There is a site called: “I Write Like”. The titular program invites the visitor to submit a sample of writing for analysis. After analysing it, the program works out, by using a formula called “Bayes’ Theorem”, which famous author the visitor’s writing style most resembles. Try it once or twice and it becomes addictive. I was told that “I Write Like…” — in no particular order — James Joyce, Vladimir Nabokov, Margaret Atwood, Mary Shelley, Ursula Le Guin, H. P. Lovecraft, J. K. Rowling, and Edgar Alan Poe. Now I couldn’t very well object to this verdict. I’d be happy to be compared to any of them. The trouble was that some of them were so different as to be mutually exclusive. Surely, H. P. Lovecraft is nothing like James Joyce or Margaret Atwood? No, not by any stretch of the imagination.
I know a thing or two about Bayes’ Theorem. It is very powerful. But it is not so much infallible as open to abuse. And so I devised a test for the application in this program. What I did was break down a piece of writing into three sections; A, B, and C, and try each in turn. A, B, and C, each brought the claim that “I Write Like…” Vladimir Nabokov. So far, so flattering. But here’s the thing, when I tried A + B or A + C or A+ B + C, I drew the same result: “I Write Like…”Margaret Mitchell, the author of “Gone With The Wind”! Now three Nabokovs does not equal one Margaret Mitchell. That is nonsense. You can’t add ‘three goats’ to ‘five cows’. When the Bible said: “Be fruitful and multiply” it had another activity in mind. After several more attempts, I realised that the program worked in a more or less meaningful way with extracts of about a hundred words or so. Anything longer, say two thousand words, and it defaults to Margaret Mitchell or sometimes Ian Fleming. Obviously, Ian Fleming is nothing like Margaret Mitchell, and the same can be said for any one of them. Margaret Atwood gave it a try, and on one attempt, she was told she writes like Dan Brown. I’ve had that, as well, and I couldn’t help being affronted, but I shouldn’t be. The writer of this program is trying to make Bayes Theorem do something it was never intended to do. That’s the trouble with statistics. It’s a powerful tool, in the right hands. In the wrong hands, it misleads and makes fools of us all.