Twitter works on option to undo published tweets 1:05
(CNN Business) – Twitter has largely abandoned an image cropping algorithm after determining that the automated system was biased. Some users complained that they preferred to display images of white people in tweet previews.
In a blog post Wednesday, Rumman Chowdhury, director of software engineering for Twitter’s Machine Learning Ethics, Transparency and Accountability team, wrote that the company concluded that the algorithm was biased after testing it for biases based on the gender and race. The post and an accompanying research article detail how the clipping system favored white people over black people and women over men, for example, when it was tested on randomly linked images of people of various races and genders.
Twitter’s conclusion comes months after the company said it would study the algorithm and serves as yet another example of how bias can infiltrate computer systems meant to perform tasks humans are typically exceptionally good at.
“We considered the trade-offs between the speed and consistency of the automated cut with the potential risks that we saw in this research,” Chowdhury wrote. “One of our conclusions is that not everything on Twitter is a good candidate for an algorithm and, in this case, how to crop an image is a decision that people make better.”
7 out of 10 smartphones will have artificial intelligence 0:42
Changes after complaints
In March, Twitter began testing a new way to display a full image, rather than an automatically cropped preview version, on mobile devices when a user tweeted a single image. The company said that after receiving positive feedback, it rolled out the feature to all of its iOS and Android users in May. (However, it centers images that are extremely long or wide.)
A Twitter spokesperson told CNN Business on Wednesday that the change came first to the Twitter mobile app, as this is how most people tweet and look at images.
When the image cropping algorithm was in place, every time a user posted an image to Twitter, the automated system would crop a preview version of that image that viewers would see before clicking on the full-size image. Twitter said in a blog post in 2018 that it previously used face detection to help figure out how to crop images for previews.
However, the face detection software was prone to errors. The company discarded that approach and instead focused its software on what’s known as “prominence” in images, or the area deemed most interesting to a person looking at the big picture. Prominence is studied by tracking what people look at; We tend to be interested in things like people, animals, and text, for example.
Last September, the company was asked to study its algorithmic approach to cropping images after it was criticized in numerous tweets.
Among these was user @bascule, who on September 19 tweeted: “Trying a horrible experiment … Which one will the Twitter algorithm choose: Mitch McConnell or Barack Obama?” Along with his words were two long, rectangular images.
He explained that the first consisted of a photo of the United States Senate Majority Leader McConnell at the top, which is white, with a thin white rectangle in the middle, and a photo of former President Obama, which is a black person, at the bottom. The second featured the opposite, with Obama at the top and McConnell at the bottom. With Twitter’s image cropping algorithm in use, the versions of its preview, which are side by side, show only McConnell.
A day earlier, another Twitter user @colinmadland noticed a similar preview result when he posted a photo that he said showed himself, a white man, alongside a photo of a black man with whom he attended. an online meeting; The Twitter preview was default to show only the white man.
In response to @bascule at the time, the company tweeted that it saw no evidence of racial or gender bias during testing before launching the preview feature, but said it would investigate if there were any issues with the clipping algorithm.
In Wednesday’s Twitter blog post, Chowdhury wrote that moving away from using an algorithm to crop images reduces the company’s reliance on machine learning (an artificial intelligence technique in which a computer learns itself by analyzing the data) “for a function that we agree is best performed by people who use our products.”
The Twitter spokesperson said the company plans to download the image cropping algorithm on the Twitter.com website in the coming months. The algorithm is also used in two other ways, such as when a person tweets multiple images; the spokesperson said that Twitter is also working on improvements for those uses.