Experts: Open-source code won’t solve Twitter’s algorithm problems

Releasing the source code of Twitter’s recommendation algorithms won’t solve the social network’s problem, according to a number of data-privacy and online-security experts, Wired reports.

According to MIT professor and computer scientist David Karger, in the age of machine learning what matters are not the algorithms but the data. He argues that Elon Musk could improve Twitter by making the platform more open so that others could use it in new ways.

“What makes Twitter important is not the algorithms. It is the people who tweet,” Karger said.

Brookings Institution fellow Alex Engler, who studies the impact of AI on society, shares a similar view. He says that a key moment of openness for the platform could be understanding the decision-making processes by which the algorithms are trained.

“The code is good. The data are in order. Code and data, merged into a model, may be better,” the expert said.

An anonymous company employee said that machine-learning models at Twitter are merely the tip of the iceberg. The entire system responds in real time to visitors’ behavior in a complex way. If users show particular interest in a certain news item, the corresponding tweets will appear more often.

“Twitter is a socio-technical system. It responds to human behaviour,” the source stressed.

Another employee said there is no single algorithm for promoting content in the main feed. Such decisions are the result of many different models interacting with each other based on user actions. The results are also tailored for each visitor to the platform.

“There is no ‘main algorithm’ for Twitter,” the source from Twitter’s technical team said.

Nevertheless, experts welcomed Musk’s attempts to make the company’s algorithms transparent and understandable for users and researchers.

According to Derek Ruths, associate professor at McGill University in Canada, social platforms are too hard for students to study. He added that transparency could reveal how the algorithms work and intensify pressure on other social networks.

“This could be a genuinely interesting experiment that has long been overdue,” Ruths said.

Earlier, Twitter board of directors accepted Elon Musk’s offer to buy the company for $44 billion. Primarily, the businessman intends to raise the platform’s transparency, open the source code of the recommendation algorithms, and ensure the principle of freedom of speech for all users.

In April, the European Union approved the Digital Services Act, regulating recommendation algorithms and targeted advertising.

In February, a group of American senators introduced a similar bill in Congress.

In January, China approved rules regulating recommendation algorithms.

Subscribe to ForkLog News on Telegram: ForkLog AI — all the news from the world of AI!