Editor's Column

Gender Tech

John Kilgore

But suppose for the sake of argument that the 80% figure is attainable. What then? For a dizzy moment, the fact seems to have far-reaching social and philosophical implications. If men and women are so reliably and demonstrably different, hardly more able to change their prose habits than their fingerprints, what does this say about gender equity? About free will? Could those blowhards on talk radio be onto something after all?

Then the dizzy moment passes. The ability to infer gender based on very limited data is in fact not too surprising, nor does it imply much of an ability to make more complex and detailed predictions. Give me just two words—the names of Subject A and Subject B—and I can tell you with 95% accuracy the gender of both, and if one is male and one female, I can go on to make high-accuracy judgments about which more likely owns a purse or a baseball bat, which one sometimes wears dresses, and so on. Deciding whether John or Joan would make a better doctor, airline pilot, or terrorist is an entirely different matter.

Still, who can resist a gimmick like this? Using a lazy-man’s shortcut, I cut and pasted the last 723 words of “Curses Not Loud But Deep,” an essay of mine that ran in Chronicles last spring, into the Genie’s purview. The program correctly identified me as Male, by a score of 1188 to 731. The analysis looked to be as simple in its procedures as it is puzzling in its assumptions:

The Genie, it seems, ferrets you out by a simple analysis of your vocabulary, tallying instances of words classed as either “masculine” or “feminine” and weighted from 1 (very slightly predictive) to 52 (strongly predictive). With and if are the most feminine of feminine terms, worth 52 and 47 points respectively, whereas around and what take top macho honors with 42 and 35. At this point you start scratching your head. What are you supposed to say instead of what if what is what you mean? How do you deal with iffy material without feminizing your discourse by using the word? The Gender Genie is sweetly innocent of opinions on such matters. All it knows is that, based on analysis of several kajillion writing samples (including your own, once you have concluded the session by certifying it as male or female) men are x times more likely to say more than women. Why is your problem.

Intrigued and a little miffed, I schlepped in the first two pages of “Hamlet in the Closet,” the talk I gave at our recent colloquium, only to behold a startling result: Female, by a score of 718 to 632. Using the last two pages yielded the same verdict and a very similar score. Ha! I thought: so much for that boast of accuracy.

Or was there, maybe, something wrong with that essay? It ran uneasily in my mind that Chris Hanlon had suggested (jestingly, I thought: but you never knew) that my title contained “a Freudian slip.” Well, I would show him. I went back to Agora 27:3 (March 2002), our second online issue, highlighted 1613 words of his essay there (“The ‘Virtual University,’ Public Education, and Intellectual Property”) and flung it before the pitiless gaze of the Genie.

The score: 2927 to 912: overwhelmingly male. Stunningly male. A macho tour de force. Worried, I betook myself to Marty Scott’s “On Nancy Sinatra” (Agora 28:1, September 2002), abducted 1261 words, and placed them in analysis. To my relief Marty’s lean and sinewy prose, in an article patently given to male preoccupations, tallied as female, by a nose, at 1555 to 1463. I was in good company.

Come to think of it, I had been trying to get in touch with my feminine side in that Hamlet talk. Reaching out to the audience, and all. Sure. And shouldn’t the real goal, here, be to fool the Genie, proving one’s range and cleverness in outflanking the classificatory systems that apply to mere mortals? On a sudden hunch, I turned to “How to Meet a Man,” a story I had written from the point of view of a young woman (available online, if anyone is interested). Sure enough, the first 2808 words scored as Female, 3519 to 3429.

I felt vindicated as a man and an artist.

But further research showed I was far from unique in my ability to fool the Genie; it seems in fact that writing done on the third floor of Coleman Hall has a striking tendency to do just that, flouting the usual patterns of malespeak and femalespeak. I wasn’t too surprised that a chunk of Darkroom, Mary Maddox’s dark action-thriller novel (Agora 27:3, March 2002), should register as solidly male (1459 to 1054, leaving her second only to Hanlon in machismo). But who would have thought that Bonnie Irwin’s warm reminiscence of “Professor Camp” (28:2, December 2002) would likewise register as male, 1248 to 942, or that Robin Murray’s sweetly understated recommendations on curriculum, from the same issue, would come in with a bruisingly butch tally of 1009 Male to 735 Female? Of the eight trials I ran using our own work (three of them mine, I egocentrically confess), the Genie was correct just twice, or 25% of the time—an abysmal record of failure unless you turn it on its head, and note that the program is remarkably good at predicting the gender to which EIU English teachers do not belong, succeeding 75% of the time.

Reflections like that last are why I lost interest before I performed some of the tests I had intended: George Eliot (a woman writing under a male nom de plume), Walt Whitman (gay), Molly Bloom’s soliloquy from Ulysses (male writer doing female voice), and a science fiction short story called “The Women Men Don’t See,” by James Tiptree, Jr., wherein a female author writing under a male pseudonym takes the point of view of a man puzzling over the strange ways of women. There are after all just two answers to the gender question, and the computer’s ability to get it right or wrong in these particular cases would not tell us much that we don't already know. If Molly Bloom scored Female, as I suspect she would, this might indicate that Joyce has a fine ear for at least some feminine speech rhythms, but we knew that already. If she surprised me by scoring Male, this would no doubt confirm that beneath the character’s female persona the author, Joyce, is still male—but we knew that too. The fact that an author uses when or below more or less than other authors turns out not to be all that material to other concerns. The Genie counts what it counts, unarguably right on its own terms, but the essence of gender, what it means or consists of in human terms, refuses to be simplified.

Why then would the program be better at detecting gender than human subjects confronted with the same writing samples? Simple: language on the page lacks the usual cues—tone, pitch and emphasis—with which we ordinarily “predict” one another’s gender in conversation, and do so far more swiftly and accurately than the Genie. We have never had to evolve roundabout word-counting aptitudes because we are already such natural virtuosos, able to determine gender on the basis of a single spoken word. That we are so uncannily good at this proves (if anyone should doubt it) how crucial gender recognition is in the round of social, er, intercourse. Somewhat paradoxically, however, this crucial judgment once made is hypothetical and abstract, suspended for the time being, an x whose particular value remains to be determined in particular contexts. However determinate they may prove to be in themselves, the gendered aspects of language will be, like gender itself, infinitely variable in the way they relate to differing circumstances. Whether you want to comfort a child, declare war, or tell someone she has cancer, you are perfectly capable of doing so using keywords from either list.

***

This winter’s Agora is remarkably full, and not for the faint of brain. Two reprints from prestigious professional publications give us a way of welcoming two “new” faculty members to the Department, though the work itself is astonishingly seasoned and sophisticated. Chris Kuipers, who earned his PhD at UC Irvine, comes to us by way of Cal Poly-San Luis Obispo, Chapman University, and Cal State Fullerton, trailing clouds of theory and classical scholarship. His “The Anthology/Corpus Dynamic: A Field Theory of the Canon,” reprinted by permission of College Literature, shows an alarming knowledge of matters ranging from Lewinian psychology to glacial mechanics to the post-Classical Greek epigram as it pursues one of the profession’s eternal questions: why those textbooks keep getting fatter. Jad Smith holds a doctorate from Carnegie Mellon, with specialties in literary theory, the long 18th century in Britain, and Native American Literature, and has grabbed his oar in Coleman after teaching at various other universities throughout the mid-nineties. His “Völkisch Organicism and the Use of Primitivism in Lawrence’s The Plumed Serpent,” reprinted from the D.H. Lawrence Review, does a wonderfully scrupulous job of sorting and weighing the evidence, textual and otherwise, of Lawrence’s flirtation with proto-fascist ideas early in the twentieth century. The piece strikes me as the sort of critical work that is especially valuable, these days, when our continual challenge is to honor the political connections of literature while escaping the pitfall of political reductivisim and rigid "correctness."

Those of us who have been enjoying Graham Lewis’s poetry for some time will be happy to celebrate with him the forthcoming publication of his first collection, Forever Came Today. Generous excerpts from the book are given here by permission of Water Press and Media; but don’t forget to order the hard copy, using the link included in the headnote. Another volume of poetry, Language of Mules, will be harder to find, having become a collector’s item, but for that very reason might make an especially thoughtful Christmas present from you to yourself. I believe a few last copies are still available from John Guzlowski. Many others have found a home in Poland, where the little book has been a hit, earning a number of extremely favorable reviews, including one by Nobel Prize Winner Czeslaw Milosz, who declares that “the poems are astonishing” and "Guzlowski has an amazing ability to apprehend reality." We knew that already, but are delighted to have one standard from the collection, “Cattle Train to Magdeburg” to share here along with a later, delicately ironic mirror poem, “My Mother Reads ‘Cattle Train to Magdeburg.’”

Denise Clark’s lovely little poem “The Rising” is taken from a collection in the making, Felonious Desires, and is surely the harbinger of very good things to come. Evelyn Scott, finally, gives us the article I suspect will be turned to soonest and most. Funny, outrageous, and wonderfully readable, All Stripped Off is also earnestly argued and well researched. Challenging the meaning and proper range of the term “feminism,” the essay will not convince all readers, but what good essay ever does? Read this for some serious limbering-up of preconceptions you might not have tested recently.

All in all, a real feast for the gray weeks that lie ahead. Merry Christmas, everyone.

JDK