Monday, August 02, 2004

Mind Blogglers/ or paradoxes

This one's been irritating me for two days now:

The taxi-cab paradox
In the town of Greenville there are exactly 100 taxis, of which 85 are green and 15 are blue. A prominent citizen witnesses a hit-and-run accident that involves a taxi, and testifies that the taxi was blue. The witness is subjected to tests that determine that, in similar circumstances, he is 80 per cent reliable in his colour reports. Is it likely that the taxi in the accident was blue?

First, it seems clear that we are entitled to accept what the witness says as highly likely. He has proved 80 per cent reliable in similar circumstances, and there is no reason to think there is any relevant difference in this situation. Surely what he says can be considered highly probable, and should be regarded as such in a court of law.

On second thought, if we take the long view, it seems unlikely that the witness was correct in his colour identification. To see this, consider 100 randomly selected taxi accidents in Greenville. About 85 of these accidents will involve a green taxi and about 15 will involve a blue taxi. If the witness were to report on the 85 green taxi accidents, he would report correctly in about 80 per cent of the cases and incorrectly in 20 per cent. This means that of the 85 green taxi accidents, he would report about 17 as involving a blue taxi. The 15 blue taxi accidents would presumably also yield 80 per cent correct reports, or 12 reports of blue taxis involved in accidents. Were a witness of 80 per cent reliability to report on 100 randomly selected taxi accidents in Greenville, then, there would be about 29 (= 17 + 12) blue taxi reports, only 12 of which would be accurate; that is, only 41 per cent of the blue taxi reports would be correct. So it seems more likely than not that the witness in our original case was mistaken in his report of a blue taxi.
I think I might have a solution to the paradox, but I'm not so sure. How reliable do you think the witness report is, 80% or 41%? If you choose one, can you tell me what is wrong with the line of reasoning for the other answer?
...click on 'permanent link' to read more...

No comments: