In recent discoveries about deep-rooted racism in our societies, twitter users have discovered new discrimination. Social media is considered one of the most unfiltered and free media of expression. However, it looks like that notion is not valid. Platforms like Twitter and Facebook often alter the view of posts to control the narrative. However, that is acceptable for images or posts unsettling for some audience. It is not right when you try to hide an entire race from the front page.
Racism is one of the significant problems in our society right now. No matter how much you try to educate and police the ordinary people. There is always a part of our community that holds these outdated and wrong norms. You would think that with more and more automation in society, racism would decrease. This argument makes sense because we know part of the reason racism exists is because of multicultural people in power. Computers do not belong to a single race, and they should, in theory, treat everyone equally.
However, that is partially true. Modern computing algorithms are based on the concepts of machine learning, which is teaching a computer to think like humans. Most of the multinational companies use these concepts to sort potential hiring resumes. It is fair to assume that social media platforms also employ similar techniques to control what users see. Now imagine if a machine learning algorithm is trained on data that prioritizes a certain race above others. Not only does this create a bias in the system, but it is also a new form of racism.
Coming back to twitter now, in a recent discovery, a user found out that if you post a vertical picture of a white man and a black man on Twitter. The website algorithm only keeps the white person in the frame. It doesn’t matter who the person is in the picture or their importance. One first argument could be the difference in the popularity of the two people. To refute this, the person used images Mitch McConnell and former US president Barack Obama and here are the results.
- Advertisement -
Trying a horrible experiment…
Which will the Twitter algorithm pick: Mitch McConnell or Barack Obama? pic.twitter.com/bR1GRyCkia
— Tony “Abolish (Pol)ICE” Arcieri 🦀 (@bascule) September 19, 2020
As it turns out, no matter the order or the popularity, the twitter algorithm instantly framed the site person in both of the images. This is scary because most of the people don’t open pictures while scrolling social media. Hiding people like this could cover up an important issue or and emergency.
After this tweet, a series of other tweets also began on Twitter. People used images of different people in different orders to find an explanation for this behavior. Consequently, there were several plausible explanations. Let’s go through these first to see if any of them actually make sense.
The first argument that the post author received was that Twitter prefers red tie color. As wire as it sounds, we must realize that computers work in mysterious ways, and even a detail as small as tie color can change the results. To verify this, he posted the images again. This time changing tie color to blue and guess what. Twitter also put McConnell in the frame, as you can see below.
Interestingly if you invert the colors on the picture, you get some mixed results. It is clear that the algorithm works by using the contrast of face and background. Once you invert the colors on the pictures, the algorithm gets a bit confused. Also, a few users even photoshopped the skin color of Obama. Once Obama had white skin, the Twitter algorithm put him in the spotlight. This, in turn, confirms the racial bias of this system.
The problem is not because of these specific images. You can use any photos from the internet or your own, for that matter. Twitter will give similar results in every case. A user even used the same face with different color complexions, and twitter still prefers the white one.
In conclusion, racism has now made its way into almost every machine learning algorithm for consumer applications. Racism is not just limited to social media or mobile phones and computers anymore. This article explains how the healthcare system in the US is automated and yet biased against Black people. The system works by assigning risk scores to each patient. In their study, people found out that this algorithm gives lesser risk scores to black people who are equally sick as a white person. A lower score, in turn, translates to more inferior quality healthcare for said individual.
In addition to healthcare, this disease has also spread to policing algorithms. Modern policing techniques use two types of algorithms to predict crimes and determine connections between suspects. The first one is a location-based algorithm. It tracks past events at a location and crime rates to predict crime locations. The other analyses personal data from people, such as their age, gender, marital status, history of substance abuse, and criminal record. After analyzing this data the algorithm predicts the chances of a person to commit crime.
As it turns out, both of these algorithms are also frequently biased to black people and black neighborhoods. In conclusion, computer algorithms are the backbone of automated systems in the future. It is, therefore, vital that these systems should be regulated to remove any biases in them. Otherwise this injustice will onlyincrease and get even worse.