BEIJING, Dec 20 (Reuters) – TikTok’s content recommendation algorithm is back in the spotlight after Chinese company ByteDance, the app’s parent company, signed binding agreements to form a joint venture that will hand over control of the app’s U.S. operations to U.S. and global investors, including database company Oracle.
Although the creation of this new entity represents a major step towards avoiding a ban on the app in the US, in addition to easing tensions between Washington and Beijing, there is still uncertainty regarding the ownership of the recommendation algorithm, considered TikTok’s crown jewel.
Rush Doshi, who worked at the US National Security Council under President Joe Biden, said on X that it is unclear whether the algorithm has been transferred, licensed or is still owned and controlled by Beijing, with Oracle merely providing “monitoring”.
The algorithm is considered essential to TikTok’s global success and, until a few months ago, ByteDance’s position was that it would rather shut down the app in the US than sell it.
However, in September, Reuters reported, citing sources, that ByteDance would retain ownership of TikTok’s U.S. business operations but would cede control of the app’s data, content and algorithm to the joint venture.
The joint venture serves as a support operation for the U.S. company and handles U.S. user data and the algorithm, the sources said at the time, adding that a separate division that will continue to be wholly owned by ByteDance will control revenue-generating business operations such as e-commerce and advertising.
Continues after advertising
Those agreements formed the outline of the deal announced on Thursday, two sources with knowledge of the matter told Reuters on Friday. The North American entity TikTok, controlled by ByteDance, would be the revenue generator, while the new joint venture will receive a share of the revenue for its technology and data services, the sources said.
The Chinese government has not yet declared its position on the signed agreement. Beijing made changes to its export laws in 2020 that give it approval rights over any export of algorithms and source code, adding a layer of complexity to any effort to sell or separate the app from the US.
What makes the algorithm powerful?
TikTok has shown that an algorithm, driven by understanding a user’s interest, can be powerful. Instead of creating its algorithm based on the “social graph” like Meta, TikTok executives said its algorithm is based on “signals of interest.”
Continues after advertising
The short video format allows TikTok’s algorithm to become much more dynamic and even able to track changes in users’ preferences and interests over time, getting as granular as what a user might like during a certain period of the day.
And TikTok’s positioning as an app built for mobile devices from the start has also given it an edge over rival platforms that have had to adapt their interfaces from computer screens.
The fact that TikTok entered the short video market early also gave the company a huge first-mover advantage. Instagram didn’t launch Reels until 2020, while YouTube launched Shorts in 2021, and both lag behind TikTok in years of data and product development experience.
Continues after advertising
What does the research reveal about the algorithm?
TikTok also regularly recommends content that doesn’t fit users’ interests, which company management has repeatedly said is essential to TikTok’s user experience.
One study, published by US and German researchers last year, found that TikTok’s algorithm “exploits user interests in 30% to 50% of recommendation videos,” after examining data from 347 TikTok users and five automated bots.
“This finding indicates that the TikTok algorithm chooses to recommend a large number of exploration videos in an attempt to better infer the user’s interests or maximize user retention by recommending many videos that are outside the user’s (known) interests,” the researchers wrote in the paper titled “TikTok and the Art of Personalization.”
