Jun 142007
 

Tim poin­ted me at the art­icle on devchix about bar­ri­ers women face in tech com­munit­ies; it’s cer­tainly sparked a lot of interest and reac­tions out there.

My reac­tion was two-fold: one was to think “is that how most women are?” To under­stand that, you have to remem­ber that I’ve spent my entire life past the age of 17 in groups that were pre­dom­in­antly male, first in phys­ics, and then in com­puters. I’ve often dis­covered that I look at things one way and some woman I talk to about it will see it quite dif­fer­ently. So until I read some reac­tions to this art­icle, I thought maybe it explained some­thing about the way women-only groups work that I didn’t know about, since I’m not actu­ally in any (even the book­club I’m in is has men and wouldn’t feel right to me if it didn’t; the only groups I’ve been in recently that were only women were knit­ting classes and those only last a short time).

The second reac­tion was that she makes some fairly strong state­ments that are test­able (there’s that phys­ics back­ground com­ing out): 

I have exper­i­mented with this myself using a male pseud­onym to post art­icles, and being told that the art­icles are inform­at­ive, use­ful, great. Six months later I repub­lish the exact same art­icle, using a dif­fer­ent title and a female pseud­onym, and sud­denly the art­icle is hor­rible, tech­nic­ally incor­rect, use­less. It’s a fas­cin­at­ing study.

It’s actu­ally a hard thing to test. Many people pub­lish art­icles on their blogs, so they can’t sud­denly change their name and gender for that; where else do people pub­lish these days? How much of pub­lish­ing inform­a­tion is about repu­ta­tion, where the read­ers say the per­son has been right about other things in the past, prob­ably is about this as well? That also doesn’t enable switch­ing iden­tit­ies read­ily. I would like to see some actual data and test­ing of the pro­pos­i­tion, and not just from one per­son.

Shel­ley wrote up her reac­tion; read both the art­icles as well as some of the com­ments and links for a fuller view. [At first I wrote “bal­anced view”, but until we know more about the issues, who’s to say where the centre (and there­fore the bal­ance point) is?]

  8 Responses to “Tech Women”

  1. I can only repeat what I’ve said else­where: the world is a bet­ter place when women treat men less like ter­rit­ory and men treat women less like appli­ances.

    Remove all sex­ism from the work place and the res­ults are far far worse than if every­one prac­tictes tol­er­ance, com­pas­sion and self-restraint. This isn’t pan­glos­sian; it is prac­tice and it is age-related.

    As to the lack of women, why is that the SGML com­munity had lots of them? My dis­tinct memor­ies on enter­ing the com­puter sci­ence work­force in 1980 was that it differed pre­cisely because of the pres­ence of women. Why is it the web online groups have struggled with this for so long. I noted this in VRML lists in the mid nineties and pos­sibly, it has some­thing to do with the demo­graph­ics (age and edu­ca­tion) of the mem­bers where edu­ca­tion is not a good pre­dictor of polite beha­vior.

    Shel­ley said it: respect for self and oth­ers where respect is not self-obsession OR find­ing all for­tune in other’s eyes (see Shakespeare http://www.albionmich.com/inspiration/whenindisgrace.html).

  2. Thank you for say­ing “pre­dom­in­antly male” instead of the more accus­at­ory “dom­in­ated by men” that many use when dis­cuss­ing this issue.

  3. You write “It’s actu­ally a hard thing to test.” But actu­ally I think it might not be that hard to test. How about some kind of ran­dom­ized test­ing? Start two blogs, one with a female iden­tity and one with a male iden­tity. Write a stream of art­icles. After you fin­ish any art­icle (and sol­emnly swear not to revise it fur­ther, not even the tini­est bit), flip a coin and let the coin flip determ­ine whether the art­icle ends up as male or female.

    The trick­i­est part, it seems to me, is fig­ur­ing out how to score people’s reac­tions. If you end up think­ing you see sys­tem­atic dif­fer­ences but you can’t put your fin­ger on how sig­ni­fic­ant they are, it would be quite nice to be able to com­pare your “13.8 pos­it­ive” rat­ing on one blog to “18.4 pos­it­ive” on the other, and I don’t know how to do that. But if the dif­fer­ence in reac­tion is con­sist­ently huge (as the quoted “inform­at­ive, use­ful, great” vs. “hor­rible” pas­sage leads me to expect), there would be no need to sift through stat­ist­ical prop­er­ties of numer­ical data, you could just print out the com­ments on each blog side by side and let the dif­fer­ences in tone speak for them­selves.

    That would not be a per­fect test, and I can think of vari­ous prob­lems. For example, per­haps one con­tro­ver­sial post could end up leav­ing part of one blog’s audi­ence per­sist­ently angry, so that even if you tested with 58 blog posts it wouldn’t be any­where near 58 inde­pend­ent samples. But as I under­stand it, social sci­ent­ists have developed a big bag of tricks for this kind of thing, so you might be able to use those tricks to improve my simple-minded idea and dodge prob­lems like that.

    One trick that might help would be to write a private ances­tral blog post, then edit it into a pair of fundamentally-similar pub­lic blog posts: keep­ing very-similar con­tent, but adding as many super­fi­cial dif­fer­ences as you can think of. Then flip a coin to determ­ine which superficially-different vari­ant ends up on which blog.

    If any­one read­ing this decides to do some­thing like this, by the way, I encour­age you to begin by vow­ing to release (on the web any­way, I have no idea about the prac­tic­al­it­ies of journ­als in the field) the res­ults no mat­ter what you find, and even if you can­cel the pro­ject part­way through (in which case you just release res­ults so far). One dis­turb­ing worry about pub­lished research is what weird dis­tor­tions could get intro­duced by people’s decisions “nah, not worth pub­lish­ing.” Since those decisions hap­pen in private, it’s hard to estim­ate how they tend to be dis­trib­uted, but some­times people guess that they might have some­thing to do with the per­sist­ence of some kinds of sci­entific errors.

  4. That would work — you’d prob­ably have to set up both blogs as anonym­ous to get around the fact that curi­ous people would hunt in search engines to find out more about the authors. Then the biggest prob­lem becomes mak­ing people aware of the two blogs by fig­ur­ing out how to link to them appro­pri­ately and “equally”. Prob­ably the easi­est way to do it would be to have a few people work as a group to write the art­icles, flip the coin as to which iden­tity pub­lishes it, then link to art­icles occa­sion­ally from their own blogs. It’s cer­tainly an inter­est­ing idea, and I haven’t heard of any­one try­ing it.

  5. You’re right, I don’t have any very good idea how to make sep­ar­ate blogs be exactly equally noticed, though your ideas to make the approx­im­ately equally noticed could well be good enough. On the other hand, I do have an idea for dodging the prob­lem: instead of writ­ing two blogs, write a stream of com­ments on someone else’s busy blog, and ran­dom­ize the sex of the com­menter.

  6. Avoid­ing selec­tion prob­lems (not pre­dis­pos­ing it to a given set of responses or the self-fulfilling proph­ecy) will be bet­ter with mul­tiple con­trib­ut­ors, but not per­fect. The top­ics have to be gender-neutral to isol­ate it bet­ter from gender-hot but­tons.

    You need a very pre­cise defin­i­tion of what you intend to meas­ure first. I’m not sure the art­icle that promp­ted this really has that yet. So if you aren’t test­ing the pres­ence of spe­cific­ally male reac­tions, then you are test­ing the ? for art­icles where those have gender iden­tity for the author.

    And how do you know who is which given pseud­onym­ous and anonym­ous response?

  7. Len, remem­ber­ing that both women and men can be sex­ist, I would per­son­ally be more inter­ested in know­ing the total reac­tion, and not sep­ar­at­ing out the male or female reac­tions. You’re right, to design an exper­i­ment that would reli­ably sep­ar­ate out male and female reac­tions would be more dif­fi­cult.

    Wil­liam, I guess the com­ment­ing solu­tion could work, although tech­nical com­ments are rare. And given the ori­ginal claim was that *tech­nical* art­icles were treated dif­fer­ently based on the sup­posed sex of the author, this test should stick to tech­nical and not delve into polit­ical or social art­icles or com­ments. That would be test­ing some­thing dif­fer­ent.

  8. True. If the meas­ure­ment is total com­ments by type based on gender of the author, then sort­ing them is need­less as long as the con­clu­sions drawn from the data are sim­il­arly lim­ited.

    I’d be inter­ested in sur­veys of high school seni­ors and col­lege fresh­man over career choices and the motiv­a­tions. Meg­gin­son men­tioned the uptick in entrants in the early eighties. I came back to the field in 1980 hav­ing worked a bit with com­puters in the very early 70s in col­lege. I found it bor­ing and went off to pur­sue music. Later when $$$ became a high pri­or­ity after col­lege, comp sci was the easy in to the engin­eer­ing worlds loc­ally and tech­nical writ­ing was a *build from spare parts* endeavor by the engin­eers. At the same time, comp-sci was con­sidered elite and highly paid. CompSci really only became inter­est­ing to me when I began to think of apply­ing it to prob­lems that inter­ested me unlike a piano which was always inter­est­ing in its own terms. In other words, if the key to engage­ment is inter­ac­tion, per­haps the inter­ac­tions inter­est­ing the cur­rent can­did­ates are not per­ceived as being enhanced by com­puters. Again, a value of val­ues issue.

    I do won­der if one of the drivers for the slopes in those curves for totals across cul­tures for both genders is status. My reac­tion to the web explo­sion of the 90s was the web became to that gen­er­a­tion what rock n roll was to some of my own.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

(required)

(required)