« why aren't more european women online? | Main | What's Carly Up To? »
October 29, 2003
Us, run through the Gender Genie
Just for kicks, I ran the last few Misbehaving.net posts through the Gender Genie, mentioned earlier on this site, a site that purports to determine the gender of a writer. The Results: "Why aren't more European Women Online? came up as Female Score: 567 / Male Score: 964; "The Accidental Techie" came up as Female Score: 701 / Male Score: 877; "Re-imagining our Lives" came up as Female Score: 692 / Male Score: 939; "The Opt-Out Revolution" came out Female Score: 903 / Male Score: 1203; "Simulating Women" came up as Female Score: 160 / Male Score: 161; "Social Construction of Technology" came up as Female Score: 232/ Male Score: 498.
Most of us came up as predominantly male. This isn't the most accurate algorithm (you can read about how it is supposed to work in this Nature article), but even at 80% accuracy, it's noticeable.
Posted by Caterina Fake at 02:44 PM in General | Permalink
Comments
I had to go back eleven entries in my own weblog for it to identify my writing as feminine. The ten I tried previously were all identified as masculine.
Note that the Gender Genie version at bookblog.net is a simplified version of the one with 80% accuracy. I suspect that the accuracy of the simplified version is lower.
Posted by: katie at Oct 29, 2003 3:21:32 PM
So, the Gender Genie wouldn't know a woman if he/she were in the same bottle with one. Another fun thing to run your posts through is the Wonderful Wankometer which rates the amount "Management Wank" in a page. http://www.cynicalbastards.com/wankometer/
Posted by: Jim Elve at Oct 29, 2003 4:41:01 PM
The article seems to imply a correspondence between non-fiction and masculine writing (in that both are 'informational' instead of 'involved'). That could explain it.
Posted by: Mark at Oct 29, 2003 5:10:37 PM
My own last blog entry - about digital TV -scored Female: 205 and Male: 1202. Is that a record?
Looking down the words classified as male or female (this was for the 'blog' type), my guess is that if you took a diary/journal style blog on Live Journal, you might well get those sorts of 'female' words, and blogs of the LiveJournal type are dominated by women (Live Journal has some interesting stats on its own site).
Posted by: Louise Ferguson at Oct 29, 2003 6:57:54 PM
I have such mixed reactions about the results of this program. On the one hand, I realize the algorithm that calculates it is based on actual samples of writing by men and women. So I can't just simply accuse the machine of being based on linguistic stereotypes, where men's discourse is more "scientific" and women's more "informal," (as mentioned by that article.)
And yet, the results of the program reproduce these stereotypes -- as if declaring that fundamental linguistic differences between men and women really do exist, that stereotypes are based on reality. That women who are writing "scientifically" or "formally" are simply putting on men's clothes a la Margaret Thatcher and strutting around in them.
This bothers me a lot, of course, so I wonder, instead, how the algorithm itself is calculated. If it really is based on samples from men and women's writings, from where are these samples taken? What kinds of people?
Also, it's still ridiculous to me that such a program even exists -- as if trying to find the innate gender behind the author of a text is even possible, and as if it were a pursuit people should actually be spending their time and resources on. I think ultimately that's where the biggest bias of the program is at all -- its very existence.
Posted by: AA at Oct 30, 2003 4:02:25 AM
Wow. Maybe my favourite part of the program is when you tell it that the writer is not male, and it responds "That is one butch chick." (?!)
Posted by: Anne at Oct 30, 2003 9:41:49 AM
I think Mark's right on. If the creators of this program just went out and found "writing by men" and "writing by women" without paying any heed whatever to the *context* of that writing, what the results will reflect is a significantly gender-segregated society.
Which, to me, is not exactly news.
CavLec tests male or female depending on whether I've been writing about HTML or my cats lately. :)
Posted by: Dorothea Salo at Oct 30, 2003 10:07:31 PM
Female: 0
Male: 88
Next was
Female: 288
Male: 800+
Guess they still have some tweaking to do.
Posted by: meg at Oct 31, 2003 5:23:22 PM
Hi!
I saw your website its good. Myself doing Trading business based at new delhi india. I have a great interst in foreign culture/language/women/frendshipImpport/export etc. looking forward for your response. Member French cultural centre, planning to learn french from next jan2004.
Bye
Sudish
Posted by: S. Mehra at Nov 7, 2003 4:59:55 AM
Just because your sex is female, doesn't necessarily mean that your writing is feminine. Nor your gender.
The algorithm can only analyze the nature of the writing, not the nature of your sex.
Posted by: Julie at Nov 7, 2003 5:18:55 AM
