A Turing Test for Literary Characterization

John H Reif

One of the greatest challenges for a writer is to create novel characters that are alive. There are many artifices to do this, including speaking the voice of the character journaling, and so on. But how does one know that the writer has achieved the ultimate literary goal of a complete characterization that is as alive as you and I, and yet just a literary creation?

A possible answer to this seemingly unsolvable problem might be found in a completely different field, that of computer science. Alan Turing was an early and brilliant computer pioneer and notably the one person most responsible for breaking the German WWII Enigma Cipher Code. He also devised the Turing Test for determining if a computer can think like a human, and so has achieved an “artificial intelligence” that is indistinguishable from a human intelligence. (See Alan Mathison Turing, Computing Machinery and Intelligence in Mind, A Quarterly Review, Edinburgh, 1950. Reprinted as "Can a Machine Think?" in J.R. Newman's The World of Mathematics, pp. 2099-2123.)

The modern version of the Turing Test simply consists of a sequence of emails between a human, which we will call Alice, and another entity we will call Bob who is either

(1) a computer or

(2) a human.

Bob passes the Turing Test if, after an appropriately long sequence of email exchanges, Alice cannot distinguish whether or not Bob is a human. So far no computer can be truly said to have passed the Turing Test and the dream of a computer with artificial intelligence of this sort probably will not be achieved until far in the future.

We can use a modified form of the Turing Test for the literary task of certifying that a writer has achieved a literary characterization that is indistinguishable from a true human. Our procedure for this Test consists of a sequence of emails between a human tester, whom we will again call Alice, and another individual Bob who is in this test always human. In this Test Alice is initially unsure whether

(1) Bob is a writer producing a literary characterization or

(2) the actual individual character.

The writer’s characterization has passed this Test if Alice cannot distinguish between the two, after an appropriately long sequence of email exchanges, even when Bob is actually a writer producing a characterization. If such an Alice is not immediately unavailable to aid in this Test, Bob might use instead use an internet message posting system to recruit an appropriate Alice to converse with and conduct this Test. (But we do not recommend that the author use a dating service to attract an Alice to conduct this Test, since there is a danger that if the Test is passed, Alice might in addition fall in love with the writer’s literary characterization.)

In this era of computer automation of almost any task, it may be a comfort to the writers among us to know that it is very unlikely in the immediate future that computers will be authors of literary works with convincing characterizations. For if a computer produces a characterization that passes our above Test, then it also will have passed a Turing Test and so achieved intelligence indistinguishable from a human intelligence.