Computer Says D’Oh!

There are, broadly speaking, two ways of using the word “funny” in English. The first, which relates to situations that puzzle but do not amuse, is “funny peculiar,” as when we find our car keys in the fridge or questionable charges on our phone bill. This is the sense to which Isaac Asimov alludes in his observation that “The most important words in science are not ‘Eureka’ but ‘that’s funny.’” The second relates to the sense of humour, to amusing situations that make us laugh, often because they have been deliberately engineered by others to do precisely that.

It turns out that computers excel at the first sense of “funny,” both at fostering a puzzling sense of the peculiar in their users and at recognizing user behaviours that are statistically out of the ordinary, but stumble rather badly on the second. The fact that we use “funny” to describe these very different kinds of experience suggests that both share a common grounding in the incongruous, but a formal and/or algorithmic means of converting “funny peculiar” into “funny ha-ha” has proven stubbornly elusive to humour theorists and computationalists alike. The gulf between the two is most easily bridged in the case of the pun – the form of humour on which machines have shown the most success to date – since pun-words themselves serve as a shallow bridge between multiple words and senses. But machines still find it acutely difficult to look beyond the pun, to see intention in the peculiarities of others or to imbue their own creations with comical intent. However, just as one person’s funny peculiar is another’s funny ha-ha, machines can mine the large grey area between the two in ways that make their efforts at humour more than accidental if still less than entirely knowing and deliberate.