Everything you look at and buy is known by your "friends"

Expert: Beware of apps' acquaintance-recommendation features infringing on user privacy

Mr. Wang, a retiree from Nanchang, Jiangxi, has been feeling somewhat frustrated lately: the health-preservation videos he watches in his spare time have all become known to his children, who often urge him not to believe pseudo-science, not to trust folk remedies for illnesses, and to go to the hospital when he feels unwell.

"My phone is always in my hand, and no one is beside me when I watch short videos. How do they know what I've been watching?" Only after talking with other elderly people in his residential community did Wang learn that he habitually taps "like" whenever he sees a video he enjoys, and when his "friends" come across that video, they can see the "traces" that he has watched it.

Many people have encountered the same confusion as Mr. Wang. The topic "Don't recommend people I may know to me" even trended on social media not long ago, sparking complaints about the acquaintance-recommendation mechanisms of social apps. Some netizens said bluntly that they not only don't want their social accounts recommended to people they know, but also want to block "people you may know" so as not to be disturbed or spied on; others cried out in frustration, "I just want to surf the internet quietly, but the app takes it upon itself to recommend me to acquaintances in real life. Can't you leave us some privacy?"

A recent assessment by reporters from Legal Daily of several social and shopping apps found that the recommendation-disabling feature on some apps exists in name only. Besides pushing people you may know, these platforms may also let friends and others know which content you have liked, what you have bought, and which livestreams you have watched, all through their push mechanisms.

The experts interviewed suggested that platforms should address user pain points by improving push mechanisms, and should also pay attention to the user-information collection issues behind acquaintance-recommendation features. Online operators should obtain users' consent before collecting their personal information, and platforms should clearly inform users of the scope and purpose of the information collected.

Watching a livestream gets you noticed

Turning off recommendations doesn't work

Wang Junnan, a student at a university in Guangdong, likes to play games in his spare time to relieve study pressure. Unexpectedly, his mother recently caught on to this hobby and said he was "playing games all day and neglecting his studies." After he protested, his mother actually sent him a screenshot showing the game footage and the date.

Wang then realized that the game he plays requires binding social platform information, and after each game it automatically edits the match content, which can be seen by the people in the same match and by friends. The authorizations for generating game videos and displaying account information all showed "enabled," but he remembered that he had never checked those options. And even after he later turned off these authorizations, the system still automatically generated videos, and these videos could not be deleted.

Mr. Chen, a programmer at a company in Shanghai, had an even more awkward experience. A while ago, a colleague asked him at work whether he was losing his hair, saying, "I saw you watching a livestream selling hair-growth lotion last night."

"My colleague said that when he opened the social platform, he saw a prompt saying 'Your friend is watching ×× livestream,' and just like that I was 'exposed.' I've also watched livestreams on how to treat hemorrhoids and solve men's health issues before—could my friends have seen those too and just not mentioned it? Can the platform leave me some privacy?" Mr. Chen said.

The practice of pushing acquaintances to users through algorithmic matching is common across all kinds of apps. For example, one social app not only displays "people you may know" accounts in sections such as the message bar and follower list, but also occasionally gives strong prompts through pop-up notifications. Some platforms also push "people you may know" and "possibly your friend" accounts to users. Family members, current colleagues, classmates or former colleagues, old schoolmates, even ex-partners who were blocked long ago, or new friends whose contact information was just added, all appear in the "people you may know" recommendation list.

The reporter's assessment found that most apps have an option to turn off acquaintance-recommendation services, but they usually default to the user accepting recommendations. After the reporter chose "don't recommend me to people I may know" and "don't recommend people I may know to me," although the frequency decreased, the app still popped up notifications pushing accounts of people you may know.

The reporter registered and logged into a short-video app using a phone whose contacts contained no other people, and found that many people were still recommended as "people you may know," some of whom were indeed the reporter's friends.

Buying products gets you watched

Users complain there is no privacy at all

Wang Chun, a resident of Xi'an, Shaanxi, recently discovered that some shopping apps have launched a "what others are buying" feature—matching and recommending based on users' shopping similarity, for example, "he, with 83% shopping similarity, bought a desktop ornament" and "she, with 98% shopping similarity, bought a dress."

Regarding this feature, Wang Chun called it "terrifying": "The shopping app pushes what consumers buy like posting on Moments; any platform friend can see it. If you buy intimate items like underwear, wouldn't you have no privacy at all if others saw it?!"

The reporter's review found that some shopping apps enable permissions such as "watched livestreams" and "my products can be recommended to friends" by default. Some let you directly see which friend bought what, while others vaguely show how many friends bought a given product. After the reporter turned off these permissions on one shopping app, another account still showed that the reporter had joined the product-sharing circle, though clicking in revealed neither the reporter's account nor purchase records.

Not only may purchased products be seen by friends, but products not yet ordered may also become known to friends.

Last year during "Double 11," Ms. Zhu, who works in Guangzhou, Guangdong, found that a friend actually knew what she was planning to buy. Upon asking, she learned that her shopping cart contents had been shared, so she quickly turned off the shopping-cart sharing feature—yet her friend could still see them.

"Only after consulting customer service did I learn that once a shopping cart is shared, it can't be turned off; you can only manually delete the products or cancel sharing of all products so others can't see what you've shared. From now on, if I don't want people to know what I've bought, I can only place the order immediately and can't leave things in the cart waiting for discounts," Ms. Zhu said.

Collecting contacts and other information

Precisely pushing acquaintance services

How do these apps precisely push acquaintance services to users?

An industry insider explained that apps' acquaintance-recommendation features are mainly based on contacts access, user interaction behavior, third-party account binding, and knowledge-graph completion technology. Simply put, they collect user information to build profiles, and then recommend friends based on those profiles. For example, if a user authorizes access to contacts, the platform will match the information it reads with that of other users and display successfully matched users in the "people you may know" list. If a third-party social account is bound, the platform may obtain more user-relationship information through those accounts.

Hu Gang, deputy secretary-general of the Legal Work Committee of the Internet Society of China, pointed out that the terminal devices, networks, and communication numbers used by users all have unique identification codes that can be used by online operators for precise profiling. "For example, if you and another person often use the same Wi-Fi (mobile hotspot), it is highly likely that you will be judged to know each other."

Do these push mechanisms, which make users feel "offended," constitute infringement of relevant rights?

In the view of Rao Wei, senior partner at Beijing Tianchi Juntai Law Firm, when online platforms provide services such as "recommending people you may know" to users, they should respect both the rights of users whose personal information is recommended and the demands of users currently receiving recommendation information.

"Whether the service infringes on a user's personal privacy should depend on whether the service content provided by the platform includes the user's private information that the user 'does not want others to know.' If it does, the platform should clearly seek the user's consent before providing the service, and should clearly inform the user of how to turn off this feature; only after the user clicks agree may the platform recommend the user's relevant information to third parties. If the platform collects a user's personal information without the user's consent and recommends that information to third parties, it is suspected of abusing big-data functions to infringe on the user's personal privacy," Rao said.

He suggested that when users choose a certain service in an app, they may simply want to use that service content, and may not want to follow—or be followed by others regarding—other users' activities such as "who is playing a certain game, what livestream someone is watching, who viewed my profile." In this case, the platform should respect users' right to choose services and should provide necessary options so that users can turn off pop-ups of other users' information and block other users from obtaining their activity information, thereby obtaining a clean service interface.

According to public information, a user sued an app over its "people you may know" feature on grounds of infringing personal information rights and privacy rights. The first-instance judgment found that the short-video app recommended "people you may know" to the plaintiff through the user's phone number and collected the user's geolocation and social information, which infringed on personal information rights; however, because the above information was not private, the app's recommendation of a limited number of "people you may know" did not constitute an intrusion on the user's peace of life, and therefore there was no act of infringing privacy rights.

The reporter's assessment found that even if some platforms can enable a "don't see others and don't be seen by others" feature, in practice pushing and being pushed still occur, and it seems difficult to completely avoid the disturbance of recommendation mechanisms.

In this regard, Yang Zhenyu, a lawyer at Beijing Yinghe (Nanjing) Law Firm, suggested that users can first communicate with the platform and report the problem through the platform's complaint channels. If the platform cannot respond actively, users can complain or report to the national cyberspace administration, which will investigate the platform and, if it verifies that the platform is suspected of infringing on users' personal information, will penalize the platform.

"If the infringement has caused serious consequences, such as property losses or a serious impact on the party's personal life, one can safeguard one's lawful rights and interests by reporting to the public security authorities or filing a criminal private prosecution with the procuratorate. One can also refer to the provisions on public-interest litigation in the Consumer Rights Protection Law and have an appropriate entity file a lawsuit against the app company on behalf of users," Yang said.

Collecting information must be lawful

User consent should be obtained

So, how can app companies address user pain points and respect users' real needs?

Rao Wei suggested that online operators should obtain users' consent before collecting their personal information, and platforms should clearly inform users of the scope and purpose of the information collected. Online operators should strictly keep users' information confidential and establish and improve user-information protection systems; they must not leak, tamper with, or destroy the personal information they collect, and must not provide personal information to others without the consent of the person whose information is collected.

"When providing services, online operators should give users the right to choose regarding the service content. If a user chooses to receive services undisturbed, the platform should not push information the user does not want to see. Only if a user chooses to receive information or activity updates about 'people you may know' may similar services be provided, and only on the condition that the other party's rights are not infringed," Rao said.

In Yang Zhenyu's view, given the professionalism of app development, the best choice is undoubtedly to establish professional and authoritative industry associations and have them guide the development of the app development industry. Industry associations should explicitly require app developers to put user experience first, eliminate excessive collection and use of users' personal information, and resolutely crack down on unfair standard terms—that is, practices requiring users to grant all authorizations at once. In addition to industry self-inspection, they should also cooperate with administrative departments or join with other organizations to supervise and inspect apps; once violations are found, they should announce them to the public and impose corresponding penalties, such as disqualifying them from annual excellence awards or restricting app listings—that is, prohibiting new app listings for a certain period or limiting the number of apps.