.  NOAA has an in-world island where the same meeting and training areas exist, however they also have some basic ‘tourist trap’ attractions such as a “Fly Through a Hurricane” simulation and life updated weather map for the US.  Nothing terribly useful, but fun and worth checking out at least once if you’re inworld.

Scientific Research

Finally, the real Scientific Research I began my quest to find.  Unfortunately, during my month inworld I’ve been unable to find any HPC simulation research in Second Life.  This seems to be for several reasons:

  • SecondLife doesn’t lend itself to the same geometric design constraints as simulation outputs.  Everything in SL is built from ‘Prims’, while most Visualization and Analysis packages work with raw Triangles.
  • SecondLife also does not lend itself to extremely large models, most models are limited in the few thousand “prims” and not the millions of triangles typically required.
  • SecondLife costs money
  • Putting a simulation result in SecondLife requires that the owner effectively “give up” the dataset, putting it in the world for whatever nefarious means it may come to.

Some of these can be mitigated.  Linden Labs has recently announced the “Second Life Enterprise” product that allows researchers to effectively deploy their own SL grid within their own labs, removing the public nature of the system.  This also allows researchers to change from the monthly “rental” pay system of Second Life to a more permanent “buy it once” exchange, which is more palatable to most institutions.

However, the first two requirements still are a burden. One good example of this is the “E8 Polytope” model made by Second Life user “Wizard Gynoid“. More art than science, it is a sculpture showing a mathematical construct in 8-dimensional space.  Beautiful to look at, it cannot exist in most places within Second Life because of the extreme geometric complexity.  How complex?  6,000 prims.  Paltry by data analysis standards, but too much for Second Life to handle.

However, all is not lost.  Other groups want to use SecondLife as a tool for collaborative data analysis and exploration, and alternative solutions have arisen.

The most widely known solution is called ‘ScienceSim‘, operated through a collaboration of the ACM and various research institutions and corporations.  Based on the open-source SecondLife server system called ‘OpenSim‘, it is mostly compatible with the SecondLife client but adds in more functionality for those who need it.  Allowing users to connect in their own physics engines and other computational drivers, users can either connect a grid up to a running simulation or perhaps even run their simulation directly inworld.

During the recent SuperComputing2009 conference in Portland, OR, Justin Rattner put it on the line during his keynote by demonstrating one such simulation live.  The Fern Seed project from Utah State University, developed by Aaron Duffy, runs multiple islands inside ScienceSim.

Rattner used the simulation as an example of emerging research technologies that will create future challenges and opportunities for the High Powered Computing industry. The computationally intensive project, built on the open source Opensimulator 3D Application Server, simulates a population of around 100,000 ferns in an environment with customizable terrain, soil, weather, seasons and physics. Each fern possesses a genetic code that controls how well it adapts to different environmental conditions. Based on the environment, population density, and other factors, the ferns reproduce, passing successful genetic codes to their offspring and allowing the population to evolve over time. Users can login and interact with the objects in the virtual world and with each other using avatars, providing unique opportunities for data visualization and education.

A powerful demonstration and a great use of the technology, when I visited I found it empty.  The simulations were no longer running, although all of the islands (and a degree of historical data) remained.  Hopefully I merely found them during a lull in simulation, and their research will resume soon.

Conclusion

After a month of intensely searching for scientific research in SecondLife, and by extension any virtual world, I did finally find some, although not what I had hoped.  HPC simulation and data analysis seems to be a rarity in the space for reasons ranging from the constraints of the massive datasets to “Why bother?”, but the sociological and psychological uses of the worlds are still under active research.

Addendum: (Added Late)

I meant to add this before publishing, but forgot.

Of course, everything I said in this article could possibly be rendered moot by tools like OpenSim & Second Life Enterprise, as these “Grids” could run in entirely enclosed networks, unavailable for me to access.  So perhaps data analysis and other tools are widely popular and used everyday, but not in a public-facing network where we can see them.  There is a little information available indicating this could be the case, tools like Green Phosphor, but until we start seeing more press releases and public information, we won’t really know.

And for the record, I spoke with several researchers in a wide spectrum of disciplines and none of them were actively using Virtual Worlds for scientific research.  Most of them didn’t use Virtual Worlds for anything at all, but some did use them for meetings and coordination, but took the actual research and analysis into other platforms specifically designed for it.   People are already telling me I’m wrong (only an hour after posting, wow), so maybe I’ll revisit this after I collect some new names to talk to.

Addendum #2:

Wow, so many responses.  The SecondLife & VirtualWorlds community is nothing if not well-connected and passionate.  Lots of great links in the comments, so be sure to read those.  People are already pointing out my errors & new information, so it looks like my journal into SecondLife is not complete yet.  Be sure to read one followup already online:

Also, if anyone out there meets this criteria, please drop me a line at [email protected] :

Using SecondLife, ScienceSim, or any OpenSim based project (OSGrid, ReactionGrid, etc) to visualize and analysis data from a computational simulation run on an HPC platform that is used for actual insight and action.  Not for computational research, but to actually create real actionable data for real products.

Weather simulations, blast codes, weight strains, computational chemistry, anything is a go.

Tags