Newsgroups: comp.ai Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!van-bc!ubc-cs!kean From: kean@cs.ubc.ca (Alex Kean) Subject: Abduction: Summary Message-ID: <1991May6.173412.11283@cs.ubc.ca> Sender: usenet@cs.ubc.ca (Usenet News) Organization: University of British Columbia, Vancouver, B.C., Canada Date: Mon, 6 May 91 17:34:12 GMT First, thank you for all the replies. The "explanations" are very helpful and appreciated. The origin of the word in Greek, as pointed out by Pierre Marquis, is from Aristotle [see also peirce32, pp 497]. The name \'\alpha \pi \alpha \gamma \omega \gamma \'\eta ==> (apagogy) means what Peirce called "abduction" and \epsilon \pi \alpha \gamma \omega \gamma \'\eta ==> (epagogy) means "deduction". The \'\alpha ( a ) is like ( un- ) in English and thus (apagogy) is the opposite of (epagogy). I guess Paul Scott and Michael Covington might be right is saying that Peirce was translating the word (apagogy) from Greek and "ab-duction" seemed the right matching structure. Another key observation (at least for me), as pointed out by Timothy Cain, is that the minor premiss and the conclusion are only PROBABLE. This is also emphasized in Peirce text [Peirce32, pp 53]. I always thought of "abduction" as "deduction" in reverse, which is true in some sense, but the difference is that the "explanations" from "abduction" is not necessary a consequence of the facts or even true. This is different from "deduction" where each inference step (except the axioms) is a logical consequence of the facts. Thank you again for all the replies. Best Regards, Alex Kean Department of Computer Science University of British Columbia #333-6356 Agricultural Road, Vancouver, British Columbia Canada V6T 1Z2 Tel# (604)-822-4912 @book{peirce32, author = {Charles Sanders Peirce}, title = {{Collected Papers of Charles Sanders Peirce}}, publisher = {Harvard University Press}, volume = 2, year = 1932, note = {Editors --- Charles Hartshorne and Paul Weiss} } =============== 1) FROM: Timothy D. Cain First, let me say that my New Webster's dictionary includes the definition "in logic, a syllogism, the minor premise and conclusion of which are only probable" Not exactly what I call abduction, but closer the anatomical definition! Forming a complete proof (all of the leaves of the proof are known to be true) is a deductive process, not an abductive one. Abduction provides a plausible proof, where one or more of the leaves of the proof are not known to be true (or at least, their certainty is not 100%). If any proof of a known conclusion requires the reasoner to make an assumption, that's a good sign that abductive reasoning is being used. A lot of people say Sherlock Holmes performed deduction, but I disagree. He made MANY assumptions to back up his conclusions, which is an abductive process (or at the very least, it's assumptive deduction, but that's a whole new can of worms!). ============== 2) FROM: Michael A. Covington I always thought the word "abductive" was chosen rather arbitrarily because it resembles "inductive" and "deductive". ============== 3) FROM: Paul Scott While I don't actually know why Peirce chose this term, I can offer a plausible guess: The term 'deduction' derives from 'ducere' (lead) prefixed by 'de' which, in this case, means 'from'. I guess that Peirce wanted another term to denote a different way in which one proposition may follow from another. So he looked round for another prefix that could mean 'from'. He found 'ab' -- hence 'abduction'. You will note that this explanation is itself derived through abductive reasoning! ============== 4) FROM: Raul Valdes-Perez I would suggest some selective reading in Charles S. Peirce, "Essays in the Philosophy of Science," American Heritage Series, 1957. I did some reading there, and discovered quite muddled uses of the term abduction by Peirce. In one essay he seems to mean one thing, and in a second he means another. Peirce introduced the term because he wanted to speak of scientific processes other than "induction," understood as merely formulating generalized laws from data. For example, he wished to discuss the proposal of explanatory theories, such as theories that provide a physical mechanism giving rise to observed data, in which data are qualitatively quite distinct from theory. Such an abduction would be the proposal of the mechanism underlying blood circulation in the body, starting from data on blood properties, what happens when an artery is artifically constricted by a tight knot (you get a bulge on one side but not the other), etc. Peirce also used synonyms for `abduction,' such as the `method of hypothesis.' ============== 5) FROM: Charles Elkan Abduction in Pierce's meaning will be found in larger dictionaries. It is common to twist an ordinary word into a technical term, so I don't think it's worthwhile to debate whether abduction was the best word to be twisted. Abduction means reasoning from a proposition B to a proposition A such that A -> B. The arrow may be causation or another type of implication. Deduction would be reasoning from A to B. Both deduction and abduction can be yes/no or constructive. If I tell you that A -> B and B, and I ask "A?" then you will do yes/no abduction to answer my question. If I tell you that A -> B and B, and I ask what follows, you will do constructive abduction. (Note that I am using constructive as an ordinary English word here, simply because I don't know a widely-accepted technical term for this distinction.) ============== 6) FROM: Pierre Marquis The term "Abduction" has been introduced by Aristotle in its First Analytics. In this book, Aristotle presents a theory of induction (II, 23) and a theory of abduction (II, 25) in a syllogistic frame. Synonyms are apagogy (Aristotle) and (I think) retroduction (Peirce). ============== 7) FROM: Mike Peirce also used the term "retroduction" for the same things he used "abduction" for. I don't really know his reasons for the term, but retroduction gives the image of something like "deduction in the other direction". I.e., the hypothesis generation part of the hypothetico-deductive method, but not merely generating any hypothesis, but hypotheses of particular kinds. ==============