DARPA interaction opportunity

>> The decision aids will assist the commanhder and staff in generating,
>> assessing and visualizing the consequences of employing diplomatic,
>> military, information operations and economic (DIME) actions, singularly or
>> in combinations, to achieve effects against the adversary's political,
>> military (air, land and sea), economic, social, information and
>> infrastructure (PMESII) systems. The decision aids will also assist the
>> commander and staff in constructing, visualizing and evaluating campaign
>> plans that exploit the impact of multi-dimensional effects and the
>> interaction among effects.

This sounds to me like a typical neural-net simulation. I guess,
reinforcement learning is the strongest candidate for the simulation
framework: RL tries to discover a policy for selecting actions that
minimises the expected cumulative cost, based on the estimated dynamics
of the environment. Visualisation specifics will depend on the
complexity of the simulator (number of nodes and links), but should be
zoomable at the very least. User actions are of two main types: build
the model (specify nodes, links, weighs, etc.) and run the model
(specify parameters of the existing model). The UI for both types of
actions can be reasonably straightforward.

The truly smart part here is not interaction design, but how one
determines the consequences of different actions (states of the model).
That's the intelligence of the subject matter, rather than one of
design.

Lada

p.s. You might get well-off on DARPA money, but the process is such a
pain in the neck...