Social Media networks like safely buying real Twitter followers are not new to having issues in their systems. Sometimes, these issues could lead to critics, or worse, in hearings. Naturally, they want to solve these issues before they grow into bigger problems. That is why they are constantly trying to improve their systems.
Twitter is now looking to improve its machine learning algorithm. The team aims to assess the “unintentional harms” their algorithms can cause. Specifically, they want to remove algorithmic biases that affect their users negatively. To do this, they are starting a new initiative called Responsible Machine Learning.
There is A Pressure Behind This Move
Why is Twitter doing this out of the blue? The answer is, this is not a random decision made by their management. There is ongoing pressure from the government that made them act.
Senators Cory Booker (D-NJ) and Ron Wyden (D-OR), with a House equivalent sponsored by Rep. Yvette Clarke (D-NY), sponsored the Algorithmic Accountability Act. This bill would require large companies to review machine learning-powered systems for bias.
If this bill materializes into law, the Federal Trade Commission will create rules for evaluating the “highly sensitive” automated systems of big companies. Highly sensitive in this context means they hold large amounts of information. In this case, a large amount means information on at least 1 million people or devices. Companies that act as data brokers that buy and sell consumer data are also not safe from the bill. The Algorithmic Accountability Act will apply to companies that earn more than $50 million annually.
This bill will require companies to evaluate their algorithms. If proven that they pose risks of discrimination, privacy problems, or other issues, the company in question should act fast in solving the issue. Else, the punishment shall come to them.
Making changes in the algorithm is quite a complicated task. They can indeed improve the system, but sometimes, they can also produce unintentional results. Twitter understands this and does not plan to take on this risk lightly.
So, they garnered people from their many departments to form the initiative.
Their formed ML Ethics, Transparency, and Accountability (META) will be leading the team. These people are engineers, researchers, and data scientists tasked to identify possible harm caused by changing the algorithm. Other members of the initiative were people from technical solutions, research, trust and safety, and product teams.
The initiative consists of these four pillars:
- Taking responsibility for Twitter’s algorithmic decisions
- Equity and fairness of outcomes
- Transparency about Twitter’s decisions and how the team arrived at them
- Enabling agency and algorithmic choice
Responsible Machine Learning’s mission is to eliminate gender and racial bias in Twitter’s algorithms. The other social media giant, Facebook, has made a similar effort in the past. Now, Twitter is following its example.
To do so, they are going to do a bunch of studies and assessments. They also promised to let the public know about their findings. The assessments will be up for public viewing in a few months.
The first assessment is about how different genders and races feel about their image cropping/ saliency algorithms. The issue popped up after Twitter users pointed out that the auto-cropped photo previews are favoring white faces over black faces. The company has already delved into this issue. They began testing, showing full photos instead of auto-cropped previews.
The second assessment is a fair assessment of how timeline recommendations differ across racial subgroups. Meanwhile, the third one is an analysis of content recommendations across political ideologies across seven countries.
By doing these analyses, they will be able to know how the machine learning algorithms that they use contribute to discrimination. Thus, they will be able to fix them.
What to Expect?
With the law running after them or not, what Twitter wants is to give its users a more pleasant experience. They stated that they want Twitter users to have more control and input so they can shape Twitter according to their preferences.
For instance, Twitter is looking at the concept of algorithmic choice. CEO Jack Dorsey envisions this as an “app-store-like view of ranking algorithms.” The algorithmic choice will allow users to select what algorithms control their feeds. This feature will help them customize Twitter to how they want it to be.
Aside from this, Twitter says it is working closely, and is open to collaborations, with third-party researchers. They are doing it in hopes that their partnership will help them get better results. Furthermore, they value feedback from the public. So to get them involved, they are going to share their analyses and ask for their opinions.
What Is The Responsible Machine Learning Initiative?
What Twitter is doing is beneficial not only to fight discrimination but also to advance the tech industry. Since they will be transparent in their studies and analyses, other people and companies in the field can take a peek and learn from them. Twitter also said that they are open to revealing their mistakes and unsuccessful attempts. Optimistically, they believe that this should help the industry have a deeper understanding of the topic. Meanwhile, they would profit from that by being held accountable for that information. It is a triple-win situation. Twitter wins by improving its systems. People win by gaining a safer browsing experience. And, the industry wins by gaining more knowledge. How can it get any better than that?
Truly, what Twitter wants to achieve is great, and so we are all eager to see them pull this off. If they succeed in this, it will open the gates to many possibilities and opportunities. The way we see and use social media, and the way they operate, may be changed and overhauled.
Of course, all of these are still in theory. They are, after all, still in the early stages of experiments and studies. Nothing yet is for certain. However, it is clear that they are heading in the right direction, and what they are doing is called progression.
If you have questions or you want to help with the development, you can tell Responsible ML using the hashtag #AskTwitterMETA.