Six came back. Four were useless. One he already knew. The last one made him uncomfortable. That does not mean the system produced one useful answer and five failures. The four useless answers eliminated territory. The familiar answer showed that the obvious explanation had not been missed. The uncomfortable answer found an edge he had not known was there. He does not yet know whether it is right. The discomfort tells him where to look next.
What makes this possible is that the marginal request costs almost nothing. Several answers to the same question arrive at the same moment: different explanations, objections, framings, analogies, continuations. There is little reason to stop at the first plausible one. Try the opposite. Keep the mechanism but change the example. What if that assumption is wrong. Each additional request opens another path at a price low enough that the path need not pay off.
In a conventional brute-force search, the person defines what is being sought and the machine runs the possibilities. Here the arrangement is reversed. The machine proliferates the candidates and the person takes them one by one, running each against experience, knowledge, taste, intention, bodily response, the world. This one doesn’t fit. This one almost fits. I already knew this one. Why does that one bother me? The rejection produces information, and the information changes the next request. What the person is doing across those trials is discovering the criterion, which could not have been specified in advance.
A low hit rate is therefore not a poor return. It is the demonstration. Five non-hits are affordable when they locate the sixth and constrain the search that follows. The important output may not be the answer at all.
He reads the uncomfortable one again. It is still wrong. It is wrong in a direction he had not considered.
WE&P by: EZorrillaMc&Co
Travel in Life 