Simplicity and parsimony

19 06 2014

Three or four weeks ago we decided to tackle “parsimony” for our weekly logic discussion. Why does everyone hold it up as a golden guiding principle, and what exactly is it anyway? For example, if I say “The Universe and all the life forms in it were created by God” is that beautifully parsimonious because it  invokes just one explanation, or wildly non-parsimonious because the proposed explanation centres on a complicated, uncertain and probably non-existent construct?

It was my turn to choose the readings and I overdid it, choosing too many things blindly rather than reading first and pulling out the best:

It was interesting to read about Morgan’s Canon (BTW how can I get something I think turned into a ‘canon’? If I decree something Mason’s Canon, think it’ll get adopted?) — in part to read again the reasonable view that where a phenomenon can be accounted for via simple or complex psychological processes, then the simple ones should be regarded as the default. However, it was also good to be reminded of the problems of calling animals or psychological processes “higher” and “lower” (his original, terrible terms); to learn that Morgan actually believed that any form of learning was evidence of consciousness (!); and to clarify that of course Morgan’s Canon does not mean one should never test hypotheses about cooler, complex psychological processes just because boring ones could suffice as explanations. (Thus the fact that allegedly empathic rats‘ behaviour could be explained more simply does not mean that they aren’t empathic — it just means that the ’empathy hypothesis’ has not been tested properly yet).

The Stanford Encyclopedia article was most challenging, but in some ways the most interesting, and so we tackled it again today. I probably still need to read it a few more times to really understand everything in it (it’s written for philosophers, and since we have to look up what ‘epistemology’ and ‘ontology’ mean every single time we encounter these words [not to mention not getting the knowing side references to ‘grue and bleen‘], it’s hard work). The usable take home messages for me were that there are different forms of simplicity; that these different forms often trade off against each other (an elegant explanation may not be parsimonious because it invokes new or complex phenomena; while a parsimonious explanation may be inelegant because it involves lots of clunky steps); that simplicity can come at the expense of accuracy or richness of information (think of a simple curve being driven through a mess of datapoints that it only partially explains); and that the very reasons why simple explanations are held up as best are actually rather cloudy… they may even be aesthetic as much as anything.