Jeni's Musings

I’ve been thinking about the low proportion of women in computing since reading Tim Bray’s post about the lack of women at RailsConf:

Geeks, you know, they’re admittedly obsessive about computers, but once you get past that they’re on average a pretty eclectic, amusing, and warm-hearted bunch. And in recent years I haven’t met a single one who wasn’t upset about the missing gender. If a booming female Voice From On High spoke out, saying, “Do this and we’ll rejoin your profession”, well I bet a lot of us would do whatever it was. But failing that, in the meantime the problem isn’t getting better.

Here are some graphs I managed to find. First, bachelors computer science:

and then masters:

I don’t think that looking at the intake of computer science courses tells the whole story — lots of people get into computing from other disciplines (and I think this is particularly true for markup technologists) — but it’s certainly a significant piece of evidence that the proportion of women in computing is getting lower.

Let’s focus at the cultural causes of this. The big problem at the cultural level is that there’s a vicious cycle at play. As the proportion of women in computing is low, it is perceived as a male discipline, which means that women are less likely to enter it, decreasing the proportion of women still further.

So what can we do as individuals? Well, we can attack at three points in this cycle.

First, we can shift the perception of the proportion of women in computing. We can talk up the proportion of women in computing by including the non-programming aspects of development, such as project management, training, technical writing and user interface design, where I bet (but don’t have statistics to back it up) the proportion is higher. We can boost the profiles of the women who are in computing simply by taking small steps like reading and referencing their blogs. We can make it easier for women to attend conferences by providing or subsidising childcare (heh). We can use the pronoun ‘she’ on occasion when talking about programmers.

Second, we can emphasise the female-friendly aspects of computing. For example, the nature of development means that flexible working practices are much easier to adopt than in some other professions. It’s easy for people in computing to work at home, perfectly possible to work part time and to work on small projects and so on. Also, the speed at which technology moves means that taking time out for children doesn’t necessarily put you at a disadvantage: everyone has to re-train constantly anyway!

Another tack might be to shift the focus onto those aspects of computing that women feel better at, such as communication and teamwork. In other words, make computing about more than hammering away on your own to create the fastest, most succinct code the world has ever seen. Stress working with clients, producing documented and understandable code and sharing knowledge with others. (In fact, several of the practices of Extreme Programming emphasise precisely these aspects of development.)

Third, we can try to reduce the impediments women feel entering a ‘male’ discipline. There are two factors that I can think of here: low self-efficacy and feeling the odd one out.

Self-efficacy is your belief in your ability to do something. Women are more likely to be feminine than men (duh-huh), and the more feminine you are, the less self-efficacy you’re likely to have. What’s more, you have lower self-efficacy for things that don’t fit with your gender role: for example, predominantly masculine people may feel out of their depth when they have to do some knitting, which is culturally associated with women. This is a double-whammy for feminine people (who are predominantly women) in computing: we don’t believe in ourselves anyway, and particularly not when we do something that has a cultural association with men.

Low self-efficacy makes it hard to experiment, because you’re worried that something will go wrong. I think that’s a particular problem in programming, where you have to be prepared to try different approaches to work around bugs or to get better performance. It’s hard to get out of a low self-efficacy state. To do so, you have to achieve something on your own, and that thing can’t be something you think is easy — if you don’t achieve something you set out to do, or if achieve something but get help doing it or it’s something easy, your self-efficacy goes down.

With that in mind, look at how Ruby on Rails is marketed. A big play is made of how easy it is. But if a language or framework is easy then people with low self-efficacy can’t win: if they manage to do something with it then they haven’t really achieved very much because anyone can do it; if they don’t manage to do something with it then they’re complete idiots. I’m not saying that we should advertise languages or frameworks as being hard, because obviously that can put people off as well, but a recognition of the barriers that people might face may, in a strange way, make them more approachable.

This is also an issue for trainers: we need to be able to boost the self-efficacy of the people we train (particularly women) by setting them challenging (but achievable) tasks and not giving them too much help to achieve them.

Finally, we can reduce the feeling of being the odd one out. There’s obviously the option of helping women network with each other, but speaking personally, I don’t feel the odd one out when I’m the only woman in the room: I feel the odd one out when I’m not treated like the other men, namely as an individual with a large set of geeky interests that very probably overlap with yours.

I can see how this is from the other side. On my ‘mothering’ days, it’s very rare for me to encounter a father, but there are odd ones who come to the playgroup or library. I find it very hard to strike up conversation with them. I think it’s partly because of a fear that any friendliness will be misinterpreted as an advance, partly because I worry that we won’t have anything in common, and partly because there are so many women around who are far easier to talk to. It must be incredibly isolating for the homedads.

So I know how hard it is to make someone of the opposite sex feel included but, y’know, you gotta try.

Looking back at the data, it’s interesting that the proportion of female computer science undergraduates has been this low before: the proportion in 2004 was about 25%, which is the same as in the late seventies. The peak of 37% was in 1984. I wonder what happened in the seventies and early eighties that might have made the field more attractive to women, and what changes initiated its decline?