The Department of Defense is interested in modeling human behavior in an irregular warfare scenario.
The agent-based model will incorporate the effects of individual DIMEFIL actions at a lower level of resolution than the strategic/campaign level and work out their implications, so that patterns emerge in the PMESII (Political, Military, Economic, Social, Infrastructure and Information) environment, on the level in which strategic and campaign decisions are made. Lower level rules and strategic level patterns will both be in accordance with social theories agreed upon with the government. The agent based model will include and make use of a database of the infrastructure, political, economic, and social states and relations between agents, so as to work out the implications of actions in a particular instantiation of a PMESII environment, g iven the rules of relations from micro level and macro level political, social, cultural, psychological, and economic social theories agreed upon with the government. It will use this data in concert with the application of social rules to give a plausibl e outcome of DIMEFIL actions in the PMESII environment. The simulation will be transparent: analysts will be able to access a complete description of the state of the simulation at output and during processing. The simulation will have a causal tracing tool to aid in finding the patterns of behaviors which ca use the emergent effects. After a preferably automated calibration process, the data and social theory should support the social models at the individual relation level as well as the level of groups of relations and sequences of relations.
As good as magic given the current state of artificial intelligence, but worthwhile research nonetheless.
(HT: DangerRoom.)
I see a big weakness: "Lower level rules and strategic level patterns will both be in accordance with social theories agreed upon with the government."
And which social theories would our government agree to? Would these be the same people who want women in combat, or who don't see any problem with welfare programs that pay more for extra babies? I think of most government officials as being among the least qualified to identify accurate low-level social rules.
BB: Oh yeah, there are a host of problems. Considering that people don't know how to perform the tasks mentioned, it's hard to imagine an AI system that will.