Moderated Colloquium/Working Groups

Dave Garvey moderated an open session where the participants returned from their working groups and discussed search issues pertinent to their current research and operations. The comments included the following:

• To get empirical data on search you need real people conducting real operations – the blue forces. Training exercises are not experiments, for the simple reason that often it is inappropriate for blue to fail at a training task, but true experiments must be able to fail. Also, embarrassing output may result from a training event, so in lot of cases it is hard to get truthful blue data because no one wants to look bad. It is unfortunate that the academies teach that the winner learns from the winner. “If it looks good it must be true” is a dangerous mindset in creating a true user interface. Ultimately misinformation, bad data, leads to poor results.
• In a search and acquisition situation, researchers ask, how do you quantify synergy and the added value? The boss does not vote, he makes the final decision; collaboration cannot be consensus in the Navy. Time is a priceless commodity in search and you can’t waste it by talking. The problem of quantifying sensor fusion in multiple networks needs to be addressed. Access through hardware interfaces and data standards (which are almost an afterthought) is required in order to increase possible search time.
• Blue force tracking needs to be assessed; law enforcement is a good example of a success measurement.
• In focusing on homeland security, border defense is a huge sorting problem whether concerning the number of containers in port or crossing the Canadian border. Dave Larson researched night vision goggles and visual searches, and he can still only predict lateral range to 27%.
• Weapons of Mass Destruction are another problem. The Center for Disease Control can use models to optimize the use of inoculation serum; a critical parameter would be how long the host goes undetected.
• The utility curve of information is nonlinear; just examine the coordination between target location error and Circular Error Probable. Consider if the point aimed at is the point that will actually be reached on earth. It is necessary to have the same range and bearing on the target point.
• Acquisition facilities are similar to hard physical object searches; finding your target is still a payoff. It has always been assumed that when aiming for the biggest case, everything else is a lesser case, but in the last 15 years, this has been proved incorrect. Algorithms and methodology set up to work at one distance scale will not always work on other scales.
• Another application is the demographic search (for example, social/market trend search for marketing). National strategy is an issue and policies can be set to determine risks and uncertainties.

Closing remarks by Dr. Brian McCue:

Since the moderated session discussed open problems of search, it led naturally into Dr. McCue’s session to suggest next year’s topic. After deliberation, the group determined that it would be useful to explore the intersection of exploration and exploitation: not just how to find something, but how to find something in such a way that the thing found can be successfully dealt with. Playing off a popular advertisement, Dr. McCue suggested that the topic is that there is life after search.

The concept of life after search opposes Koopman’s uniform premise that the search is over when the target is found. After the target is found, the searcher still has work to do. Some of it is target processing: in searches for smugglers, the searcher works on container detecting or rubbing swabs to see what is being smuggled. This processing takes away search time from other targets that need to be found, a balance that Koopman didn’t address.

Another consideration is that the target may be processing the searcher. This may be good because it means the target is not searching for other targets during that time. However, the searcher may think that the “game” is over when the target is found, even though the target may now be processing the searcher.

Koopman barely addresses the information the searcher obtains by not finding the target. The point of uniform optimal search may contain information as to where other targets are. In terms of negative search, the searcher is usually happy not to find the target in a place where the searcher doesn’t want to find it.

Dynamic re-allocation exists because of false targets. If the searcher believes something to be a target, then realizes it is not, he may have departed from the search pattern. The searcher is then required to dynamically reallocate to continue the search. With the advent of new sensor development empirical work also needs to be done. Koopman also ignores the herding effect, which increases density of targets and possibly pushes the target into the searcher’s lane.

Enemies could also use decoys to pull the search off the target; for instance, the Coast Guard faces smugglers who use many decoy operations. A common maneuver for smugglers is to place three cocaine containers on their ship. One container, deeply hidden, holds the real supply. The other two are the decoys, one holding a small amount of the drug, the other empty; each placed where they could be found. The smuggled cargo then has a greater chance of going undetected. Small drug-running boats use a similar operation. Four runners are sent out as red herrings while the drug-filled boat has the opportunity to get past the Coast Guard undetected.

An example of the Hidden Object Problem is the ratio of fewer nuclear bombs to bombers. As the number of bombs goes up the number of bombers goes down. Game theory migrated into academia in the 1960’s but recently has made a comeback in the Air Force. A lot of CNA work never got declassified out of laziness; if the classified works of the past were made public, researchers would be able to work out many more issues.