From abduction to ‘bad science’

8 08 2014

Starting  on my “back-blog” at last.  A few weeks ago we tackled ‘abduction’, as part of our quest to understand how induction and deduction are used during research. When our tame philosopher last year mentioned that abduction was a third way to draw conclusions, we were initially horrified, but by now we had got over that enough to read the Stanford Encyclopedia on it.

Abduction is sometimes called ‘inference to the best explanation’: it draws conclusions based on what seems to be the best explanation for an observed pattern of events.  Abduction differs from induction in that it draws conclusions about what is likely based on proposed causal mechanisms rather than statistical frequency.  However it resembles induction in that what is inferred is not necessarily true, even if the premises are correct (you may not have identified the best possible explanation correctly, for instance, or the best-seeming explanation may not actually be the right one because strange things can happen). Thus while in deduction, true premises guarantee true conclusions (as in “All As are Bs; Z is an A; therefore Z is a B”), induction and abduction can only be used to argue what is likely to be true.

What are its other weaknesses? Abduction presupposes that you can identify all possible explanations for an observed pattern, and  then that you are well equipped to pick out the best one.  Obviously this is a much more subjective process than one might like (for example, apparently people tend to greatly over-rate the probability of simple explanations compared to more complicated ones: yet more evidence of the seductive power of simplicity). But despite this, it can be … um, not valid, because that only applies to deduction… but ?? reliable???/??sound??/other complimentary words, and so everyone uses it, both informally and in science, all the time.

Although this was a fun discussion, this, our third or fourth time use of the SEP, confirmed for me that reading articles aimed at philosophers really is too hard. There was a section on Bayesian Confirmatory Theory that we all had to skip (I was hoping Bayes had done a TED talk or something, but nope: he died in 1761), many other sections that didn’t really sink in even after multiple readings, and the liberal use of words and phrases like ‘qua’, ‘to wit’, ‘explicating’, ‘inter alia’, and others is just distracting. So, we are having a break for the rest of the summer, and then in the fall, shifting to pastures new: ‘Bad Science‘ by David Goldacre.