====== Abduction (logic) ====== A form of logical inference which tries to find the //best// explanation for an observed phenomenon. Abductive inferences are not necessarily //true//, but rather (at best) //sound//, which means that their result is correct with a high probability. Example: > [//Observation//:] the street is wet. > Possibly the street cleaning has washed the road, or someone has washed their car on the road, or it has been raining. > Of all these possibilities, //rain// is the most likely. > Therefore, we assume (as long as there are no other clues) that //rain// was the reason why the road is wet. The conclusion is obvious, and – as long as one is aware that it only holds with a certain probability – also //sound//. However, one must also be aware that the conclusion must be able to change at any time if further information becomes available: e.g. the further observation that only the section of the street in front of the house is wet, but not the rest of the street; or that we see a neighbour carrying the high-pressure cleaner into the garage, etc. ===== Other Names ===== * Abductive reasoning * Abductive inference * Retroduction * Ἀπαγωγή [apagōgē] / ap­a­go­ge. ===== Description ===== The result of an abduction is not //necessarily true//, but merely a //plausible assumption// that still has to be verified. It does, however, allow us to justify //predictions//, which would not be possible with other forms of logical inferences ([[glossary:deduction|deduction]] and [[glossary:induction|induction]]). In particular, //abduction// is the only one of the three forms of inferences that can be used to //expand// knowledge: we can indeed create //new knowledge// by inhering the most likely explanation for observed phenomena. For more on this, see [[wp>Philosophy of science]]. On the other hand, an inference to probability comes with numerous limitations. If these are not carefully controlled, a number of errors can occur, some of which are described in the following articles: FIXME **To be added later.** ===== Examples ===== ==== Medical diagnosis ==== A typical [[wp>Medical diagnosis]] is a good example of //abductive reasoning//: from the information obtained by means of [[wp>Medical history|anam­nesis]], such as symptoms, previous diseases, etc., as well as from expertise and experience, a //most likely// diagnosis is inferred. Surely everyone has heard stories where such medical diagnoses have subsequently turned out to be inadequate or even completely wrong. Unfortunately, this is in the nature of things, since a //necessarily correct// diagnosis would only be possible with //complete knowledge// of all relevant factors, which is practically impossible ([[:sherlock_holmes_fallacy|Sher­lock-Holmes-Fallacy]]). ==== Sherlock Holmes’ deductions ==== Contrary to what [[wp>Arthur Conan Doyle]] makes his famous detective say, virtually all of Sherlock Holmes’ conclusions – at least those found in the original books – are not //de//ductions but //abductions//. If Holmes concludes, for example, on the basis of the soiling on the clothing, that someone has been in a place where a specific type of soil predominates, this is a //sound// inference and also very likely //correct//, but not //necessary true//. The pollution could also have happened indirectly, e.g. by contact with a soiled carriage, or the person could have worn already soiled clothes - possibly by another person. Undoubtedly, these explanations are less likely than the one put forward by the detective, but in real life they cannot be ruled out as easily as in a novel – where, of course, the author knows what will eventually turn out to be "true". ==== See also ==== * [[glossary:deduction|Deduction]] * [[glossary:induction|Induction]] * [[logic:formal_fallacies:fallacy_of_division|Fallacy of Division]] ==== More information ==== * [[wp>Abductive reasoning]] on //Wikipedia// {{page>templates:banner#Short-BG-Logic&noheader&nofooter}}