In my daily life, I am surrounded by people making astounding claims about computational power. Many of these claims are true. Insurance underwriters can increasingly predict our mortality based on our habits and behaviors and the more data they have, the more accurate they can be. Psychologists at Lancaster University have set computers to the task of figuring out whether Edgar Allen Poe killed himself after a descent into severe depression.
The researchers conclude, after having the computer compare Poe’s late-life writing through a database containing words, phrases and images that typically connote depression and the suicidal impulse, that Poe was depressed but did not take a direct hand in his death.
It’s a worthy project, but I’m skeptical for a couple of reasons. First, the database is necessarily bereft of writing samples from people who don’t write while suffering depression and from people who don’t write anything in advance of suicide. Second, I wonder if you can compare the writing of somebody like Poe, who is practiced at writing in character, to a database of anything, even if you’re using his letters and journal entries.
Not saying the study has come to a correct or incorrect solution, just that computers can’t know everything.