Now back in New York after a quick visit to Harvard, having spent the way up writing letters of recommendation and nomination and the way back down writing more letters of recommendation and nomination, when I wasn’t wondering whether or not a train that rattled and shook so much could possibly have been designed by an artificial intelligence, it occurred to me to wonder, in addition, whether or not an artificial intelligence could possibly be programmed to write all these letters of recommendation and nomination. This is also a response to this comment on an earlier post. The comment claimed that a musician “has already written the answer” to one of my questions about artificial music performance. I’m sure that it was not written by a musician because it has two authors, and I’m not sure it’s the answer, but it’s certainly an interesting and thought-provoking answer, and it inspired me, as I said, to wonder whether another person (composed of two or more people) would already have been inclined to write the answer to this question about artificial letter writing. Or whether a machine would write a letter — to me, for example — answering my question.
I put a fair amount of time into writing letters of recommendation (and nomination) because I generally like the people I’m recommending (and nominating), which is very different from “liking” them. A machine could easily find their social media footprints and “like” them but can it like them? And if not, can its letters really substitute for letters written by a human? After all, the department or institution that is considering hiring (or promoting, or rewarding) the candidate should be indifferent to the candidate’s non-objectifiable personal qualities and should only pay attention to facts — like citation statistics, for example. And it turns out that machines are pretty good at reporting facts and even turning them into stories that are enjoyable, or at least “like”-able. A machine willing to look beyond the mere citation indices and to explore the citations themselves will be able to write “the candidate’s work has been called interesting and thought-provoking” and every word of that will be objectively true.
Of course, having stepped this far out on a limb, one has to entertain the possibility that the entity doing the citing is (itself? themself?) a machine, and if you haven’t actually met the candidate but are only familiar with the candidate’s work — which is more likely than not if you are a letter-writing machine — then you have to be willing to imagine that the candidate may well also be a machine. Whether I would have been able to follow this line of thinking to its logical conclusion if the Acela train had not made me seasick, or whether the line of thinking is itself a symptom of my seasickness, is a matter only the reader can judge, and then only if the reader knows what seasickness is like, and whether or not a machine has been invented that can experience seasickness, or whether it would even be ethically permissible to invent such a machine, are not questions I can begin to answer.