Assessing My Personal Gender Bias on Twitter

I am a feminist. I believe that men and women should be given equal opportunities and that - as far as possible - we should have a civic society composed of equal numbers of men and women.

At least... I think that's how I feel.

Most people believe that they are rational and live up to their own ideals. But can we ever really know our own subconscious biases?

The Geena Davis Institute on Gender in Media have released a series of study about gender biases. One of the more interesting findings was that when the male:female ratio in a group is 70:30 - both genders think the group is at parity. This is due, in part, to popular entertainment sticking almost rigidly to these gender ratios.

I honestly believe that I follow a roughly equal number of men and women on Twitter. I certainly see a lot of women in my feed. But that's just a feeling - what do the data say?

I decided to look at two different aspects of my Twitter usage.

  1. How many women do I follow? (i.e. I choose to see their tweets).
  2. What's the ratio of male:female tweets that I see? (i.e. if Alice tweets twice and Bob only tweets once, it's 2:1)

Limitations

The most obvious limitation is that not every tweeter is a person - some are faceless corporations. Some company accounts are definitely gendered - take Cosmo magazine for example - but I think it's easiest to separate them out.

Those tweeters who are people may not choose to reveal their gender - or it may not be immediately obvious. I don't intend asking people directly what their gender identity is.

Finally, there are many other ways I could slice this pie. I could look at race, gender, political leaning, etc - but gender is the most visible and therefore easiest to identify.

I'm basing my assessment on the user's name and their photo.
Male and Female Twitter-fs8
In this example, I've no easy way of knowing whether @Lateral is male or female. Sorry!

I did consider using computer vision recognition and running the names through a stochastic model - but I'm off sick today so decided to manually count each follower. For something like this, the human brain is faster than installing and configuring software!

So, for this investigation, I'm dividing the tweets that I see into Male, Female, Unknown, Corporate.

Followers

As of now, I follow 1,158 Tweeters. At first glance, that's a ridiculous number. There's no way I can have any form of meaningful interaction with that many people. What that doesn't reveal is that many of those accounts are dormant or seasonal. Some, no doubt, I have no need to follow. But since Twitter killed TwitCleaner it's impractical to go through all those old accounts and remove the deadwood.

I manually looked through the 300 accounts I most recently followed. I was fairly confident that it would show want an equal opportunity tweeter I was. You're welcome to look at my Twitter friends directly, if you'd like to verify the data.

300 most recent Twitter friends
Gender Number
Male 130
Corporate 87
Female 80
Unknown 3

Youch. I follow more corporate accounts than women. And the M:F split? 62: 38. Not too far off the 70:30 split.

How depressing! There I was, thinking that I was breaking the mould - when instead I am, basically, a litlle better than average.

Because Twitter doesn't let me retrieve all my followers at once, I had to split my research into batches of 300. After looking at my 600 most recent friends, I was even more depressed.

600 most recent Twitter friends
Gender Number
Male 227
Corporate 179
Female 140
Unknown 3

Exactly the bloody same ratio! True, some of the corporate accounts are likely to be run by women... but I think that's just clutching at straws.

I think the m:f ratio here is interesting because it's accounts I've chosen to follow. For some reason, my brain only finds women about half as worthwhile following as men. I think that's troubling.

After going through all 1,158 accounts I follow (I was off sick today - what else was going to fill my time?) here are the final results.

All Twitter friends
Gender Number
Male 582
Corporate 323
Female 245
Unknown 7

Oh. Em. Gee. Honestly, I didn't fiddle the data - take a look at my friend list. But that's a Male:Female ratio of exactly 70:30.

I've gone from thinking I was a new wave feminist man to realising I'm just part of the patriarchy. Ok, maybe not that dramatic, but I honestly thought that I was following a much more gender balanced crowd than that.

I may not be following lots of women - but perhaps I'm seeing lots of their tweets?

What Do I See

Of course, these women could be tweeting a lot more than the men. And, perhaps, men could be retweeting a lot of women.

Perhaps.

Twitter's tools only let me download the last ~800 Tweets I've seen. Twitter have no interest in letting you travel further back in time than that, which is a shame. Today - a Monday - that's around 4 hours worth of Tweets. Normally I don't read them all, nor do I scroll back extensively through my timeline to catch up on what I've missed.

With more time - and a more cooperative Twitter API - I could take a look through the last month. But it is what it is.

4 Hours of Tweets
Gender Number
Male 345
Corporate 165
Female 136
Unknown 19

No need to grab a pocket calculator. That Male:Female ratio is stuck at a stubborn 72:28.

I've no way of knowing how representative a sample this is. Obviously the ratio will ebb and flow depending on the time of day and what's happening in the world. I could also de-duplicate the results and try to find out how many unique voices I hear.

Thoughts

Firstly, these data are really only representative of me. I've not included anyone else in this investigation.

Secondly, it's really important to take an impartial look at your own behaviours. I would have sworn up and down that I followed a a better ratio than 70:30.

Thirdly, there are many different ways to measure bias. For example, looking at who people retweet - although, even by Twee-Q's measure I'm still a 70:30 guy.

Fourthly, men and women both use Twitter in roughly equal numbers - so I can't blame the platform for my performance.

We can't control gender equality in politics, TV, education, or employment - but we can control the number of distinct voices we choose to expose ourselves to.

Take a moment to look at your social circle - whether it's online or not - and see if you could make some adjustments to increase your daily dose of diversity.

In the meantime, here's a funny video with Geena Davis.


3 Responses to “Assessing My Personal Gender Bias on Twitter”

  1. Cristian Vrabie (@cvrabie) Image of Cristian Vrabie (@cvrabie)

    I think this might be more of a consequence of your job than your bias. We tend to follow people with which we share an interest and we derive many of those from our job. Our industry is undoubtably skewed. While this is not how it should be, it's the reality at this point. As a long term solution to your problem we should encourage more girls to become engineers :)

    Reply
  2. Jinian Image of Jinian

    As you allude to with the retweeting measure, the popularity-contest aspect of Twitter will tend to magnify existing biases and propagate them even to people who think they're doing fine. If someone is looking for accounts to follow, the ones they're likely to see recommended are more likely going to be coming from men. Ratio perception is definitely an issue, but the system enhances the problem.

    Reply

What Do You Reckon?