Stanford Virtual Worlds Research

This article contains selected notes on the some of the research conducted at Stanford University on virtual worlds and the interaction of humans within virtual environments. It is based on sessions held during the Media X conference. Pat Hanrahan defined a virtual world as a “networked multi-user distributed environment”. But the audience reaction was altogether less technical, and more oriented towards the social implications of such environments.

Stanford is one of the few universities that can not simply be accused of climbing on the virtual worlds band-wagon: People like Nick Yee were examining these environments at long before they were regarded as a suitable topic for serious research. Related sessions on workplace application and DKP and the archiving of virtual worlds/games will be covered by separate articles.

Why Use Virtual Environments for Medical Training?

LeRoy Heinrichs spoke on the use of virtual medical rooms for training medical students.

It is cost effective, even when developing bespoke software: Conducting a live training exercise in a physical hospital costs about $50,000 per day, and can only train a relatively small group. Stanford’s first virtual patient model cost almost $1 million to develop, yet in the long run is still cheaper than physical-world exercises.

Initial analysis of performance is not yet conclusive, however early signs suggest knowledge does transfer to real practice, and virtual training is just as good as other methods.

The business case for virtual worlds is ultimately a critical driver to their success outside of their traditional (game or social) environments. Medicine is a fundamentally expensive business, so even with custom software, one user can make a saving. Other sectors may be slower to follow, waiting for the cost to drop. Cost are likely to drop by sharing development costs between multiple projects – either industry-wide initiatives, or through the development of platforms for virtual worlds, which will transfer most of the costs on to a single provider, who can then share those costs between many customers.

Size Matters

Renate Fruchter revealed that visual size does matter. Ideally people should appear on screen life-size: In most cases that means a bigger screen!

Jeremy Bailenson outlined some of Nick Yee’s research behind the “virtual mirror”. The virtual mirror is a technique that changes the visual identity of a person’s avatar while in a virtual world: Their avatars literally look into a mirror and take a different form.

The experiment is useful in understanding the consequences of an apparently fluid online identity, and determining whether self-perception theory (and similar) transfer to avatars: If you don’t know how to act, you look at yourself, particularly your uniform, and that determines your behaviour.

Height is important. In the physical world, height correlates to confidence and personal income. Through the use of an “ultimatum game”, where avatars negotiate a deal, it was possible to show that a 10cm difference in avatar height increased the value of that avatar’s deals in their favour.

Physical attractiveness of avatars was also tested by examining “interpersonal distance”: If you like someone, you will tend to stand closer to them. And they’ll disclose more information.

Finally the effects of age were tested by morphing pictures of one’s self to show the passing of years. The older the avatar, the more the subjects were prepared to invest in their retirement.

Further detail on some of these topics can be found at The Daedalus Project.

Virtual Worlds, Serious Work, and Collaboration for DKP

Byron Reeves (Stanford University) spoke to the Media X conference about how experiences from virtual worlds could be transferred into working life. This article summarises his talk, and contains personal analysis of the potential for using DKP (Dragon Kill Point) systems to measure contribution to collaborative activity.

Playing Puzzle Pirates at Work

Take a dull job such as that of a call centre worker. Now take the online game, Puzzle Pirates. Strip out the puzzling part, and add in the dull job. What do we get?

  • Metrics about the performance of yourself and others – highly detailed feedback loops that are largely missing from most regular jobs.
  • Through these metrics, a way to identify issues with team performance, giving…
  • An easy way to notice and resolve human issues within the team.
  • A way to make money that relates directly to performance within the game.

Why Might This Work?

Some possible reasons:

  1. Worlds are popular. People like playing them! Reeves was unusual among academics in acknowledging the huge popularity of teen-orientated worlds like Habbo Hotel, and down-playing relatively unpopular titles like Second Life.
  2. A new “gamer generation” is emerging. Even without the online component of games, these features aspects of competition, failure, risk and feedback. It is reasonable that this generation will come to expect to work using collaboration tools with features that match.
  3. Well understood recipe for creating a great game.
  4. Emotional involvement. Byron Reeves showed how heart rate increased by the value of 10 [presumably beats per minute] when playing with another human-controlled avatar, rather than a computer-controlled agent. This implies a performance gain when human collaboration is present.
  5. Technology: Worlds are easier to build, and “better”.
  6. Painful long-standing problems in enterprises might be solved. For example, large proportions of workers are “out of the office”; have limited employee feedback; do fundamentally dull work; and require emotional contact with other humans to innovate.

Dragon Kill Points as a Measure of Contribution

Dragon Kill Points (DKP) might be used as a way to value contributions to collaborative environments such as wikis. DKP is a way of resolving how to share finite loot among a group – originally from killing dragons in Everquest, now from any encounter that requires a group to complete.

The application of DKP to other collaborative environments was not fully developed. So let me try.

Loot is the primary reward from most collaborative activity in an game such as World of Warcraft (probably where DKP is currently most used). At the most advanced stages of the game a hostile creature might require 10 or 25 people to kill, yet only yield 2 or 3 items of loot. An equitable method of distributing loot is critical to long-term motivation of players.

Pragmatic random distribution of loot is one method: Players those avatars would benefit from the loot are invited to roll a virtual 100-sided dice, and the highest score wins the loot. The process is not entirely without social mediation. For example, one player might pass (forfeit their roll) to allow another to win loot that the first player knows they particularly need. Likewise rolling on loot that the rest of the group perceive the player doesn’t really need is likely to cause a social backlash. Pragmatic random distribution of loot is easy to administer and well suited to small groups comprising players that might not regularly play together.

However, pragmatic random distribution does not account for long-term contributions: One player might attend one session, gain a rare loot, and stop contributing to further sessions. Meanwhile another player might attend multiple sessions and gain nothing.

DKP is an alternative method. It creates a tally of points based on contribution to group activity. Loot is then distributed based on the volume of points a player has banked (and is prepared to spend) from earlier contributions. DKP is generally used where:

  • Groups are composed of many people, typically 10 or more.
  • Groups are formed out of a limited set of people that often play together.
  • A low volume of loot is generated relative to the time commitment required to generate it.
  • Groups routinely split play sessions between activities which generate different amounts of loot. For example, learning/practice (“progression”) vs gathering loot from already familiar activities (“farming”).

If DKP sounds simple, it isn’t: A DKP system is a complex construct, with different ways to measure contribution and balance the flow of loot to players. Agreeing that balance is a highly social activity, and failure to get the balance right can break-up long-established groups.

Group stress (“drama”) caused by the requirement for a complex DKP system may be one of the reasons for the growing importance of tokens in World of Warcraft. Group activity yields tokens, rather than loot. The tokens can still be traded for loot within the game. However tokenization removes some of the requirement for groups to balance the value of different items of loot.

DKP as a Currency

Edward Castronova and Joshua Fairfield have already mused on some of the economic aspects of DKP. But there are some interesting tangents that have not obviously been explored.

DKP is a meta-currency where the value of the currency is based on the values players place on one another’s contribution. Oddly this makes DKP far more like a modern physical-world currency than the formal in-game currencies created and balanced by game designers. Most modern currencies are valued on nothing more than trust – even if most users of currency never realise.

DKP systems effectively create many different currencies, each balanced and exchanged between a tiny number of people. The economy this creates is so devoid of complex economic mechanisms, and so obviously balanced by social interaction, that it might be mistaken for barter; but it isn’t.

Applying DKP Elsewhere

The value of DKP is in the ability of a group to allocate their own collective set of values to the results of collaborative activity. The value of the currency is a reflection on the group itself.

Applying DKP to a wiki-type collaborative environment is problematic: Contributions are not equally balanced within the group – the classic 1%-9%-90% pattern, where most contribute nothing, and few contribute a lot. While DKP might seem an ideal way to resolve this imbalance, and give the 1% the credit they deserve, we must remember that the DKP system’s balance is a social construct: The system will naturally be primarily designed by the 1%, and so will be biased to reflect their needs or perceptions of value. So DKP resolves nothing.

Administering DKP tends to be complex and time-consuming. DKP is not just technically complex (which might be eased through better software tools): Its value-system is an ever-changing function of the group itself.

That all assumes DKP will always be established through negotiation between those involved. We could theorise that eventually standard approaches will develop, that later generations of players will come to recognise and accept a standard approach. But standardisation would merely create another traditional currency system. Such a currency would be less arbitrary than some formal in-game currencies, since its value would genuinely reflect the work of players, and would not have to be carefully balanced by those designing the world.

It is not clear that DKP can be applied to any collaborative situation. However it may form a currency that better reflects players’ effort than one designed by those operating the virtual world. Consequently it does have a lot of potential for further development.

Paul Saffo on The Revolution After Electronics

Paul Saffo spoke to Stanford’s Media X conference on the art of predicting the future. Specifically predicting which technology will come to dominate the next decade. Paul’s talk may at first seem somewhat contradictory in nature: Demonstrating how to do it, while simultaneously showing it can’t be done. This article summarises the talk.

30 Year Cycle

Every 30-50 years a new science turns into a technology. With approximate dates:

  • 1900: Chemistry
  • 1930: Physics
  • 1960: Electronics
  • 2000: Biology

We are now on the cusp of a revolution from electronics to biology. The precise inflection point, the point of change, may not yet be clear.

Paul noted that Thomas Watson’s famous misquote, “I think there is a world market for maybe 5 computers”, was made in 1953, right on the cusp of the electronics revolution: Aside from the fact that he was talking about a specific machine, and not all computers, the quote is a good example of how it is difficult to predict the future at such points of radical change.

Forecasting the Future

The goal is not to be right, but “to be wrong and rich”: It is easy to take the view that one cannot forecast. If you do attempt to forecast you will still mostly be wrong, but the very act of trying will increase your chance of success over those that do not try.

The further away from a point in time you predict into the future, the greater the level of uncertainty. The difficulty in forecasting is finding a balance between being too narrow and too broad. Forecasting might use wildcards. The “hard part” is to be wild enough.

Typically forecasts for a new product or technology’s introduction are linear: The magnitude of the amount of use of the technology is forecast to grow steadily with time.

Reality tends to be represented as an S-shaped curve: In the early stages the magnitude of use is below the expectation generated by the linear forecast. Usage then rapidly grows, such that the actual usage rises above the prediction in the later stages. The result is that in the first part, forecasters tend to over-estimate performance, while latterly they under-estimate performance. Venture capitalists tend to have linear expectations, and so are disappointed in the early stages, while failing to see the later potential.

Robots and Inflection Points

Stanley, winner of the 2005 DARPA Grand Challenge. Paul Saffo used the example of DARPA’s annual competition for robot-driven cars. In the first year only a handful of competing robot drivers made it out of the starting gate. No car completed the challenge. The next year 22 out of 25 robots got further than the leader in the first race.

The example gives a quantifiable measure of how the technology is developing, year to year.

Spotting the inflection point, the place at which real, dramatic change starts to occur, can still be hard. Sometimes it can be spotted using data which has been ignored or hidden. Sometimes it is a case of looking for what does not fit. The anonymous quote, “history doesn’t repeat itself, but sometimes it rhymes”, is apt. Look back in time as far as you look forward.

The good news is that if you miss an indicator, you still have lots of time to spot another.


Paul contested that the last three decades had been characterised by a dramatic cheapening of a component technology, which in turn had led to the widespread use of a product:

  • 1980s: Cheap processors led to the processing age. The result, widespread use of PC.
  • 1990s: Cheap communications lasers led to the access age. The result was the network infrastructure to support the World Wide Web.
  • 2000s: Cheap sensors are leading to the interaction age. Applications are currently missing, but widespread use of robots appears to be the future.

Biology and Electronics

Electronics is building biology, and Paul expects that eventually biology will rebuild electronics: These technologies are far from isolated.

An example of developments in electronics progressing biology can clearly be seen from work on the human genome. A well funded government-backed project was beaten by a far smaller project. The smaller project was able to successfully deploy robots, with the results that the cost of the work dropped by a factor of 10 each year. The government project had been funded based on the cost of technology at the outset, and initially failed to adequately respond fully to the changing cost structure.

The creation of the first artificial genome in January 2008 may yet prove to be the inflection point.

Trust Instincts at Your Peril

“Assume you are wrong**” (** and forecast often)

Paul used the example of the sinking of a US naval fleet near Honda, on the west coast of the United States, on 8 September 1923. The fleet had been navigating using a forecasting technique called “dead reckoning”. The coastline had a (then) new technology available to assist navigation – radio direction finding. This allowed a bearing to be given between a land station and the fleet.

The radio direction finding gave an unexpected result that did not match the forecasted position. The lead boat in the fleet concluded that their position was more favourable than anticipated (closer to their destination), and turned sharply… straight into the rocks they had been trying to avoid. The 11th boat in the fleet did not trust the judgement of the lead boat, and when the fleet turned, it hedged its bets, slowing and waiting to see what happened. It was one of only 5 ships from the fleet not to run around.

The morale of the tale: Hedge your bets, but embrace uncertainty. Or as written once on a tipping jar:

“If you fear change, leave it in here.”

Divergence of the Species

The question was asked, will biotech lead to a further aggregation of wealth? Yes. The electronics revolution had itself deepened inequality. Biotech raises a particularly ugly spectre which extends beyond wealth, to life itself. The wealthy would be likely to use their wealth to extend their lives. The ultimate outcome – species divergence. Currently the rich tend to benefit from better health care, and so extend life. But biotech is likely to create a lot more options.

Michael Malone on The Protean Corporation

Michael S. Malone is perhaps best known for his work defining the “Virtual Corporation” in the early 1990s. At Stanford’s Media X conference he proposed the next iteration of organisational development – the Protean Corporation. The topic forms the basis of Malone’s next book. This article is based on his talk.


The total number of consumers is growing exponentially. Wireless broadband covers an ever-increasing amount of territory. The US may become the first truly “entrepreneurial society”, with skill-based work that never last more than a few years, where people never plan to do the same work forever: A mixture of creativity and volatility. The increasing size of the customer base will lead to larger organisations. Simultaneously, competitive threats can appear from anywhere, particularly in fast-moving technology sectors.

The result is two contradictory forces:

  1. Centrifugal: Technology enables workers to be spread out.
  2. Centripetal: Humans still need a sense of legacy and wider purpose; and are inherently social creatures. The “fatal flaw” of the Virtual Corporation was that once everything has been pulled apart, nothing is left.

Throughout history, from pre-corporations (such as early modern trading companies and guilds), through Taylorism to the virtual/adaptive/wired organisations, two trends can be seen:

  • Increased autonomy of employees, with greater communication between them.
  • Reduced management control.

The Protean Corporation

The paradox is simple: How to build an enterprise that lasts, while still being flexible and adaptive?

Michael used the Quantum atom to demonstrate the shape of things to come: An organic form, in constant flux, which retains its core. The design attempts to recreate the structures within Hewlett Packard, where a group of long-term employees remained at the core, with the traditional enterprise formed around them.

The Protean Corporation has three parts:

  1. Core
  2. Inner Ring
  3. Cloud

The Core are the permanent staff – people that have been with the business since as long as anyone can remember – “the immortals”. Their role is to protect the culture of the company, which they do somewhat informally. For example, they might be highly regarded by other employees for their experience or ability to get a result out of the organisation. Largely unseen, they are the people that make the organisation run smoothly. They may not be immediately apparent to the senior management, and as such they need to be protected from a new CEO – they are likely to be accidentally culled along with the rest of the workforce.

The Inner Ring are the traditional full time employees. They manage and operate the business. Their job is to recruit the Cloud.

The Cloud are 90% of the organisation. Their employment might last a matter of hours or days. They might work remotely, never having met their employer. The cloud is so transient that they might make errors before they have time to learn. It is critical that the Core is able to watch over the Cloud, and maintain the company’s culture and standards.

The role of the company’s board is merely to adjudicate and not to manage – to act like the company’s Supreme Court.

Competence Aggregator

The Protean Corporation will be fixed in perpetual motion. The most important role in such a corporation will be “competence aggregators”. These people pull individuals together for specific projects, much like creating start-up companies within the corporation. Competence Aggregators exist within the Cloud, but are still governed by the Core. The Competence Aggregators will be the new superstars of the economy.

Private and Public

The shape-shifting Protean Corporation can exist in both public and private sectors of the economy. To achieve this is the zenith of the concept.

A key problem remains: There is no way to accurately value the Protean Corporation. Its assets are intangible, and not reflected in conventional accountancy-based corporate market valuation. It is a similar issue to that which limits social entrepreneurs – there is no way to measure the performance of non-profit organisations. The ultimate limitation on the whole process is the lack of a market for intellectual capital.

The Protean Corporation in Practice

Wikipedia was cited as an example of a protean-like corporation.

I personally recognise the existence of both the Core, Inner Ring and Cloud from the Open Directory Project. The Core was part-formalised as “Editalls” – floating editors that had no fixed role, but which were always trusted and experienced veterans of the project. The Inner Ring consisted of “Meta” editors (and later Admins), who appointed everyone else, and took the formal leadership role. However, neither group entirely matched these roles. The Cloud, the regular editors, were just as described by Malone: The majority of the organisation, often with very limited ties to the project, many moving on after a short period of work.

The Core is also commonly found within British local government: In most long-established authorities there are a handful of people who both provide a sense of stability, and can simply get things done that nobody else can (usually through some combination of contacts and experience). Without these people I suspect that much of local government would be rendered totally dysfunctional (as close to collapse as a public body can become).

It was noted that Intel had originally shunned the concept of the Virtual Organisation, yet had subsequently developed into one “by walking backwards into it”. For example, only 20% of its “employees” are now traditional permanent staff. Far more contribute “virtually” or as suppliers. Yet all need access to company data and systems, so have to be trusted. A fifth have never met their boss face-to-face, and half of those never expect to: Such an organisation is logically already facing the challenges that the Protean Corporation seeks to answer.

Michael S. Malone’s book is called The Future Arrived Yesterday: The Rise of the Protean Corporation and What It Means for You.