Connect with us

AI

Twitter will abandon its racially bias image cropping software

The software contained some pretty significant bias.

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

Twitter has decided to abandon its image cropping software after it was found that the software contained some pretty significant bias. Users found that the software had a tendency to prefer showing white people over Black people in the image preview that you see in a tweet.

In a blog post on the social media giant’s website, Rumman Chowdhury, a software engineering director for the company’s META (machine learning ethics, transparency and accountability) team posted findings by the team that did, in fact, find certain bias in the software based on race and gender.

These findings come from months of research dating back to late last year when users began noticing the problem. One user in particular, @bascule, posted an experiment showcasing the bias.

They added a picture with Barack Obama and Mitch McConnell to see which face the Twitter algorithm would focus on. No matter how the faces were positioned, the software always chose Mitch McConnell’s face, unless colors were reversed.

https://twitter.com/bascule/status/1307440596668182528

Twitter quickly responded, saying that internal testing was done before the feature launched, but obviously there was more that needed to be done. Working with several third-parties, the company conducted research that ultimately found significant errors in the software.

As of now, the company has almost completely abandoned this software. Twitter currently uses a model that shows the entire picture whenever possible. For pictures that are too large, the app will center crop them down to the maximum possible size.

As the company’s blog post mentioned above, what was really learned here is that machine learning is not always the answer. Just because you can, doesn’t necessarily mean you should. When it comes to things such as images on social media, a more hands-on approach may be the better option.

Have any thoughts on this? Let us know down below in the comments or carry the discussion over to our Twitter or Facebook.

Editors’ Recommendations:

Follow us on Flipboard, Google News, or Apple News

Staff writer at KnowTechie. Alex has two years of experience covering all things technology, from video games to electric cars. He's a gamer at heart, with a passion for first-person shooters and expansive RPGs. Shoot him an email at alex@knowtechie.com

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

More in AI