Genre Gender.

This fascinating website uses an interesting mathematical algorithm to predict the likely gender of the writer.

Now, I’m not sure how I feel about my writing being “gendered” (presumably if an algorithm can predict such a thing with a fair degree of consistency, our own subconscious reading of the text might infer a gender as well, though perhaps with less accuracy). I think ideally, I’d like to be almost perfectly balanced between “masculine” and “feminine,” especially since I have always striven to balance both energies in my own personality.

The article accompanying the website breaks down the development of the algorithm and explains a bit how it works:

When the Israeli stylometricans, as they call themselves, study a text, they scrub it clean of everything that’s ”topic specific” — in other words, no ”gown,” no ”princess,” no ”keg,” no ”bullet-resistant.” This is how sophisticated language analysts work these days. They ignore the obvious stuff and concentrate instead on the seemingly unobtrusive little tics that the writer and reader barely notice.

What happens when you strip writing of “topic specific” words? You’re left with the basic grammatical structure of a text, how sentences are shaped and slung together with conjunctions, prepositions, pronouns and such. Before even reading the article, I’d begun to notice the algorithmic pattern that it describes:

Similarly, what the gender-identifying algorithm picks up on is that women are apparently far more likely than men to use personal pronouns — ”I,” ”you” and ‘’she” especially. Men, on the other hand, prefer so-called determiners — ”a,” ”the,” ”that,” ”these” — along with numbers and quantifiers like ”more” and ‘’some.” What this suggests, according to Moshe Koppel, an author of the Israeli project, is that women are more comfortable talking or thinking about people and relationships, while men prefer to contemplate things.

I’m not sure if I buy all that. Anyone who knows me knows I am absolutely obsessed with “contemplating things” (why the hell do I write poetry if not because it affords me the opportunity to focus an intense gaze on objects and events that might otherwise be overlooked?)… and yet it is the interconnection, the relationships, between things (and, by extension, between people) that makes things in themselves fascinating. Few people can write engagingly about pudding or void or anything homogeneous—complexity, i.e. a thing’s relationship to itself, as well as contrast make all “things” and people interesting and worth discussing. If anything, this algorithm might help lay bare the sexual stereotypes so embedded in our culture and upbringing, but as usual, I doubt very much if it reveals any biological necessity for such stark differences.

Of course, I always have to be the stubborn one to shrug off the rules. I couldn’t help but play around with this “Gender Genie” for a little while, plugging in various examples of my work, seeing what results it returned. Eventually, I began to notice a very interesting pattern

As might be expected (by anyone except the writers of the aforementioned article, apparently), examples of my more formal essays and book reviews tended to be overwhelmingly “masculine” in their use of language, while less formal blog entries returned more balanced results, often just on the cusp, though still tending towards the work of a “male” writer. These results make sense to me, though—not only has my nonfiction writing always been very analytical and structured (something which the algorithm only picks up on ostensibly in terms of conjunctions), but whenever I am writing “about something” like a theory or a text, isn’t it inevitable that my language is bound to be more abstracted from “people and relationships”?

Next, I plugged in some examples of my poetry, and the results at first surprised me. For some reason, I expected my poems to be more “feminine,” because poetry has always provided me with a space inside which I could allow myself vulnerability and uncertainty, in which I could freely explore “people and relationships” in creative, intuitive ways. But instead, my poetry too came back as heavily masculine (especially, it seemed, when I was writing from a female’s perspective, either my own or a fictional character’s; however, in poems written from the perspective of “the Orphan,” a male character within a particular collection, the writing leaned more towards the feminine… in some ways, this pleased me, as I had been going for an intentional “role reversal” of stereotypical masculinity in this particular group of poems—it would appear I succeeded). On further consideration, I can understand why poetry may be more “masculine.” I have always been taught that poetry should be made up of intense, concise language, and eschew “weak” or wishy-washy words (especially unnecessary conjunctions and vague pronouns, which also happen to make up a large percentage of “feminine key words” the algorithm screens for). Because my poetry has always sought metaphors for “people and relationships” in the world of “things,” it may appear to a “topic blind” analysis to be masculine in nature.

Last but not least, I plugged in excerpts from my fiction and short stories. Lo and behold–finally, I am writing like a girl! (Should I be rejoicing?) My fiction scores for femininity were almost disconcertingly imbalanced (though perhaps this is partly due to the fact that my major work of fiction right now has almost an entirely female cast of characters, quite intentionally). But then, what is fiction, but the perfect genre for straight-forward explorations of “people and relationships”? Breaking down my long story into segments, often times when my main character, Kim, appeared alone, working at her art or contemplating recent events, the language veered towards the “masculine” again, while scenes involving multiple characters interacting with one another invariably registered as “feminine,” even when all the characters involved were male.

So what does this mean? Seems to me that genre, much more than someone’s sex, is likely to influence the apparent “gender” of a text. (My god, that’s almost poetic!) Of course, the “Gender Genie” asks you to select what genre your work appears in (supposedly so that it can make the appropriate adjustments), but experimenting with the same text plugged in under varying options often returned very similar results. So perhaps the algorithm isn’t perfect just yet. Or perhaps it applies best to the kind of mass-produced magazine articles and serialized novels being written today, while more serious or creative or “transgenre” works slip through mathematical blind-spots.

What may be more interesting is to compare the algorithmic results with our ability to predict through “guess” or intuition the sex of the writer, and to explore how readers experience these differences, especially as they manifest as a “preference” for one kind of gendered writing or another. Do readers, male and female, appreciate “masculine” nonfiction but expect more “feminine” language in their fiction? Do males generally prefer masculine language and females feminine language regardless of the genre? Is there an appeal, to each sex, for writing with traits from the opposite gender? Many, many questions…. But o no, I’m afraid I’ve started “contemplating things” again! No wonder I can’t land a man.

Postscript: In case you were curious, this blog post is rather “masculine,” with a score of 1903 (male) to 1571 (female).

~ by Ali on November 23, 2008.

6 Responses to “Genre Gender.”

  1. On the other hand, the NYT article linked to above, and written by a man, is twice as masculine as it is feminine, so perhaps my writing is more balanced than I realized…

    I’d be interested to see other people experiment with their writing, plug in some excerpts and see what they come up with. Feel free to share your results here, if you like…

  2. I may have choked it to death trying to feed it a whole novel. The smaller chunks of my w-i-p that I put through came out 5891 (female) to 5161 (male); 4208 (female) to 3699 (male); and 6506 (female) to 6745 (male). That latter score was for a long section that was largely dialogue, whereas the first two had more narration. Given that all the main characters are female (not to mention other issues), I am pleased. Much else that I submitted (fiction and non-fiction) came through male, apart from a piece I wrote about the death of my father which was 1263 (female) to 934 (male).

    I am still puzzled by the basic assumptions that underlie this. Presumably the original algorithim was based on texts for which the sex of the author was known. But on what basis were these chosen? Were they taken from a variety of cultures and historical periods? Did they take into account any gender bias in the publishing industry?

    There was a discussion recently on this (I’m afraid I forget where) that seemed to suggest that only men write serious, philosophical texts about ‘important’ subjects. It was so loaded with biased assumptions, I didn’t follow it through. But thinking about it now, I do wonder if that assumption was behind the decision to publish. That is, this book is by a man, it must be serious, important (politics, economics, history, philosophy), well written etc. This, of course, is to deny all the important works by such luminaries as Rachel Carson, Mary Midgley, Dora Russell, Joanna Russ, Virginia Woolf, any one of the women mention in Dale Spender’s book, etc etc

    I also wonder if, as you suggest, different styles, approaches, and different subject matter (no matter how much ‘topic specific’ material is stripped) affect the shape of the piece and the basic structure of the language. We tend to be taught to write academic essays in a particular way which will have developed in a largely masculine education system. There is a certain sense to it, but it is interesting to speculate how eucation and our approach to non-fiction would be different had our society been more equal (or dominated by women rather than men). Indeed, is it dominance rather than equality that skews how we write and how we interpret (analytically or synthetically) what others write.

    OK. I think I’m beginning to wander in aimless circles, here. Too many hours trying to get the other damned computer to work. Ho hum.

  3. I definitely agree about the many many problems this algorithm brings up. I found the link posted on another (Druid) blog, by a teacher who uses it as a fun classroom exercise for his students (he claims that it’s almost always accurate, but that’s obviously not the case with my own work, nor with yours). I think it goes without saying that subject matter shapes language and form (the old “medium is message” theory again). It was certainly a revelation for me that my writing in different genres could be so distinct–obviously I knew my style varied depending on what I was writing, but seeing it as quantifiable data was something else. I also agree that there may be bias in the publishing industry itself (I recently read an interview with a publisher who said it’s “so much easier to sell an author than to sell a book”–this was in a conversation about a recently best-selling woman writer who “writes novels about animals” and has a “gentle, warm personality”–sounds like a very gendered view of the author if you ask me).

    I’ve always thought the argument for gender-as-biology was incredibly shaky, just a kind o pseudo-scientific mask over basic sexism, really. I’ve often been told that I’m less “feminine” than the average female, but then does this mean that I’ve somehow been “trained” against my own biology (unlikely, especially considering the pressure to conform to gender stereotypes), or merely that I grew up in a household (and attended daycares and schools during my formative years) that were largely free of such enforced gender roles, intentionally guarding against them. I certainly don’t feel as though I’m less myself, or less of a female… which to me suggests its the gender roles themselves that are the false reality, the false duality.

    But now I’m rambling. In any case, it definitely gives me some food for thought… Makes me want to experiment with my own writing, for instance trying to write a short story with mostly “masculine” language. What would such a story be like? Could I write an essay in “feminine” language without sacrificing meaning? Even if the gendered terms are inappropriate, certainly it suggests some legitimate differences in writing styles, and a way of “measuring” those differences…. Interesting, anyway…

  4. Also, I’d really be interested in plugging in some work by authors in the GLBT community, as well as some of the other women writers you mentioned. Also, perhaps, excerpts from personal journals of writers in all genres (if available), as a way of seeing how the more “free associative” (and presumably more natural) writing compares to works that are intentionally revised and sculpted for publication. All sorts of questions…. Why isn’t somebody studying this, like, NOW?!

  5. Indeed, much to ponder. And thanks for sharing.

  6. Ali,
    I usually don’t respond to blog discussions of my work, but your comments are very thoughtful, so I’d like to address one point (also raised by Grum). The models were trained on works of fiction (for which author gender was known) and are intended to be used to classify fiction writing only. (We have other models for non-fiction and blogs, but the Gender Genie people used the ones for fiction.) In fact, when you put non-fiction into these models, the answer is almost always “masculine” due to genre constraints, regardless of the author.

    Best,
    Moshe

Leave a Reply