Page:Sm all cc.pdf/111

 most search problems cannot be cast in this form. Furthermore, the technique fails if one of the answers may be incorrect.

An old parlor game begins with everyone except the subject selecting some object in the room. Then the subject attempts the seemingly impossible task of identifying that object. As he wanders around the room, he detects people’s involuntary reactions to his proximity to the target object. Surprisingly often, the following technique helps to solve equipment problems: take the equipment apart and put it back together. Perhaps this procedure will reveal the problem (e.g., a corroded connection) or perhaps it will solve the problem without revealing it (e.g., a loose connection). This procedure is counterproductive, however, if one puts the equipment back together incorrectly and creates another problem; it is much harder to troubleshoot two independent problems than to identify one.
 * Use a search method that indicates both direction and distance to the object at each search step. On the few occasions that this technique is feasible, it is extremely efficient.
 * Consider the probability of missing the object even if it is in the search path. Decide whether it is more efficient to do a detailed search first or to do a quick reconnaissance first and, if unsuccessful, then do a slower search of the same area. Do not eliminate any area unless there is virtually no chance of having missed the object.
 * Consider possible impacts of the search itself both on the object and on the difficulty of finding the object. For example, will the search technique preclude re-search? If so, we must be certain that the initial search does not overlook its objective.

Searching for submarines requires one to consider the impacts of the search: the submarine may detect you before you detect it, and it will take evasive action. Naturalists have the same problem with finding wildlife. Hunters sometimes take advantage of this phenomenon by using beaters to flush the game.
 * For multidimensional search problems, search one dimension at a time. For example, when searching an area of ground, run parallel lines, with minimal but finite overlap of swaths. For multiple independent variables, check one variable at a time. For example, when tuning several controls for maximum sensitivity, maximize response with each separately, then repeat the procedure if interactions are possible.
 * Consider the possibility of the object being present or visible only intermittently. This possibility applies not to physical objects but to a problem or searched-for phenomenon. Intermittent phenomena usually require long-term monitoring rather than, or in addition to, a more conventional searching of different areas at different times.
 * Consider the possibility that two independent factors are helping to hide the object. Would the search procedure be successful regardless of the relative importance of these two controls, or does it