Creating Artificial Personalities, not necessarily Artificial Intelligence

As I have often written about here, there is a difference between a simulation and an educational simulation. A simulation is a model of something, often predictive and diagnostic. An educational simulation is an experience that includes some simulations, balanced with other things, all for the sake of creating transferable behaviors or perspectives. Thus a simulation of a cockpit may have every knob, while an educational simulation during an early level may just have the most critical ones.

Likewise, in educational simulations, there is an increasingly need to differentiate between creating and using an artificial intelligence and an artificial personality.

Artificial Intelligence

It seems that AI's generally fall into four overlapping categories:

• Expert system: A system used to solve a task that provides judgement or other help that takes the place of or augments an experienced professional. A doctor may use an expert system to suggest reasons what might be wrong with a patient.
• Model of human brain: A computational model that strives to both use the same techniques as a human brain and provide similar output. For example, using an artificial stroke on a neural network may not destroy any data, but simply make it take longer for the program to access it.
• Adaptive problem solving: An algorithm that has a goal, sensors, and parameters, but otherwise may grow and evolve in response to an unpredictable environment.
• Skynet.

However, few educational simulations or serious games will delve into any of these AI areas. They are computationally difficult and intensive, and as with the cockpit example, not always the best to meet learning objectives.

Having said that, we still need to feature avatars in any program with a learning objective of helping people "deal" with other people. And these avatars cannot simply be driven by broad branching algorithms. They have to be dynamic and responsive.

Artificial Personalities

Literally, as crass as this sounds, we have to have avatars that can serve as backboards to enable players to repeatedly practice actions and calibrate their actions against the response. These avatars need to be consistent, and often have visualizeable inner workings. Let me give two broad examples from two projects that I have done.

In Virtual Leader, the model of leadership frames that a leader has to

• surface relevant ideas,
• ensure a power base,
• moderate the tension and working environment,
• before driving hard to complete the right work.

The artificial personalities that are represented by the avatars in Virtual Leader, then, all represent distortions of that model. Some characters just wanted to complete any work and were intolerant of side conversation. Others wanted to moderate or lower tension no matter what. They would switch off tough ideas and praise people a lot. Still others wanted to amass power, sucking up to power figures, building alliances using common assets for personal gain, and taking ownership no matter what. (Now admittedly, this model was fairly computationally heavy. )

In a new simulation, I am modeling the personality of single individuals through the three distinct attributes of (a) what the person is doing, (b) what the person is intellectually thinking, and (c) what the person is emotionally feeling, all on a single conceptual map. As a result, the avatar may be asked to do something, and then do it once, but then abruptly stop doing it (showing compliance not commitment), because his/her hands are not aligned with either head or heart (i.e. it is one thing to wear your seat belt once, it is another thing to see yourself as someone who wears seat-belts, and to understand the risks of not wearing one). Or as the avatar gets stressed, the influence on his/her behavior begins to be influenced more by where his/her heart is than where his/her mind is. This may seem complicated conceptually, but the algorithms and modelling are no tougher than making a sim of a bunch of cars driving around a parking lot. Further, the artificial personality is totally transparent and can be visualized to a student of the sim.

Key Attributes of Artifical Personalitites

Given this, the three key attributes of an artifical personality are:

1. Computationally light-weight (runs in Flash), and more often a form of fuzzy logic;
2. Provide behavior models that are instructional to engage, aligned with the learning goals;
3. Are dynamic enough to both respond to a student in an open-ended way and also allow for a programmer/level designer to create (at least) three or four distinctive varieties (a good physical analogy would be how shifting just five variables can change the behavior of a vehicle from dynamically handling like a sports car to handling like a garbage truck).

Leading, not Following, Academics and Computer Games

This work into creating satisfying Artificial Personalities typify the challenge of our industry. On one hand, we don't get much help from either academic research (my degree in Cognitive Science from Brown University is not overly useful here) or computer games. And we will never create characters of indisputable accuracy and predictiveness to real world situations (a primary success criterion for academics). But on the other hand, we can create experiences that radically change and improve the behavior of students (a primary success criterion for educational simulations), are really engaging, will work their way over to games in some form, and we can still create small algorithms that turn out to be more powerful than dissertations in illuminating behaviors of real people.

My fourth book, Learning Online with Games, Simulations, and Virtual Worlds, available for pre-order

I am pleased to announce that my fourth book, Learning Online with Games, Simulations, and Virtual Worlds: Strategies for Online Instruction , is now available for pre-order on Amazon. This book goes into selection and implementation detail around Highly Interactive Virtual Environments. Along the way, it prepares instructors for the new challenges of developing a true culture of interactivity and helping students "learn to do" not just "learn to know."

This book is aligned with, but not part of, my Simulation Trilogy, which will be complete this summer, and is made up of (and can be read in any order):

1. Simulations and the Future of Learning: An Innovative (and Perhaps Revolutionary) Approach to e-Learning (the blue one)
2. Learning by Doing: A Comprehensive Guide to Simulations, Computer Games, and Pedagogy in e-Learning and Other Educational Experiences (the red one)
3. The Complete Guide to Simulations and Serious Games: How the Most Valuable Content Will be Created in the Age Beyond Gutenberg to Google (the green one)

I realize there may be some confusion between them (especially because so far, no new book replaces an existing book before it), so feel free to ask any questions about what subjects each will or will not cover, and I (and hopefully others) will try to answer.

As anyone who does this type of research knows, kind words go a long way, so thank you to so many who have encouraged this continued effort.

Need Help in Collecting Examples and Methodologies of Evaluating Simulation Effectiveness

I have asked for, and received, examples of rigorous evaluations of sims from many people in the past. Over the next few months, I am now going to assemble all of those along with new reports. I would like to be able to better both frame and answer the questions, "What can we expect Sims to do?"

This is not a report for a specific client (although a few have kicked in - thanks!). Thus I can share everything, and will make my journey as transparent as I can. If you have any work that you feel has helped you down this path, I would be grateful to learn about it.

Best,

Clark

Virtual Worlds, Games, and Simulations: The Challenges of the Next Five Years

Across my own client base, there are increasing questions about how the next five or so years will play out. Here are some thoughts on the peeling away of the "post-literacy" onion.

2009: Understanding and Procuring forms of Highly Interactive Virtual Environments

Organizations will continue to explore their understanding of Highly Interactive Virtual Environments. They will seek to resolve the differences in structure, use, best practices, and metrics between virtual worlds, serious game, and educational simulations.

More organizations will purchase access to virtual worlds. In corporations, they will use them primarily for building communities and bridging distances, although about 80% will be greatly underused.

Meanwhile, organizations will commission their own full, self-contained sims from scratch for foundational skills sets (mostly using external vendors, which will continue to have 8 months to 1.5 years development time), and slowly buy (and often modify) off-the-shelf sims. We will see a proliferation of stand-alone mini-games (although often connected to online communities) as the dominant model

Both socially focused virtual worlds (as opposed to the still under-developed area of virtual worlds as platforms for sims) and self-contained sims, when done well, will work better than people realize, creating a rethinking of the multitude of flawed assessment methodologies such as Kirkpatrick. However, organizations will still pursue the Sisyphean task of "managing through metrics," trying to come up with an ROI for either an active community or the development of save-the-company skills.

Within universities using virtual worlds, the activity of students building basic interactive content will become a critical and increasingly dominant rationale for the continued use of the environments. Schools that do not focus on the students' role of building interactive content will wind down their use of virtual worlds for easier tools, such as enhanced virtual classrooms with basic 3D emotiveness. At the same time, the military will continue to push the leading edge of using sims to develop soft-power through the application of critical interpersonal skills.

Newspaper publishers and book publishers, as well as schools and internal corporate training functions, will find themselves in increasingly dire shape. And the sale of interactive applications, such as via iTunes and for Android, will continue to flourish. IBM will launch a new initiative into this space.

2011: Authoring in HIVE environments and presenting increased complexity

In about three years, we will see the widespread availability of robust and easy to use authoring tools and environments, mostly with the functionality described in the upcoming The Complete Guide to Simulations and Serious Games (and foreshadowed in http://www.pivote.info/). While small vendors will be launched to initially meet these authoring needs, these tools and capabilities will be increasingly be aggregated by the biggest software vendors. And the availability of these tools will enable large organizations to bring in-house sophisticated authoring capabilities (as students who grew up authoring in Second Life enter the workforce). The time it takes to build a useful sim will be reduced asymptotically to about four weeks, but larger budgets will be available for more sims that take multiple years to build (the short and the long of sims development reflect the maturity of the tools and the market value of the potential respectively).

New genres will continue to be created around "learning to do" content, embracing the simplifications or pieces of the Big Skills (aka Twenty-first century skills). Linear content will be viewed with increased suspicion as thin and ineffective,

Distinct parts of Highly Interactive Virtual Environments will merge and be seen as a continuous whole. Students and teachers will expect the smooth ramp up and ramp down between the real world, the open virtual world, the fun game, and the relevant simulation.

The communication gap between students and teachers will be a national epidemic. Institutions supporting schools will try, and fail, to build simulations around traditional content, such as biology and literature. Second Life will go out of business as the last corporate customers follow the younger generations to better looking and more dynamic, but also more splintered, environments. Ironically, as the virtual world market fragments, the plaftorm for sims will converge. Adobe Flash will run everywhere (including hacked future versions of Xbox's and Playstations) and be the common authoring environment of choice, enabling schools to assign sims without babysitting hardware. IBM will launch a new initiative into this space.

2013: Rethinking Knowledge and Subject Matter Experts

By 2013, we as a culture will finally be rethinking the possibilities and necessities of captured wisdom. Research organizations and consulting groups will reluctantly reject the easy lens of linear content, and due to competition and client requests follow a Research and Analysis Process for simulation based content, even when not building a simulation. Reports will talk about actions, systems, and results, not just processes and tips. Search engines will be significantly challenged, with huge investments and infrastructure trapping them in old content, as people realize, "You can't learn leadership from Google."

Increasingly, everyone from The MacArthur Foundation to Accenture will default to producing interactive content over passive. Reports produced will not be binders but experiences; not bullet points and inspirational quotes but equations, interfaces, and dynamic relationships. This new research will cycle back into increasingly detailed simulations themselves. Journalism will go away as a distinct college major and career.

Open source simulation design will flourish, and be compatible with the professionally created content. With the popularity of the $49 laptop, China and India will both announce that a majority of their school curricula across all ages will be simulation based. Computer game makers will enter the educational simulation and serious game space for real here, as they see there is a market for finished goods, but they will be too late to create a real brand. They will still manage to wipe out large tracts of smaller companies. IBM will launch a new initiative into this space.

2015: Applying the New View of Knowledge and Wisdom

Moving forward, school curriculum in the US will then be really retooled around teaching innovation and stewardship (and other Big Skills or "21st Century Skills"), as the seminal philosophy of Tim Ellis are rediscovered and embraced. The first Pulitzer prize to a simulation will be announced in 2015, as well as the last use of multiple-choice standardized tests (after years of decline). The last textbook publisher will fold. Pure linear content will be looked at the way we listen to scratchy phonographs. Finally, and truly, the most valuable content in the world will be educational simulations and serious games. IBM will launch a new initiative into this space.