Whyville and Numedeon


Late last summer I posted a technology based explanation for Why Whyville exists.  Recently, I was asked to provide:

 “a comprehensive understanding of what Numedeon (the company that runs Whyville) does and what it aims to achieve “

I hope readers find the answers not only uncharacteristically brief but also interesting:


The Problem:  What is the primary problem we are trying to solve?  

Numedeon Inc. was born not so much in response to a specific problem, but instead as a result of what we saw as an enormous opportunity to use the Internet to positively support and affect the lives of children.  In launching Whyville.net in 1999, Numedeon established the first virtual world Internet site that used games, social networking and a strong community structure to engage children directly, providing a platform for our users to explore and understand an ever widening set of issues relevant to their lives and futures.  Now having engaged literally millions of children, Numedeon continues to be driven by the opportunity and the need to address subjects that improve children’s lives, from nutrition, to social responsibility to mental health.  Numedeon also increasingly plays a role in helping more traditional organizations including governmental organizations; NGOs and schools reach and engage a generation of children now generally referred to as “digital natives”.

Entrepreneurial Insight: What is the core revelation informing your approach?   

In the mid 1980s, Numedeon’s founders were already inventing how to directly engage children using online play (i.e. games) on networked computers.  Learning from our users, we soon realized that the Internet was most effective not as “push”, but instead as  “pull” technology.   While most efforts to influence children then and even today continue to push information at them, Numedeon’s social game-based technology pulls them in, in the process making the learning their own.  Critically, through this process we also learned that success depends on forging an active partnership with our users, allowing them ownership through their own creative participation.

The Program Model:  What are the key elements and how do they fit together? 
14 years later, Numedeon is still involved in actively running, managing and expanding one of the largest and most engaging virtual world spaces for children on the Internet.  Half of our employees are responsible for running our sophisticated community management system, judged by many to be the best on the Internet.  The rest of the company is involved in designing and building new Whyville learning projects in collaboration with an ever-growing list of sponsors, many of whom have already been working with us for many years (e.g. The Getty Museum, NASA, the CDC, WHOI, the Field Museum).   In addition, every member of the company spends time in Whyville (as the famous “CityWorkers”), interacting with and learning from our users.   Increasingly, Numedeon is also providing support for teachers and schools using Whyville in their classrooms.
Direct Impact: What are the primary results of your organization’s work?
With 7.5 million registered users, Whyville.net is one of the largest virtual worlds of any kind and absolutely one of the largest serving education.  With an average of 30 minutes per login, Whyville is also one of the stickiest sites on the Internet for kids.  Depending on the time of year, use ranges from 100,000 to 300,000 uniques per month, average age 13, and 72% female.   On average users stay active for more than a year, with many engaged over multiple years.  With the company’s focus on learning impact, our virtual world engine provides a wealth of detailed use information.   Typically, Whyville activities attract hundreds of thousands of users, playing millions of games.  The company also has an open door policy towards academic research, making Whyville one of the most independently studied educational web sites.   Whyville has already been the subject of one book with a second currently in development.  (I am writing one as well  JB  ).

Systemic Impact:  articulate your theory about how, in the long-term, macro-level change might result from your work.

Arguably, nothing has greater potential for positive systemic social impact than effective education.  For 14 years, Numedeon has been a principle innovator in the use of digital technology to deeply engage children in learning.  However, when Whyville launched in 1999, few teachers, schools or parents understood the potential educational power of social gaming or virtual world technology.  Accordingly, Whyville was launched as an informal educational website with no direct connection to the formal education system.  Now 14 years later, Whyville’s use in the schools is growing, and Numedeon is receiving increasing support for classroom-based interventions.  As the nation’s educational system continues to “go digital”, our years of accumulated expertise, metrics, and large Internet foot print put us in a unique position to influence the dramatic changes sure to result, while also providing a wide range of organizations more effective mechanisms to directly reach the children they seek to serve.


Vision:   Where will the world be in 2050 and how will you have contributed?

The structure of the modern educational system, its schools, classrooms, and textbooks were established in the 15th and 16th centuries (see Momento Mori below) to deal with a fundamental problem in scalability:  Growing populations needing to be educated by a still small number of educators.  The Internet provides a fundamentally new way to scale learning, the consequences of which for 2050 are already manifest now.  The Internet breaks the distinction between formal and informal education, bringing the real world into the classroom and vice versa.  Web-based learning is self-paced, and learner-centered with children seeking their own achievement levels.  Information no longer needs to be distilled into a textbook for distribution.  Progress can be measured by actual achievement, rather than indirectly through paper and pen assessments.    And perhaps most importantly, social play will once again be the basis for learning, engaging children across national and geo-political boundaries.  These are all features of Whyville today.


Personal Story:  Why are you here? On the face of it, we are here because two Caltech students, Mark Dinan (physics) and Dr. Jennifer Sun (visual psychophysics), decided that a life in science was not likely to be as rewarding as a life spent improving children’s wellbeing.  While Caltech students, both decided to work for the Caltech Precollege Science Initiative, co-directed for 17 years by our other founder, Dr. Jim Bower, as a hands-on science intervention in California Public Schools.  For his part, Dr. Bower’s interest in educational reform considerably predates his professional interest as a neurobiologist building computer models of the brain.  However, given his computational expertise, in the early 1980s, he already anticipated the power of simulation/game-based learning.   As a former student at Antioch College, Dr. Bower had also experienced first hand the power of student centered inquiry-based instruction.  As the son of a minister deeply involved in the civil rights movement and a psychiatric social worker employed in poor inner city school districts, Dr. Bower was raised in a family that believed both in giving back, as well as in the power of learning and education to change human lives.  Whyville’s success reflects the unique contributions, but shared motivations of its founders.


Critical Decisions:  Why are you still here?

Numedeon’s founders have made a long series of difficult business decisions over the last 14 years to keep the company alive.  At the start, we had to decide whether our decidedly social entrepreneurial intent was best organized as a non-profit or for-profit.  After years sustaining long-term projects with short-term grants, we decided to give sustainability in the marketplace a try.  Second, founded during the rising days of the dot.com bubble, we could have raised considerable development money in exchange for a loss of corporate control.  Suspecting that we were well ahead of the market, and wanting the freedom to develop our approach without the complications of venture capital, we decided to bootstrap the company instead.  That decision, which we do not regret, gave us the freedom we needed, but at some cost in stress.  Through it all, the principle metric that has sustained and directed us, is the enthusiasm of our users.


Why are we doing this and where did Whyville come from??

Well, I suppose, in some sense, the reason that I finally decided to jump into the blogosphere was to try to provide an answer to those questions.  As may perhaps now be apparent, the answer is kind of complex – and has become more complex as time has gone on and it is ever more apparent why we would have done this “had we known then what we know now”.  The Whyville team and I will be forever grateful to Whyvillians for that ongoing education.

However, although still early in the overall exposition, perhaps you will indulge an effort to explain why, in 1984, I THOUGHT I should start playing with games embedded in social (virtual) worlds as a mechanism for engaging elementary and middle school aged children in science education.

The answer then had to do with computer simulation technology and the importance of models as a tool for understanding complex things, Whyville’s core technology.  It also came from a deep sense that many, even in science and especially in biology, did not themselves understand the importance of model-based discovery.

Three weeks ago I helped to organize a meeting in Cambridge England celebrating the remarkable accomplishment of Sir Allen Hodgkin and Sir Andrew Huxley 60 years ago, who built a mathematical model as a tool to understand the way in which neurons communicate with each other electrically.

Without mentioning Whyville of course (few in neuroscience know about my ‘other life’), my opening talk at the meeting asked how it is that the results of the Hodgkin / Huxley model have been largely accepted, while their modeling methods still seem foreign to so many biologists.  After all, IT HAS BEEN 60 YEARS!!!  In fact, while giving the meeting introduction, it occurred to me that the meeting in Cambridge might have been the first in the entire history of biology to be organized around an actual model.   Physics has been organized around and by models for 400 years.

This observation (another example of learning by doing), in turn inspired me to submit the commentary copied below to Nature Magazine, which has just rejected the publication because of “pressure on space in our pages”.

No such pressure here however 🙂  although I realize that this commentary may stretch the interests and willingness of those who follow this blog.  I am hoping that those who do fight through it might better understand Whyville’s origins as well as my own perhaps somewhat over-assertive commitment to Whyville as an idea.


Commentary on:  “60 Years of the Hodgkin-Huxley Model.  In celebration of the 60th anniversary of the publication of the Hodgkin-Huxley model of the action potential”, Cambridge UK July 11-13, 2012.


The contrasting role of standard models in biology and physics: considering the 60 years anniversary of publication of the Hodgkin Huxley Model for the neuronal action potential.

The announcement of the Higgs Boson on July 4 attracted widespread attention among physicists and the general public in large part because it confirmed a theoretical prediction made almost 50 years earlier regarding a particle key to the relation between elementary particles and the forces between them. As such, the discovery of the Higgs boson has been reported as a triumph for the partnership between theory and experimental practice in physics.  A week after the announcement of the discovery of the Higgs Boson, a symposium took place at Trinity College in Cambridge, England, celebrating the 60th anniversary of the original publication of the Hodgkin-Huxley (HH) mathematical model for the initiation and conduction of the neuronal action potential which provides a fundamental mechanism for communication between neurons.  Like the Standard Theory of elementary atomic particles, the original publication of the HH model both unified a diverse set of experimental observations and made a series of predictions for phenomena not yet observed, or at the time observable. As was made clear in the symposium at Trinity College, experimental research in the subsequent 60 years has largely confirmed those predictions and placed the HH model at the center of our understanding of the electrical activity of nerve tissues throughout the animal kingdom from the squid’s giant axon to Human brain cells.

While Alan Hodgkin and Andrew Huxley received a share of the Nobel Prize in 1963 for their work, the success of their model in predicting the ionic processes underlying the generation and propagation of the action potential remains largely unheralded even within neuroscience.  Most neuroscience textbooks instead simply refer to the HH model as a “description” of the ionic basis of the action potential, failing to include any discussion of the scientific process represented by the model, or its role in organizing and leading 60 years of subsequent experimental and theoretical investigations. Typically, there is also no mention of the fact that the HH model today provides the basis for most ongoing efforts to build realistic models of brain circuits and understand brain function and dysfunction.  While the discovery of the Higgs Boson is lauded as a triumph for the Standard Theory of elementary particles, the similar accomplishment of the model built by Hodgkin and Huxley is largely neglected.

Prior to the HH model in the late 19th and early 20th century there was considerable disagreement and confusion about the cellular and biophysical mechanisms responsible for the action potential.  While the action potential itself had been recorded as early as the mid 1860s by the German physiologist Julius Bernstein, there was considerable debate regarding both the ions involved and the mechanism(s) responsible for their movement across the membrane.  In 1937 Alan Hodgkin showed that the action potential depends on regenerative changes in electric charge movement across the membrane, with the change in potential propagating down the axonal fiber.  Contrary to the then prevailing view (associated with Bernstein) that these ionic movements resulted from a transient breakdown in the axonal membrane, Hodgkin and Huxley working together showed that the action potential exhibits a brief transient period when the internal negativity of the membrane potential becomes positive, an “overshoot”, requiring a more sophisticated membrane mechanism than previously assumed.  After World War II, Hodgkin and Huxley returned to their experimental work using a state of the art feedback amplifier to perform voltage and space clamp measurements on the squid giant axon.  Combining the voltage clamp with ion replacement experiments, they measured for the first time in detail the flow of potassium and sodium ions crossing the membrane and their corresponding conductance changes generated during the action potential.

This experimental work was published in a remarkable series of 5 papers in the Journal of Physiology in 1952. While the first 4 described the experimental results, the crowning achievement was the 5th paper, which included the mathematical model itself in the form of 4 ordinary differential equations.  Even today, this sequence of 4 + 1 represents one of the best, and perhaps also one of the clearest demonstrations of the value and proper use of models in biology, exemplifying the links between theoretical ideas and experimental studies.

Hodgkin and Huxley themselves were very aware of this unifying use of their model, making it clear in their paper that more than a description of the phenomena, the model was an essential investigative tool.  Thus, they state:

In order to decide whether these (experimental) effects are sufficient to account for complicated phenomena such as the action potential and refractory period, it is necessary to obtain expressions relating the sodium and potassium conductances to time and membrane potential.”

expressions” in this case of course being the model’s equations which both provided concrete definitions of the processes involved as well as a means to link separate experimental results into a larger understanding of mechanism.  In addition to helping coordinate the experimental results, the model was also used to explicitly rule out mechanisms that were inconsistent with observations:

“…  we shall consider briefly what types of physical system are likely to be consistent with the observed changes in permeability.”


The object … is to show that certain types of theory are excluded by our experiments and that others are consistent with them.”

In this way, Hodgkin and Huxley used the model to test and reject existing ideas about the origin of the action potential, including, importantly, their own:

Consider(ing) how changes in the distribution of a charged particle might affect the ease with which sodium ions cross the membrane … we can do little more than reject a suggestion which formed the original basis of our experiments (Hodgkin et al., 1949).”

To this day, perhaps the highest (and rare) mark of any model is to rule out the author’s own previous beliefs and speculations.

Beyond testing proposed mechanisms, perhaps the greatest achievement of the HH model was in making a series of predictions related to membrane mechanisms not yet described and data not yet obtained or obtainable.  Specifically, the core model prediction was that the movement of sodium and potassium ions through the membrane are independent and controlled in different ways.  While Hodgkin and Huxley could not have known the underlying biophysical mechanism at the time, their model, in effect, predicted not only the presence but also the core functional properties of individual membrane bound ion channels not clearly identified until the invention of patch clamp recording techniques for which Erwin Neher and Bert Sakmann shared the Nobel Prize in 1991.

It is our view that in a science still dominated by descriptive studies, in which the large majority of submitted grants and research projects do not reference or include a quantitative theoretical basis for the work, the history and success of the HH model stands as a testament to the value of modeling, theory and simulations in understanding complex phenomena.  By not emphasizing the predictive nature of the HH model, and the relationship between the construction and testing of the model with experimental data and the subsequent success of its predictions, we deny our students knowledge of a critical component of the scientific process and one of its greatest successes to date.

the plague unveiled at the symposium which now pays tribute to the experimental and theoretical work performed at Trinity College on the neuronal action potential

As was clearly apparent at the symposium at Trinity College, the HH model, like the Standard Model of particle physics, continues to provide the quantitative underpinning for our understanding of the electrochemical properties of the brain and in particular, how its neurons communicate with each other and with the outside world. The HH model and its derivatives also provide the foundation for almost all efforts to build biologically realistic brain models including the compartmental modeling techniques introduced by Wilfrid Rall and his collaborators in the 1960s.  All the major simulation software packages, including GENESIS and NEURON are based on the HH equations, as are large-scale simulation projects like the Blue Brain project aiming to model the mammalian cerebral cortex. These computational efforts, however, continue to involve a relatively small number of neuroscientists and an even smaller number of experimentalists. Perhaps, revisiting the history of the HH model, and presenting the model as a set of predictions rather than the now accepted description of the action potential, might provide a pathway for more neuroscientists, and perhaps even more biologists in general to value, understand, and participate in modeling studies.