China is taking a first step toward regulating algorithms. How that experiment goes could help Western regulators understand what to embrace—and what to avoid—as they ponder tougher controls on Western social-media giants too.
China launched a sweeping three-year plan last month to regulate the use of algorithms, setting itself up as a potential trailblazer as governments around the world step up regulation of Big Tech. According to draft rules released in August, companies cannot use algorithms which lead to addiction or excessive spending. Users should also have the right to opt out.
The broad-based regulations, if implemented strictly, could fundamentally shake up the business models of many successful internet companies. For example ByteDance, the owner of TikTok, has succeeded largely by recommending catchy content with the help of its powerful algorithm.
To be sure, some aspects of China’s proposals are clearly targeted at maintaining government control. Guidelines from the internet watchdog say algorithms should uphold core socialist values and promote positive energy. Democratic societies are unlikely to accept such strictures, and even more benign rules would likely face court challenges.
But watching how China’s moves work out—and how large any collateral economic damage ultimately is—could still prove useful to other countries which are also grappling with the enormous societal impact of internet companies. The European Union proposed a bill in April to regulate artificial intelligence systems in some so-called high-risk uses like critical infrastructure, college admissions and loan applications. In the U.S., Congress recently conducted a hearing on Facebook after The Wall Street Journal’s investigations into the social-media giant.
The biggest problem for regulating algorithms is how opaque they are. That’s becoming a bigger issue as more decisions are made by machines which learn through crunching a vast amount of data. It’s not easy, sometimes even for the creators of algorithms, to pinpoint the exact reason why an artificial intelligence makes a particular decision. Biases embedded in the training data could unknowingly seep into the decision-making process. And algorithms can also narrowly focus on some objectives, like amplifying viral content, without considering other impacts. Moreover, they are also continually updating, which makes regulation even harder.
That’s a big challenge even for China, which has more powerful tools at its disposal. Another problem is how to make the algorithmic process more transparent and accountable, without taking too broad a brush that could stifle all innovation, especially at smaller companies. That is one big risk Beijing takes by being a first-mover—it may reap immediate benefits, as it sees it, in terms of enhanced social control and fewer ugly side effects like addiction and indebtedness. But it may also squash the potential emergence of any new Bytedances in the process.
Algorithms have become an integral part of everyday life. Regulations may finally need to catch up—but how to go about it remains a difficult puzzle. Investors in U.S. internet companies, and their detractors, should both be watching China’s experience closely.
Write to Jacky Wong at jacky.wong@wsj.com
Copyright ©2021 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8
