How Chinese Communication Scholars Understand Human-machine Relationship?

In 2015, the first news story written by Dreamwriter, an automated writing system developed by Tencent, was released on the Finance Channel of the Tencent website(J. Liu, 2017; Y. Liu, 2015). Around the same time, Xinhua News Agency, a leading Chinese media organization, formally deployed a robot reporter, known as “Kuaibixiaoxin” (快笔小新) to produce news. One year later, another robot reporter,“Xiaomingbot”(张小明), was launched by an influential news platform “Jinritoutiao” (今日头条) and was used to cover the Rio 2016 Summer Olympics similar to a human journalist. Following that, Jinritoutiao developed a data-driven media tool, known as “Meitishiyanshi” (媒体实验室) primarily to support editorial decision-making through seeking news topics, analyzing audiences, and visualizing data.

In this short amount of time, the advantages of machine-assisted content production on efficiency, scalability and translation accuracy, have drawn increasing attention from Chinese communication scholars. A giant wave of artificial intelligence (AI) - powered applications have been adopted into communication processes, particularly in content production and interaction, such as robot reporters, ChatGPT, and writing algorithms. Reviewing existing literature, researchers commonly understand these different forms of non-human intelligent agents as machines  (Mou, 2018). By examining their benefits and challenges in content production and interaction, a conceptual framework of understanding human-machine relationship has been well established in academic discourse.

Human guidance is important

Some scholars have articulated an ideal picture of journalism in the future which is grounded in effective collaboration between humans and machines.

Most Chinese scholars emphasize that rather than replacing humans, using machines could expand the boundaries of human capabilities. It has been demonstrated that under human guidance, machines can be used to collect and analyze information and patterns that are difficult to detect or even invisible to humans, find clues for journalists, assist in editorial planning, and portray target consumer profiles for publishers (Peng, 2016). When algorithms are used to draft stories, journalists may spend more time refining content. Although the integration of machines into the field of journalism has caused concerns of jobless, scholars emphasize that human resources could be shifted from elementary tasks to complicated ones that need more innovative ideas and in-depth thoughts, such as judging, analyzing, interpreting, forecasting, and polishing content (Deng, 2016; Yu, 2018). In this sense, the application of machines may be seen as pushing media practitioners toward a higher level of professional growth.  

However, it must be noted that all journalism practices outlined above are under human guidance, as highlighted in literature. Scholars constantly stress the dominant role of humans in human-machine collaboration, mainly due to technological limitations. For example, robot reporters only can produce scalable news stories based on text templates. Machine-generated stories are often boring, not readable enough to satisfy audiences. Writing algorithms may accelerate the circulation of fake news. Training data for machines is important, as algorithms tend to learn, reproduce and even amplify the biased patterns in training data. Biased output may negatively influence the decision-making in real world. To address these challenges, guidance and supervision from humans as algorithm designers and machine developers is essential. Another reason highlighted by scholars is that machines cannot generate essential elements of communication, such as empathy, care, trust, and mutual understanding that emerge in interactions (Wang, 2018).

By illustrating machines’ disadvantages, Chinese communication scholars attempted to highlight the indispensable role of humans in guiding practices of producing content by using machines.  

Machine: not intelligent

Astonished by consecutive successes of Alpha Go against human Go players between 2015 and 2017, Chinese communication scholars started exploring differences between human intelligence and machine intelligence. For Alpha Go cases, Yu (2017) stressed that although machines demonstrated excellence in processing standardized rule-based tasks, they would become incapable when the rules changed. By contrast, humans can act as rule designers who can change the rules for machines. 

However, the increasing adoption of intelligent machines like Alpha Go and the rise of ChatGPT in 2022 has led scholars to worry that humans may over rely on AI algorithms to make decisions (Zhang, 2018). As a result, machine intelligence may play a dominant role in decision-making processes (Liu, 2023). 

Interestingly, the majority of scholars regard this risk to be a result of lacking comprehensive understanding of human-machine relationship. Scholars emphasize that machine intelligence is derived from learning human information, knowledge, and patterns of thinking. Likewise, Generative AI, such as ChatGPT, does not create any new knowledge in its own, but rather rearranges existing human knowledge and reproduces them in creative visualized formats (Zhu, 2023; Liu, 2023). Essentially, AI generated content is a recombination of existing knowledge and is heavily dependent on the way a human questions, as well as information, corpus and behavior patterns provided by humans (Peng, 2023; Hu, 2023). Even in the existing copyright legal discourses, machines are considered as assisting tools, rather than authors with creativity. All works generated by AI only represent the will of AI designers or trainers (Xiong, 2017). 

By stressing the fact that content products by machines or other AI-powered tools are results of learning from human society, communication scholars aimed to highlight that only human can decide how intelligent that machines can be in the future.

Rethinking machines’ contribution 

Despite fears of job loses owing to a giant wave of AI applications, communication scholars note that the creation of new job opportunities in journalism industry, for example the Associated Press appointed an automated news editor specifically responsible for identifying workflows suitable for automation (Xu, 2017). 

Another interesting insight drawn from Chinese communication scholars is that machines can be used to better understand human society. This view is supported by the fact that machines, when functioning as simulators, enable researchers to examine how human society operates by modelling real society or constructing a “pseudo-environment” which represents human activities in a comprehensive manner (Peng, 2024).

Therefore, in the area of communication, AI-powered tools can be used to simulate and analyze public opinion to better inform decision-making. 
Communication scholars also perceive machine as a “mirror” or the “other” in social interaction. The notions of “mirror” and the “other” originated primarily in philosophy and were later developed in social psychology and sociology. According to Cooley (2017), as social beings we live with our eyes upon our reflection. Applying Cooley’s core idea to “presentation of self”, Goffman (1959) indicated a mirroring process that when interacting with other humans, individuals are likely to perform identities based on anticipated audiences’ reactions. 

In the age of AI, interacting with intelligent machines becomes common. Communication scholars extended Cooley’s idea by arguing that individual performance might be adjusted in response to machine-generated feedback, and in this sense, machines have begun functioning as self-perception mirrors (Peng, 2024; Liu, 2023). In other words, machines, rather than traditional humans, may become the “other” in interactive contexts, and individuals might perceive, reflect upon, and construct their sense of self through engagement with machines. This finding indicates that AI-mediated interaction may offer a new pathway for self-reflection, although awareness is needed to avoid excessive reliance on algorithms.

Conclusion

To sum up, existing literature indicates that Chinese communication scholars adopt a human-centered approach in understanding human-machine relationship. On one hand, human’s dominant roles, such as rule making and meaning creating, are consistently emphasized in literature. In this sense, human area is considered as a fundamental normative baseline while machines are recognized as assistants that are expected to operate within a human-defined world. On the other hand, machines are acknowledged to contribute to a better human society by expanding human capacities and offering a new pathway to self-understanding. It is clear that Chinese communication scholars do not conceptualize human-machine relationship merely as a technical issue, but a social problem involving multiple actors, forms of agency and responsibilities. 

Existing research by Chinese communication scholars does not suggest that we should simply treat machines as tools. Rather, it calls for enhanced understanding of respective responsibilities of humans and machines in communication activities, which is crucial for establishing an explainable, accountable and transparent human-machine collaboration framework in the future. 

References

Cooley, C. H. (2017). Human nature and the social order. Routledge.
Goffman, E. (1949). Presentation of self in everyday life. American Journal of Sociology, 55(1), 6-7.
Liu, J. (2017, April 18). 写稿机器人DreamWriter，究竟是如何写出一篇新闻稿件的? Yicai. https://www.yicai.com/news/5268559.html 
Liu, Y. (2015, September 10). 腾讯发布首篇机器人撰写的新闻称不会抢记者饭碗. China News Service. https://www.chinanews.com.cn/cj/2015/09-10/7517183.shtml 
Peng, Lan 彭兰 (2016) 智能化:未来媒体浪潮——新媒体发展趋势报告(2016) [Smart media: Future trend of media—New media development trend report (2016)].国际新闻界 [Journal of International Communication] 32(1): 6–24.
Peng, Lan 彭兰 (2023) 从ChatGPT透视智能传播于人机关系全景及前景 [The Panorama and Prospect of Intelligent Communication and Human-Machine Relationship(2023)].新闻大学 [Journalism Research] 204(4): 2–16. http://doi.org/10.20050/j.cnki.xwdx.2023.04.005 
Liu, Hailong 刘海龙 (2023) 新常人统治的来临：ChatGPT与传播研究 [The Panorama and Prospect of Intelligent Communication and Human-Machine Relationship(2023)].新闻记者 [Shanghai Journalism Review] 484(6): 11–20. http://doi.org/10.16057/j.cnki.31-1171/g2.2023.06.008 
Mou, Yi，Xu, Kun 牟怡 许坤 (2018) 什么是人机传播——一个新兴传播学领域之国际视域考察[].江淮论坛[Jiang—Huai Tribune] 2: 149-154. http://doi.org/10.16064/j.cnki.cn34-1003/g0.2018.02.023 
Xu, Xiangdong, Guo, Mengmeng 许向东 郭萌萌 (2017) 智媒时代的新闻生产:自动化新闻的实践与参考 [News production in the age of smart media: Practice and thinking of automated journalism].国际新闻界 [Journal of International Communication] 32(1): 29–47. http://doi.org/10.13495/j.cnki.cjjc.2017.05.002 
Yu, Guoming, Yao, Fei 喻国明 姚飞 (2017) 试论人工智能技术范式下的传媒变革与发展——一种对传媒未来技术创新逻辑的探析 [A preliminary discussion of media transformation and development in the paradigm of artificial intelligence: An exploration of the logic of future technological innovation in media].新闻界 [Press Circles] 1:39-43. http://doi.org/10.15897/j.cnki.cn51-1046/g2.2017.01.005 
Zhang, Chao 张超 (2018) 新闻生产中的算法风险：成因、类型与对策 [Algorithmic risks in news production: Causes, types and countermeasures].国际新闻界 [China Publishing Journal] 13: 38–42.
Hu, Zhengrong 胡正荣 (2023) 颠覆与重构：AIGC的效用危机与媒介生态格局对比 [Disruption and reconstruction: The utility crisis of AIGC and the comparison of media ecological patterns].新闻与写作 [News and Writing] 8: 48–55.
Xiong, Qi 熊琦 (2017) 人工智能生成内容的著作权认定 [The copyright for content generated by artificial intelligence].知识产权 [Intellectual Property] 3: 3–8.