Simulations Show Us Things We Didnt Know We Knew

Peter Coffee: As IT moves from self-contained systems to multiplayer partnerships, lessons from simulation apply to emerging application and network testing needs.

Having despaired last week that mindless idiots control the movie business, its ironic that I should be inspired this week by news of animated characters with brains.

The special effects software called "Massive," which controls the battle scenes in the "Lord of the Rings" film trilogy, reminds me of what drew me into computing in the first place: the idea of populating artificial worlds with independent actors, each controlled by individual rules, and turning them loose to see what would happen--and learning, in the process, things that might help us re-write (or, at least, learn to accommodate) the rules that control the real world.

Application developers and IT architects need to give more attention to this kind of thing as they enlist more complex platforms, such as distributed computing grids and Web services. Emergent behaviors, combining simple actions to produce unexpected results, must occupy a larger share of developers attention as their applications inhabit a less controlled environment.

The payback for taking on that challenge will be high. Simulation expertise developed in purely technical realms, such as network or Web site scalability testing, may give developers entrée to higher-level strategic discussions. Even when we cant actually make the world a better place to live, perhaps we can at least carve out a position from which we can better compete.

Even simple examples prove the point. The behavior of highway traffic invites us to model each car and its driver as an object that responds in obvious ways to its immediate environment, and to see if the group behaves in ways that we might not expect. Sure enough, the dangerous "shock wave" crowding of high-speed traffic quickly emerges. Armed with a mental model that anticipates this effect, individual drivers can at least reduce their risk by placing themselves in the open spaces that arise between the clusters, or even adjust their own behavior in ways that smooth the flow for all. Holiday travelers take note.

For any business affected by group behaviors, at various points in the supply chain or the marketplace, the implications are obvious. Modeling your own behavior, and even merely assuming that your competitors act likewise, may tell you things that help you be the exception to the rule.

That possibility, I hasten to add, is not a guarantee. In one of my MBA subjects at Pepperdine, back in 84, we went through a full-weekend computer-based simulation of a competitive consumer appliance market, with each team playing the part of a company. As it turned out, every team decided to go for volume, competing on price, though not without vigorous debate in (ahem!) at least one case, and all of us cut R&D and quality control and other costs to the bone.

In the subsequent debriefing, one of our professors said hed found it a frightening re-enactment of the U.S. automobile industry before the rise of Japanese competition. The simulation did its job better than we did ours.

If youre going to use simulation, therefore, to model systems in which human decisions are important parts of whats going on, its essential to run things in a way that challenges people to do things that they might never try in real life. If people are afraid to try something wildly different from current corporate strategy, because it would be politically unacceptable for that rogue approach to succeed, then the value of the exercise is lost. Injecting some artificial players into the exercise may free up peoples thinking. One has to consider the possibility that competitors just dont think the same way.

Theres a general problem, of course, with simulation in that it can only even try to reflect behaviors that weve already noticed. A simulation of the orbit of Mercury, if based on Newtons formula for gravity, will be wrong--not because of errors of approximation, tempted though we might be to think so, but because that formula doesnt reflect Einsteins more complete description of what makes things move.

My middle-school son recently worked in his geometry class with a remarkable piece of software, developed at West Point, for designing and testing bridge designs and competing to produce the least costly solution: It doesnt even attempt to model the aerodynamic behaviors that have produced some spectacular failures in real life. Be duly warned.

Tell me what youve learned from simulations.