Toward a new pragmatics for human-computer interaction
FILE PHOTO Chatbots powered by large language models, such as DeepSeek, exemplify a widely used form of human-computer interaction.
Pragmatics, the study of language use in communication, focuses on the relationship between signs and their interpreters (users). With ongoing technological advances in artificial intelligence (AI) and large language models (LLMs), computers’ capabilities in language use and generation have grown rapidly, raising an important question: how can we address the many practical challenges of language use in human-computer dialogue to meet the growing demand for such interactions? Developing a new human-computer pragmatics could offer promising solutions.
Background of human-computer pragmatics
Today, human-computer dialogue systems are widely applied across sectors such as retail, banking, healthcare, education, research, and customer service. Since Stuart K. Card and others introduced the concept of “human-computer interaction” (HCI) in 1980, theoretical research on human-computer dialogue has evolved beyond technical considerations to embrace social and humanistic dimensions, with communication studies contributing to the field.
Since 2002, when American scholar Ayse Pinar Saygin and Turkish scholar Ilyas Cicekli first conducted a pragmatic analysis of human-computer conversations, to 2024, when Chinese scholars Quan Zhi and Chen Zhiwei introduced the term “human-computer pragmatics,” researchers have concentrated on several key issues: debating the necessity and feasibility of pragmatic research on human-computer dialogue, conducting preliminary analyses of certain pragmatic phenomena, and evaluating the pragmatic capabilities of LLMs.
At present, human-computer pragmatics remains in its infancy, as evidenced by the lack of specificity and systematic focus in research objects, the limited adaptation of classical pragmatic paradigms, and the absence of a coherent theoretical framework.
Pathways for developing human-computer pragmatics
The development of a new human-computer pragmatics can follow a trajectory beginning with empirical research, advancing to theoretical construction, and culminating in application. This emerging field encompasses four major areas of research.
Analysis of pragmatic phenomena and errors in human-computer dialogue: First, researchers can examine the characteristics of pragmatic phenomena in computer-generated discourse (such as meaning, presupposition and givenness, speech acts, reference, coherence, and context), comparing them with human discourse to explore patterns of interaction and mutual adaptation. Second, studies can analyze emotional expression, politeness, humor, rhetoric, power dynamics, and cognitive patterns in computer-generated discourse, again comparing these features to human communication to better understand their similarities and differences. Third, pragmatic errors in these core and macro-level phenomena can be identified, their causes diagnosed, and solutions designed. Researchers can observe human responses, computer self-adjustments, and processes of mutual adaptation.
Comparison of pragmatic performance and capabilities of representative LLMs from China and abroad: First, comparisons between human-computer dialogue and interpersonal dialogue can be drawn by utilizing relevant corpora. Second, researchers can assess representative Chinese and foreign LLMs with respect to core pragmatic phenomena, macro-level pragmatic topics, and discursive and conversational features. Differences can be analyzed within the theoretical frameworks of classical pragmatics, LLM development, and communicative competence, allowing for a systematic evaluation of each model’s strengths and weaknesses.
Theoretical construction of human-computer pragmatics: First, by analyzing how human-computer dialogues adhere to or deviate from classical pragmatic principles, researchers can formulate pragmatic principles specific to human-computer dialogue. Second, findings can be synthesized to establish a theoretical framework for human-computer pragmatics, which can be extended to modalities beyond language and to scenarios beyond everyday conversation.
Application of human-computer pragmatics: First, practical issues encountered by the public in everyday human-computer conversations can be identified, with targeted solutions and user education programs developed. Second, challenges faced by human-computer interface designers can be addressed to assist in the creation of more user-friendly applications. Third, researchers can help identify the needs of regulatory bodies and industry associations to inform policymaking and governance.
Developing human-computer pragmatics holds significant theoretical and practical implications. On one hand, it responds to societal needs and national strategies in the AI era, offering insights to support governments, industries, and the public in advancing the deployment of human-computer interaction technology and enhancing its regulation and governance. On the other hand, it expands classical interpersonal pragmatics into the domain of human-computer interaction, promoting interdisciplinary integration and advancing the development of the “new humanities and social sciences.”
Liu Xingbing is a professor from the School of Business English at Sichuan International Studies University.
Edited by WANG YOURAN