PORTLAND, Ore. Armed with cognitive models of human behavior, Sandia National Laboratories is aiming to enhance soldiers' performance with knowledge augmentation, while simultaneously diminishing our enemies' effectiveness by second-guessing their next move.
"We knew that to model humans on our side or the threat side, we had to have higher fidelity in our cognitive models. We needed models that really behaved as humans behave," said John Wagner, manager of Sandia's Cognitive and Exploratory Systems and Simulations Department at Sandia National Laboratories.
Sandia National Laboratories recently passed a milestone in its quest for higher-fidelity models, answering an internal Grand Challenge to build a cognitive model that would function like a human and automatically instantiate it into models of particular humans from writings about them. Now they say their cognitive models can predict any person's future behaviors just by inputting text about them, their daily activities and travel records--information that can be gleaned from public records, the Internet and private databases.
"Using patterns instead of rules, we have achieved the goal of the Grand Challenge--building a software framework that we can populate automatically from text and spatio-temporal behavior," said Wagner.
As a result of its success, the Cognitive Science and Technology Program has been upgraded to a strategic initiative, thereby making it a permanent part of Sandia's national security development efforts. It will be funded internally with $2.8 million for 2007-2008. The two-pronged effort will build models of individual soldiers to augment their knowledge, as well as models of the personalities of threats like Osama Bin Laden, to help predict their behavior.
Sandia's cognitive models aim to encapsulate the expertise of specialists, improving the training experience and shortening the time necessary to become competent in new skills.
"If I can build a cognitive model of you as a war fighter, then I can use that model to improve your performance. I can prevent you from making mistakes, I can help you get situational awareness faster, I can give you a near perfect memory, I can manage your workload so you never feel fatigue; there are a lot of things that I could do to augment your performance," said Wagner.
By switching from outmoded rule-based expert systems, Sandia chose to instead use pattern-based artificial intelligence that employs semantic networks to store knowledge and statistics, thereby predicting actions. They also wanted their models to feel like humans feel, which meant including simulations of fatigue and other emotions from real situations.
Now the researchers believe they have the tools to build cognitive models of people perceived as potential threats, in order to predict their behavior in response to current events.
"The important factors that always dictate what our response to a threat will be [are] our predictions of what the threat's actions will be next," said Wagner. "So we are not just doing human factors modeling anymore; we are trying to model the decision-making processes of a human based on his most recent patterns of behavior."