Twitter Bias: We Listen When Men Talk Tech and Women Talk Diversity

I write quite a bit about women in technology. I’m also an enterprise startup CEO, a linguistics PhD and a fan of astronomy. I regularly tweet about all these things. One day a few months back, I had a hunch: My tweets about women in tech seemed to get significantly more engagement than the others.

Going back and looking through my history, sure enough: Over a three-month period, my tweets about women in tech averaged five times as many favorites and nine times as many retweets as the rest.

I started paying attention to other technologists I follow. Was it just me? Or did my pattern apply to other women who tweet both about tech and about their experiences as women in the industry? Would I discover different Twitter engagement patterns if I compared men’s and women’s tweets?

In all, I selected 100 men and 100 women in tech, and analyzed their tweets over an eight-week period. All average at least three original tweets a day, and have more than 1,500 followers. The women tweet about both technology and diversity (often gender, but not exclusively gender). The men tweet about technology and at least some social justice topic (25 of them tweeted about diversity during the eight-week period). 20 of the women and 17 of the men are people of color — not enough, but substantially more than we see in the tech industry diversity stats getting published.

Here’s what I found:

Men tweet more often, and fewer of their tweets are about diversity.

I captured 39,233 original tweets from women and 48,158 tweets from men over the eight-week period. The women average just about seven original tweets a day, while the men make more than 8.5.

More striking is the distribution of topics. 23,634 of the men’s tweets are about core tech or the business of tech (49 percent). By contrast, just 12,160 of the women’s are (31 percent).

Some 15,008 of the women’s tweets are about gender and other diversity issues (38 percent). A puny 2,648 of the men’s are (5 percent). More than half of the men’s tweets about diversity come from the 17 men of color that I included in the study, whereas women’s tweets about diversity are evenly distributed across white women and women of color.

Men’s tech tweets are five times more popular than women’s.

The men’s tweets show higher engagement across the board, averaging 12.6 favorites and 11.3 retweets across all topics. The women average seven favorites and 4.9 retweets per tweet.

Narrowing to tweets about core tech only widens this already substantial gap: The women’s tech tweets average 3.2 favorites and 2.7 retweets a piece, while the men’s get 16.4 favorites and 15.5 retweets apiece. That is, the men’s tech tweets are more than five times more popular than the women’s.

Strikingly, the engagement gap holds up even when comparing men and women with similar numbers of followers overall. The men in this study averaged 8,484 followers apiece, while the women averaged 7,899.

Twitter pays more attention when women talk about diversity than when they talk about tech.

Women’s tweets about gender and other diversity topics see higher engagement, getting 11.7 favorites and 10.3 retweets on average. This sounds promising, until you consider that the women’s highest engagement tweets — the diversity tweets — still place them below the men’s average overall.

On the other hand, the men’s diversity tweets significantly underperform their others, garnering just 7.4 favorites and 4.4 retweets apiece. In other words, men’s lowest engagement tweets — the diversity tweets — perform about as well as the women’s average overall.

The Twitter data suggests that the tech industry may have begrudgingly found a way to listen when women talk about their experiences as women in the industry, but hasn’t yet afforded them equal attention in the broader conversation about technology and business.

This is concerning on a number of levels. If the conversations about tech and diversity in tech don’t include the same people, it’s hard to imagine how they can productively come together. Especially if the people talking about diversity in tech come from marginalized groups that have limited sway in the broader industry outside those conversations.

Talk about perpetuating the cycle of unconscious bias: The pattern amounts to keeping women and other underrepresented groups at the kids’ table while the grown-up majority discusses the serious topics. As long as the conversations that evolve the center of the industry remain exclusionary, it’s just not enough for a handful of companies to release their diversity numbers every year and hope for the best.

I’m proud of the work we’ve done at Textio to help companies detect unconscious bias in how they communicate, and I believe that software can change habits and behaviors for well-intentioned individuals. But this data shows bias so deep and across such a large population that it’s almost invisible — until you look at the statistics, where it’s egregious.

It’s a lot of work to hire people outside your network and to build a climate of inclusiveness within your organization. It’s not actually that hard to follow them on Twitter. We can do better.

Kieran Snyder is the co-founder and CEO of Textio, a machine-learning company that optimizes and removes bias from job listings and other critical business documents. She holds a PhD in computational linguistics, and has published widely on text analytics and gender. Reach her @KieranSnyder.

This article originally appeared on Recode.net.