View profile

The racial bias in Twitter’s image-cropping AI

The racial bias in Twitter’s image-cropping AI
By get.Africa Weekly • Issue #70 • View online
Mangwani akanaka,
That’s “good morning” in Shona
Twitter was originally a text-based social media network.
In its early days, users had to rely on third-party apps like Tweetphoto to upload and share their images. Tweetphoto later became Plixi and was eventually acquired by Lockerz.
Over time, Twitter has been re-engineered to handle images natively. But every once in a while, its users experience the kind of glitches that remind everyone of Twitter’s origins — and that’s as a text-based social media network.
Doesn’t see colour 
Twitter timelines are constrained spaces, hence the 140 and, eventually, 280-character limit. To fit images of different dimensions into user timelines, it uses previews.
And to determine what part of an image should be contained in the preview, Twitter uses, or should I say previously used, a cropping tool that works with an algorithm called a saliency model.
Saliency models are trained on how the human eye looks at a picture as a method of prioritizing what’s likely to be most important to most people. It then crops out the rest of the image.
Over time, Twitter users started noticing that the model, while highly effective as a cropping mechanism, had some significant biases.
Last September, Twitter user @bascule conducted an experiment.
In a tweet, he posted two photos of the then-US Senate Majority Leader Mitch McConnell and former President Barack Obama. In one image, Obama was featured at the top, in the other, he was at the bottom. However, Twitter’s algorithm cropped both images to show only McConnell in the previews, indicating that it found him more “salient” than Obama.
When experiments like this got to Twitter’s attention, the company ran experiments of its own and came up with several findings, two of them stood out to me:
  1. Demographic differences: In a lineup with both black and white people, white people were often cropped ahead of black people. And in a lineup of people of both genders, men were often picked ahead of women. 
  2. Male gaze: The cropping tool picked a woman’s chest or legs as a salient feature.
Read Twitter’s full report.
The opportunity
Following its findings, Twitter has now suspended the use of its image-cropping tool. 
I’ve written about why, because of the bias in today’s most widely-used computer vision systems, it’s too early to deploy the technology for sensitive use cases such as law enforcement and mass surveillance. 
However, this is different.
Twitter’s saliency models aren’t being used to make decisions that can impact an individual’s civil liberties. What this episode does, however, is underline the issue of bias. 
But there are no easy solutions. 
At the first glance, it may seem that the image-cropping tool errors can be articulated purely as a fairness concern. 
And researchers have developed technical ways of defining algorithmic fairness, such as requiring that models have equal predictive values across groups or requiring that models have equal false positive and false negative rates across groups.
In the example of the Twitter saliency model, there was a 4% difference in favour of white individuals, fairness implies that the disparity should be approximately 0%.
However, there are limitations in using formalized fairness metrics in making an assessment. Regardless of the statistical results of the analysis, the risk of representational harm is still present; representational harm is a situation that occurs when people underrepresented in a data set do not have the choice to represent themselves as intended.
For example, default camera settings are often not optimized to capture darker skin tones, resulting in lower-quality database images of Black people. This means that while choosing the salient parts of a picture that contains both lighter and darker complected individuals, some form of bias is likely to persist.
The challenge for the entire ML community is, therefore, to find a holistic, long-term way to solving this bias problem. But the challenge for the African ML community, in particular, is to lead this development or at least play an active role.
Working in the African AI/ML space, I often question why the ML equivalent of Lockerz can’t be developed here. It’s a big ask, I know, especially considering that the community is still in its infancy. But being the continent with the largest population of Black people, the bias affects Africans more than anybody else.

get.Africa is a weekly roundup of the most important stories in African tech. To support, follow us on Twitter, subscribe to our YouTube channel, share this issue or send us an email. You can also check our archives.
Credit: Marek Piwnicki
Credit: Marek Piwnicki
What's the Big Deal?
E£: Egyptian startup Homzmart, a furniture and home goods marketplace, has closed a US$15 million Series A funding round to help it speed up its already impressive growth. Homzmart’s platform enables retailers by helping them easily reach customers. The capital raise was led by MSA Capital.
E£: Egyptian logistics and last-mile delivery company Bosta has raised a US$6.7 million Series A funding round to scale its operations and expand. Bosta currently covers all of Egypt and claims to have a 95 per cent delivery success rate. The round was led by global VC fund Silicon Badia.
E£: Sequoia leads $5M pre-seed in Egypt’s 1-month-old digital bank Telda. The bank plans to digitize how Egyptians save, send and spend money. This is Sequoia’s first venture in MENA.
+ ₦: Future Africa’s recent report reveals it has achieved most of its set goals for the year, including the $1 million investment in female-led startups. According to this report, the firm has invested $3 million in 13 startups, that’s double of its total investment in 2020.
MTN misses out as Kenya's Safaricom wins sole telco licence in Ethiopia
Cryptocurrency markets in the red: The story of a rookie trader
Much ado about API fintech in Africa and the hype is justified
Does Facebook have something to hide?
The case for an innovation and startup act in Ghana
Charts + Graphs
Credit: Salient
Credit: Salient
A new study looks into 61 of the biggest healthtech companies in Africa, it goes into their funding and business strategies. Full report.
+ Reads
+ Data privacy in Nigeria: The government is warning Nigerians against WhatsApp’s new data privacy update, but the government’s new IMEI policy is throwing up privacy questions of its own.
+ Day of reckoning: The South African government is launching an inquiry into anti-competitive practices in online markets; market leaders like Takealot and Uber are expected to be on the hot seat.
+ Based on logistics: Nigeria’s government approves two new, but familiar, regulations for NIPOST. But what does this mean for startups in Nigeria’s logistics space?
+ Getting left behind: While African countries like Kenya and SA continue to roll out 5G, Nigeria’s senate has made the strange decision to put 5G plans on hold over health concerns.
+ E-commerce in Tanzania: A pursuit for the crazy? A deep dive into the operations of e-commerce companies like Duka Direct and Sarafu.
Watch + Listen
Why it's Too Early for African Law Enforcement to Adopt Facial Recognition Technology
Why it's Too Early for African Law Enforcement to Adopt Facial Recognition Technology
More on Nigeria's IMEI registration
On SEC sub-broker license, Abeg, and Paystack in SA
Overheard on Twitter
Andy Obuoforibo
Why should FG have IMEIs of everyone's devices by default, and as a requirement for telcos to grant network access ("whitelist")? Globally, the standard is for consumers to give their IMEI data AFTER a theft ("blacklist/blocklist").
Why can't FG get warrants on a per case basis?
I'm writing a book on remote work for African businesses
I'm writing a book on remote work for African businesses
+ Register a .ng domain name for your business today.
+ To make whiteboard animation videos like this for educational or marketing purposes, please send us an email.
Did you enjoy this issue?
get.Africa Weekly

African tech in a language you'll understand. The most important stories contextualized and delivered to your inbox every Monday. Curated by Chiagoziem O.

In order to unsubscribe, click here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue
get.Africa Media, Nigeria