- Twitter users believe they discovered a racial bias in an algorithm the platform uses to automatically select which part of an image it shows in a photo preview.
- Many argued that the auto-cropping tool showed a white bias after testing the theory with photos of Black and white people, cartoon characters, and even dogs.
- However, others who tested the theory generated results that did not support this idea. Regardless, most users admit that these experiments have their limitations and agree that the current results at least show that this is something worth looking into.
- The company released a statement saying it tested its system for bias in the past but admitted it needs to conduct further analysis of it. Online, Twitter employees seemed to welcome the public discourse and the company promised to share its results as well as further actions it may take.
Potential White Bias
Twitter responded to concerns over its automatic cropping algorithm Sunday after users believed they discovered a racial bias in the tool.
In 2018, Twitter began auto-cropping photos in its timeline previews to prevent them from taking up too much space in the main feed and to allow multiple photos to appear in the same tweet. To do this, the company uses several algorithmic tools that focus on the most important part of the picture, like faces or text.
However, users recently began to spot issues with the algorithm. The first person credited for highlighting a potential problem was PhD student Colin Madland. He made his discovery while highlighting a different racial bias he thinks he found on the video-conference company Zoom.
Madland tweeted that when his Black colleague uses a virtual background on Zoom, his head is erased. When he uploaded examples to show this happening to his Black colleague and not himself, he noticed that Twitter was only showing his own face in its preview.
Soon after, others followed up with more targetted experiments. Cryptographic and infrastructure engineer Tony Arcieri, for example, tweeted out two long images with Senate Majority Leader Mitch McConnel and Former President Barack Obama.
The two photos have the politicians stacked on top of each other in different orders but with white space in between them. The experiment showed that Twitter would focus on McConnell, no matter what order the photos were stacked in.
Another user found that the algorithm even focused on McConnell when two photos of Obama were present in a single stack.
I wonder what happens if we increase the number of Obamas. pic.twitter.com/sjrlxjTDSb— Jack Philipson (@Jack09philj) September 19, 2020
A similar white preference appeared in examples of Black and white men in suits, Simpsons characters Lenny and Carl, and even black and white dogs.
Examples That Don’t Support White Bias Theory
Others looking into this theory of a white bias found results that did not support the idea.
For example, one user found that photos of Obama were cropped for the preview over photos of Donald Trump.
Still, some researching the trends noted that these experiments do have their limitations and are likely influenced by tons of other factors. Some believe the algorithm recognized high profile figures or considers brightness and contrast, among other photo elements.
Twitter’s Chief Design Officer (CDO), Dantley Davis, even suggested that the choice of cropping sometimes takes brightness of the background into consideration.
However, ohers found examples that rejected that idea. Regardless, all these tests did a lot to convince people that there was something worth looking at here, including Davis, who has been experimenting himself.
He’s not alone in his research. In fact, plenty of other Twitter users have been going to great lengths to track their results as they try to study what is going on.
Twitter Promises to Investigate
On Sunday, a Twitter spokesperson eventually released a statement admitting that the company had work to do.
“Our team did test for bias before shipping the model and did not find evidence of racial or gender bias in our testing,” the company explained.
“But it’s clear from these examples that we’ve got more analysis to do. We’ll continue to share what we learn, what actions we take, and will open source our analysis so others can review and replicate.”
Davis also isn’t the only employee that has appeared to welcome all of this public discourse. The company’s Chief Technology Officer, Parag Argawal tweeted, “This is a very important question. To address it, we did analysis on our model when we shipped it, but needs continuous improvement. Love this public, open, and rigorous test — and eager to learn from this.”
See what others are saying; (The Next Web) (The Guardian) (Mashable)
Robinhood Crypto Trading Crashes Twice as Dogecoin Multiplies in Value, Enraging Users
- Robinhood users found themselves unable to buy or sell cryptocurrency Thursday night, an issue reminiscent of the app’s decision to restrict GameStop trades earlier this year.
- While Robinhood resolved the problem within a matter of hours, it came amid a massive rally on Dogecoin, a cryptocurrency that started out as a joke. The app’s crypto services briefly went down again Friday morning as the rally continued.
- Robinhood has denied that its crypto trading outages were an intentional effort to drive down Dogecoin prices and instead blamed the outages on “unprecedented demand for Robinhood Crypto services.”
- By Friday morning, Dogecoin briefly soared to $0.45, more than 400% of the value it had at the beginning of the week and more than 4,500% of the value it had at the beginning of the year.
Robinhood Crashes Amid Dogecoin Rally
The joke cryptocurrency Dogecoin has surged more than 400% this week alone, but around 10 p.m. EST Thursday night, the free-to-trade app Robinhood tweeted that it was “experiencing issues with crypto trading.” In turn, that caused many of the app’s users to find themselves unable to execute trades.
Dogecoin first began to spike Tuesday ahead of the market debut of the cryptocurrency exchange Coinbase, which raised $86 billion in its first day of trading. That morning, one Dogecoin amounted to about $0.07. By midnight, it had doubled in value. Those gains continued Thursday evening when Dogecoin spiked to around $0.33.
That may not seem like much, but if a person invested $1,000 in Dogecoin when it was selling for around $0.01 at the beginning of the year, by Thursday evening, that person would be sitting on a small fortune of around $33,000 before taxes.
Robinhood Users Angry Yet Again
Many Robinhood users found themselves frustrated when they were unable to sell off their existing dogecoins, especially since the cryptocurrency’s value was rapidly falling.
In fact, within the matter of just over an hour, it had dipped to around $0.25. Using the last example above, that would mean thousands of dollars of missed opportunity.
“Are you going to cover my account?!?” one user asked Robinhood when she found herself unable to sell her dogecoins. “This is a technical error, not my own risk. Ive been trying to execute this transaction for almost two hours! None of my crypto comes up!”
Are you going to cover my account?!? This is a technical error, not my own risk. Ive been trying to execute this transaction for almost two hours!— JT (@JenninNYC) April 16, 2021
None of my crypto comes up! pic.twitter.com/rZ4qoXostn
This is not Robinhood’s only bout with controversy. Earlier this year, the company infamously blocked its users from buying GameStop stock during a frenzy that sent shares from under $20 to nearly $500 at one point; however, Robinhood still allowed users to sell their existing shares — a move that even if it lacked the intention, had the effect of attempting to drive share prices for GameStop down.
Though CEO Vlad Tenev later argued that the company “had no choice” but to restrict buying, Robinhood’s decision nonetheless sparked the ire of its users and even prompted Congressional investigations.
Many Robinhood users were quick to point that out Thursday when they once again found themselves unable to execute trades. Some even accused the company of more nefarious intentions.
Service Restored… Until It Went Down Again
At 11:46 p.m. Thursday night, Robinhood tweeted that crypto trading had been “fully” restored.
“Like others, we were experiencing unprecedented demand for Robinhood Crypto services, which created issues with crypto trading,” the company said. “We’ve resolved the issue and apologize for the inconvenience.
Multiple times since Thursday evening, the company has denied that it intentionally halted crypto trading to affect Dogecoin prices.
“Unprecedented demand for Robinhood Crypto services created temporary issues with crypto trading,” a Robinhood spokesperson told the New York Post Friday. “That’s it, plain and simple.”
On Friday morning, Dogecoin went on to spike at a current 52-week high of $0.45; however, it soon dipped back into the mid- to upper-thirty-cent range, where it remained around 3 p.m. EST.
Meanwhile, amid the surging demand, Robinhood experienced yet another crypto outage around 10:30 a.m. EST Friday. Just before 11 a.m., it said that trading had been restored for most customers.
See what others are saying: (New York Post) (Business Insider) (Coindesk)
Child Safety Advocates Urge Facebook To Scrap Plans for Instagram Kids
- Nearly 100 child safety experts and international organizations sent a letter to Facebook Thursday criticizing its plans to develop an Instagram app for children under 13.
- Facebook claims the app will offer parental controls and is meant to create a safer space for kids, who are often lying about their age to access the normal version of Instagram.
- Still, critics point out that children already on Instagram are unlikely to switch to a kids version. Many also cited concerns about screen time, mental health, and privacy, arguing that younger children are not ready for such a platform.
- U.S. Lawmakers expressed similar concerns earlier this month, saying, “Facebook has an obligation to ensure that any new platforms or projects targeting children put those users’ welfare first, and we are skeptical that Facebook is prepared to fulfill this obligation.”
Instagram for Kids
An international group of 35 organizations and 64 experts, coordinated by the Campaign for a Commercial-Free Childhood, released a letter Thursday urging Facebook to abandon its plans to release an Instagram app for kids under 13-years old.
Plans for Instagram Kids have been public for about a month after Buzzfeed News obtained emails about the app in mid-March. Since then, there have been widespread concerns about how such an app could affect children.
Thursday’s letter argues that a version of Instagram targeting under-13-year-olds raises concerns about privacy, screen time, mental health, self-esteem, and commercial pressure. Stephanie Otway, a spokesperson for Facebook, said the company understands the concerns presented by the Campaign for a Commercial-Free Childhood.
“We agree that any experience we develop must prioritize their safety and privacy, and we will consult with experts in child development, child safety and mental health, and privacy advocates to inform it,” she said.
“The reality is that kids are online. They want to connect with their family and friends, have fun and learn, and we want to help them do that in a way that is safe and age-appropriate. We also want to find practical solutions to the ongoing industry problem of kids lying about their age to access apps,” Otway added, noting the reality of how many children interact with age-gated apps.
Unlikely To Stop Children From Joining Regular Instagram
The idea that children would just switch to Instagram Kids received pushback from the Campaign for a Commercial-Free Childhood. In fact, the group’s executive director, Josh Golin, pointed out that most kids who are currently on Instagram are between 10 and 12-years-old, and they likely wouldn’t migrate over to Instagram Kids because it will be perceived as “babyish and not cool enough.”
”The children this will appeal to will be much younger kids,” Golin explained. “So they are not swapping out an unsafe version of Instagram for a safer version. They are creating new demand from a new audience that’s not ready for any type of Instagram product.”
It’s unknown exactly how the app would work, but it would feature content similar to what is allowed in other age-appropriate apps, such as YouTube Kids. One of the few details given out so far is that Instagram Kids will be ad-free and feature parental control options.
Concerns over Instagram Kids has also come from lawmakers. On April 5th Senators Edward Markey (D-Mass.) and Richard Blumenthal (D-Conn.), alongside Representatives Kathy Castor (D-Fla.) and Lori Trahan (D-Mass.), sent a letter to Facebook CEO Mark Zuckerberg expressing concerns that “children are a uniquely vulnerable population online, and images of kids are highly sensitive data.”
“Facebook has an obligation to ensure that any new platforms or projects targeting children put those users’ welfare first, and we are skeptical that Facebook is prepared to fulfill this obligation.”
See what others are saying: (TechCrunch) (BBC) (NBC News)
Retail Sales Jump Amid Stimulus Spending, Unemployment Claims Plunge To Pandemic Low
- The Commerce Department released a report Thursday recording a 9.8% spike in retail sales for the month of March.
- That surge was largely driven by stimulus check spending, with restaurant, sporting goods, clothing and accessory, and auto sales all being among the top-performing sectors in retail for the month.
- Coupled with that news, the Labor Department reported that 576,000 unemployment claims were filed last month — a pandemic low.
- That figure is still significantly higher than the roughly 200,000 weekly unemployment claims filed before the pandemic.
Retail Sales Spike
U.S. retail sales for the month of March jumped 9.8% from February, according to a Thursday morning report from the Commerce Department.
That spike is largely thanks to the most recent round of stimulus checks from Congress.
March was the best month of retail spending since May of last year, which at the time saw an 18.3% gain following the first wave of stimulus checks.
Sales in the bar and restaurant industry rose 13.4%, making them among the retail sectors that saw the biggest spikes last month. That’s largely a result of relaxed lockdowns stemming from the country’s current pace of around three million vaccinations a day. Meanwhile, sporting goods spending rose 23.5%, clothing and accessory sales rose 18.3%, and motor vehicle parts and dealer sales rose 15.1%.
“Spending will almost certainly drop back in April as some of the stimulus boost wears off,” wrote Michael Pearce, senior U.S. economist at Capital Economics, “but with the vaccination rollout proceeding at a rapid pace and households finances in strong shape, we expect overall consumption growth to continue rebounding rapidly in the second quarter too.”
Unemployment Hits Pandemic Low
The retail sales data came around the same time that the Labor Department released this past week’s unemployment figures, which dropped to a new pandemic low of 576,000 claims.
That’s a massive difference from almost exactly a year ago when 6 million people filed for unemployment in a single week. It’s also a significant decline from the 769,000 people that filed jobless claims last week, especially since some analysts had predicted there would be around 700,000 jobs lost with this week’s report.
That said, unemployment claims are still much higher than the around 200,000 a week that were being filed prior to pandemic closures.
“You’re still not popping champagne corks,” Diane Swonk, chief economist at the accounting firm Grant Thornton, said according to The New York Times. “I will breathe again — and breathe easy again — once we get these number[s] back down in the 200,000 range.”