Social media is flooded with repetitive but largely meaningless content. Behind it is a massive industry.
Photo from CFP
By SHE Xiaochen
YU Meng makes ads for a living, but she feels more like an assembly line worker than an artist. Her job is to find out what's popular on Douyin and make something that’s almost exactly the same. After that, she checks a few boxes on who she wants to send the ads to ‒ age, location, and sometimes occupation ‒ and blasts it out to them. Every day, she has to make seven ads to meet her quota. Every week, her team is evaluated on click-throughs, playtime, and conversions. Her employer, a fast-growing ed-tech company, is always hungry for more clicks, so the content mill keeps on churning.
“I would say that around 99 percent of what I make is garbage,” said Yu. “It’s quite okay if my videos have exactly the same plot and almost the same script as many others. Douyin is flooded with them. There is a massive industry that feeds the internet this garbage, some of it highly specialized garbage.”
LI Hui works at a PR firm and his specialty is Weibo. When a client’s hashtag needs a bump, he comes up with a few dozen “content templates” in less than two hours and assigns them to hired trolls. (Influencers are also up for hire but charge more.) His performance is measured by the number of likes, comments, and reposts, as well as whether the bumped hashtags appear on the trending list. He is blasé about hiring trolls and can spot troll-generated content at a glance.
Big tech has shown a high level of tolerance, or even encouragement, to such content. After all, they make money from clicks and attention. Jiang Lingling, who writes promotional articles for a smartphone company, said her job is to “scour the internet for juicy stories” and rewrite them. The racier the plot, the more clicks it gets.
An operations manager at Toutiao, a Bytedance-owned news app, acknowledged that repetitive and largely meaningless content has a cumulative value. Toutiao doesn’t have its own editorial agenda. It makes recommendations purely based on traffic The algorithm will pick up an article even if it’s merely a soup of trending words and lacks cogent sentences, as long as enough people are reading about the associated topic. The hope is that the wisdom of the crowd will eventually prevail. Toutiao has 280 million users: Chances are, they will sort things out.
But many industry insiders worry that relying on traffic alone will backfire. A former editor at a news app that has already shut down said even to the last day, the debate on whether to recommend racy stories to lower-tier cities was never settled. Other platforms are trying other approaches. Toutiao, for example, has started to make curated content.
By the definition of big tech, none of this gossip or these annoying ads are actually spam. Spam is pretty much confined to hate speech, violence, or manipulating stock prices. Such posts are removed as a matter of course. That is Wei Ning’s job. His team trains algorithms to identify spammers and block their accounts automatically.
“Offending accounts share common traits in user-profiles and activities,” Wei said. Although blocked users can always register a new account and start posting again, many will find it too costly to beat the algorithm and give up.”
This doesn’t mean human moderators will be gone any time soon. About 3 percent of algorithm-flagged posts need to be reviewed by humans. And then there is the arms race. Spammers adapt new tactics as quickly as the algorithms root out the old ones