Twitter staff machine learning engineer Naz Erkan and Senior Director of software engineering Sandeep Pandey announced in a blog post Jan. 29 that the company is partnering up with researchers from UC Berkeley to improve the use of machine learning.
The new research initiative focuses on improving the performance of machine learning in social networks, according to Erkan and Pandey in the blog post. The study is being led by two professors, Moritz Hardt and Benjamin Recht, from the campus department of electrical engineering and computer sciences, or EECS.
According to the blog post, this partnership will help “increase the collective health, openness, and civility of public conversation” on Twitter.
Machine learning is used as a method of data analysis and, according to the blog post, has a crucial role in powering Twitter and enhancing user experience on the platform.
The team at UC Berkeley will be collaborating with a Twitter team to “create a research program that has the right mix of fundamental and applied research components to make a real practical impact across industry,” according to the blog post.
The company said this partnership with Hardt and Recht is “one of many steps” Twitter is taking to involve itself in machine learning, along with sponsoring an upcoming Association for Computing Machinery Conference on Fairness, Accountability, and Transparency.
The study also has a focus on algorithms because, according to the blog post, safeguarding the company’s algorithms against exploitation from individuals or groups is essential.
Erkan and Pandey added in the blog post that the consequences of exposing algorithmic decisions and machine learning models to a large audience are poorly understood.
“Even less is known about how these algorithms might interact with social dynamics: people might change their behavior in response to what the algorithms recommend to them,” Erkan and Pandey said in the blog post. “As a result of this shift in behavior, the algorithm itself might change, creating a potentially self-reinforcing feedback loop.”
The blog post said that by working with a team of researchers from UC Berkeley and combining the academic and industry perspectives, the company will be able to do fundamental work in its developing space and apply the study to improve Twitter.
“In general, if you think about machine learning, it’s composed of three pieces. The first piece is the interesting part for Twitter, such as finding out about what people talk about,” said Gerald Friedland, an EECS adjunct assistant professor. “For machine learning, you use the data. For example, you can increase the revenue by finding patterns.”