While realistic graphics and immersive environments are improving, properly replicating human behavior continues to challenge both model and simulation developers. (U.S. Army)
Graphics are coming along pretty smoothly, and with game engines constantly evolving and upping the realism ante, it seems as if hyper-realistic visuals for simulations — the uncanny valley notwithstanding — are not far away. The more interesting field of development will be properly modeling human behavior in games and simulations. It’s a challenge that will require input from everyone from game designers to psychologists.
The reasoning behind how individuals act is complicated, groups of individuals more so. But when it comes to designing realistic games that will help soldiers train or models to be used for planning, designing a functional, complex, normal world of people is the tricky next step.
Cultural differences, life experiences and access to information all feed into how humans make decisions. Game designers and modelers don’t want to control each avatar individually, but these avatars still need to make cognitive, individualized decisions as people would in the real world.
Furthermore, force-on-force combat is no longer the norm for soldiers. As irregular warfare rises, training soldiers to think in different terms grows more important. What is the impact of building a school or digging a well? What are the consequences of non-combat acts, and how can soldiers leverage them to bring peace and order? How do you roll all of these factors into a virtual training scenario or a planning model? It comes down to data. With more data, researchers can engage in social science to figure out what motivates people to perform certain actions.
The amount of data available for an American shopper in Kansas is far greater than what is available for an Afghan leader in Kandahar. It is collected every time a person logs onto the Internet or buys something from a grocery store or searches on Google.
That said, the government has been collecting information from intel sources, residents, people on the ground, and polls and surveys. There is an ongoing effort to collect data, but gathering it from all demographics and geographic regions can be difficult.
Those data then have to be validated, made accessible, culled for the “important” data that will properly feed a model, and interpreted and tested. It’s not an easy road, but it is a necessary one — something that the Department of Defense acknowledged when it started the six-year Human Social Culture Behavior Modeling Program in 2008.
The inability of simulators to accurately represent friends and foes is an ongoing and critical problem. The way out is to accumulate more data about what drives decisions in various cultures and harness it to breathe life into the simulations and models of the future.