Changing my gender to male on Google+ improved my What's Hot experience
Earlier today, +Kimberly Chapman
told me about a little experiment she'd done. She'd opened a tab on her browser and directed it to plus.google.com/explore
, a stream that enables you to see the hottest content on Google+. In another tab, Chapman changed her gender on her Google+ profile from female to male, and then loaded the Explore page into another tab.
The difference was astonishing. "As a girl, I get stupid warm-fuzzy meaningless shit," Chapman writes on a post (http://goo.gl/ZDBxT
). "As a guy, I get nerdy stuff. This is not okay. I get that [What's Hot] algorithms are based on what people click, like, share, comment on, etc. Fine. But I challenge anyone to give me one good reason why there should be such a drastic difference in less than ten seconds by simply changing my gender, other than institutionalized sexism about what girls and guys apparently like."
I tried the experiment myself just a moment ago. Below are ten posts that graced my screen, organized into pairs. I have paired them based on their placement on my stream and no other factors. The example image shown below, for instance, shows a post with a quote and one with a bar graph -- the quote was the fifth post on my Explore stream when I was female and the bar graph was the fifth post on my Explore stream when I was male.
Note that I have not changed anything other than my profile. And despite my extensive interaction with the social network, when I'm female, gadgets might as well not exist. When I was male, I saw two posts referencing #io13
, but there was no mention of the event when I was female. In fact, the closest thing to tech news I got as a female was an article about how to become an influencer on Pinterest. Needless to say, this article didn't show up when I was male.
Neither did the nausea-inducing feel-good quotes on pictures. +Yonatan Zunger
, Chief Architect of Google+, has responded to the issue, saying: "What's Hot is based both on properties inherent to the post -- how people have reacted to it when they saw it, +1's, reshares, and so on -- and on properties of the viewer, especially profile info. Gender is one of those signals, as are quite a few other things. From what I can tell, gender is having a relatively large effect on the result set, because our models (based on people's actual responses to seeing these things) seem to show a pretty sharp gender difference in response."
He agrees the model isn't sufficiently sensitive, adding that a team is currently working on content recommendations that better reflect individual user preferences. He assures users that "there was no human editing to say that women like X and men like Y."