China’s first AI-themed dance drama, “Deeply In Love With You,” was staged at the National Center for the Performing Arts in Beijing on Aug 26th, 2023. Photo: Jia Tianyong/CNSphoto

---

Artificial intelligence is an area of inquiry not only for natural sciences and engineering, but also for the humanities and social sciences. In China, the application of integrated human-machine intelligent systems in the humanities and social sciences was predicted in the 1980s. This vision is now becoming increasingly tangible. 

Characteristics of generative AI 

Today, AI applications have remarkable ability to permeate, diffuse, and disrupt, raising concerns over runaway technologies and social disorder. Understanding the main characteristics of the latest generation of AI can help get to the core of the problem. 

First, in addition to emulation and prediction based on existing data, generative AI can learn from this data to generate new content, thus manifesting a certain level of creativity. 

Second, generative AI built upon large language models are to some extent able to give human-like responses to prompts or questions submitted by humans in natural language. This is one of ChatGPT’s most impressive features. 

Third, breakthroughs in functionality are not merely a result of data volume and model parameters reaching a certain scale. Reinforcement learning from human feedback (RLHF) and human-machine intelligent interaction during use can further improve functionality and output. 

Fourth, generative AI has ushered in the automated knowledge production mode of “large model + human feedback” marked by higher speed, lower costs, broader scope, and significantly enhanced efficiency of knowledge production as compared with traditional knowledge production methods. 

Fifth, while generative AI has the basic ability to assess the correctness, accuracy, and value of the content it generates, human information literacy, knowledge systems, and critical thinking abilities are still needed. Both the input of data, such as feeding and training, and the output of data, such as judgment and utilization, now require human capabilities that differ from those required in the past. 

AI-driven research 

Recent trends suggest that the dynamics of cutting-edge science and technology largely lead and even determine the generation and movement of research focuses in the humanities and social sciences. The accelerating integration of socio-economic development and emerging technological advancement will gradually expand the range of AI application from natural sciences and engineering to the humanities and social sciences, giving rise to AI-driven humanities and social science research. This new mode of knowledge production is conducive to liberating productive research forces and unleashing innovation potential. 

First, the barriers to knowledge access are further reduced. Equipped with AI’s powerful information processing capabilities such as search and classification, researchers can access massive quantities of knowledge anytime and anywhere. Second, the overall knowledge supply is enriched. Researchers can acquire more data once hard to obtain, thus breaking through silos between different fields and disciplines. Third, preliminary processing of content by AI can free researchers of some repetitive and basic-level tasks in knowledge production. Fourth, AI can provide supplementary information relevant to the research subject, enabling the diversification of research plans. Fifth, AI can assist researchers in writing routine texts, such as literature reviews, research proposals, article abstracts, and syllabuses. 

AI-driven humanities and social science research can be integrated into education and talent development. AI can assist teachers in generating teaching material such as syllabuses and assignments, and answering student questions. When used properly, AI can help students develop critical and innovative thinking skills, as well as independent thinking and problem-solving abilities. 

AI can facilitate the development of new research methods and ideas, contributing to the formation of emerging disciplines and interdisciplinary fields, while also helping foundational, critical, and fringe disciplines find new growth engines. In terms of knowledge production process, AI not only allows for faster and more efficient management, dissemination, and application of knowledge, but also provides a more convenient technological environment for knowledge dissemination, which makes knowledge more open. 

Humanities and social sciences research in the age of AI calls for infrastructure that can further promote academic innovation. Existing digital infrastructure in the humanities and social sciences should be fully integrated and utilized. Adherence to the principles of fairness and inclusiveness is essential. Innovation platforms should actively serve researchers. 

Traditional databases need to be optimized to store larger volumes and a greater diversity of literature, so as to make information retrieval more convenient. Platforms should improve the user-friendliness of user interfaces and experience, while also delivering personalized and differentiated services that cater to the individual needs of each researcher. 

It is necessary to establish multi-stakeholder consultation mechanisms, sharing mechanisms that end digital monopolies, and interactive mechanisms aimed at increasing user engagement. Ethical risks posed by new technologies should be addressed promptly, which entails clarifying and adhering to certain ethical principles. 

In summary, the involvement and support of AI can help humanities and social science researchers gain a deeper understanding of the nature of the human mind and its behavior, and explore the destiny of human civilization in an increasingly technology-oriented future. 

Lei Huanjie is a research fellow in the Institute of philosophy at CASS. 

 Edited by WANG YOURAN