For Chinese journalists, the battle to control AI is uphill

In 2019, my research partners and I began tracking popular news articles about artificial intelligence on Chinese social media to answer two questions: What aspects of AI and algorithmic systems have been on the radar of Chinese journalists, and what journalistic techniques do they use in their reporting?

After four years of research and analysis of hundreds of articles on WeChat, China’s most widely used social media app, we found mixed results. On the positive side, we found that the rise of AI has opened up space for critical journalism in China, investigating the broader impact of technology on work and life. On the other hand, we found very limited use of advanced techniques to explain to readers the power, errors, and biases of AI algorithms—what researchers call “algorithmic accountability reporting.”

Instead of resorting to advanced techniques, such as reverse-engineering an AI image generator by trying different inputs to understand how it might be discriminatory, Chinese journalists rely mainly on traditional reporting methods, including field research, interviews and expert testimony.

The trend stands in stark contrast to China’s major media companies’ ambitious commitments to applying AI in newsrooms, a trend that dates back to the 2010s and has accelerated with the increased focus on large language models since 2022. Newsrooms across the country are racing to introduce “virtual anchors,” automate news editing, and experiment with creating videos using generative AI. As traditional media outlets around the world struggle to stay relevant, these efforts aim to demonstrate that journalism is still worth investing in.

While artificial intelligence is being touted as potentially revolutionary for the news industry, this doesn’t seem to have translated into advanced, complex reporting on the technology itself. How can this gap be explained?

First, we found clues in the academic and professional backgrounds of the journalists. Among the 23 high-profile AI-related investigations we analyzed, the majority of the journalists behind these stories have a background in social or economic journalism. While they may have experience covering social issues such as workers’ and consumers’ rights, they likely lack the resources and expertise to pursue algorithmic accountability reporting involving more advanced techniques.

Industry insiders told me that the biggest barrier to hiring journalists who are experts in using advanced techniques is a lack of resources in newsrooms. Talent is hard to find and expensive. More and more resources are being devoted to opening up new channels for content distribution, from app development to social media, rather than producing original reporting.

As a result, most newsrooms remain labor-intensive places, with reporters and editors working in much the same way they did a decade ago. In addition to their already heavy workloads, Chinese journalists are increasingly being asked to help with the business side of their news operations, such as finding institutional clients.

To be clear, I am not telling Chinese journalists how to do their jobs. Some of them have accomplished remarkable feats using traditional reporting methods, and in some cases, tech giants have tweaked their algorithms following high-profile investigations. However, as AI becomes more prevalent in our daily lives, it becomes increasingly necessary to examine these systems more critically.

Without the ability to conduct in-depth investigations into AI, Chinese journalists face imperfect choices: rely on outside experts, who can provide technical knowledge but may not have information about specific AI systems, or on the tech companies themselves, which rarely give journalists access to proprietary information or their engineers. According to journalist Lai Youxuan, most systems engineers at food delivery giants Meituan and Ele.me rejected her interview requests for what later became her groundbreaking investigation into food delivery platform algorithms in 2020, citing “corporate confidentiality.”

This undermines the chances of Chinese journalists holding AI systems to account, and the chances only get worse as AI systems become more complex. To improve their chances, Chinese newsrooms should ideally allocate more resources to training staff in advanced techniques, hire more staff with technical reporting backgrounds, and encourage exchanges between journalists and AI experts. We have seen growing interest in organizing such training and exchange programs.

However, in a context where many Chinese newsrooms are struggling financially, it might be more realistic to start by training all journalists in the basics of AI: what is AI? How is it developed? What problems arise at each stage of its development? Mastering AI in newsrooms is important because journalism itself is not immune to AI-induced bias. For example, social media algorithms can present search results tailored to a journalist’s user profile, which increases confirmation bias. As AI adoption in newsrooms grows, clear guidelines on the ethical use of AI throughout the content creation chain, from newsgathering to publication, need to be established.

With greater overall AI literacy, journalists will not only be better positioned to conduct critical journalistic investigations into AI algorithms, but will also be more likely to adopt AI technologies into their work more productively and, more importantly, more responsibly.

Ultimately, a newsroom’s most valuable asset is its people, not its technology. Just as AI in the broader sense should always be beholden to humans, not the other way around, newsroom leaders must also remember that AI-powered reporting tools and advanced reporting techniques, as important as they are, are only worth investing in if they are used by high-quality journalists.

Ji Xiaolu, a postgraduate student in media studies at the University of Amsterdam, contributed equally to this article.

Editor-in-Chief: Vincent Chow; pPortrait artist: Wang Zhenhao.

(Header image: Visuals by VCG, reissued by Sixth Tone)