At the end of 2022, the emergence of ChatGPT created a ripple effect in the scientific and technological community and ignited widespread interest and debates in the field of higher education. As a representative tool of generative artificial intelligence (GenAI), ChatGPT has rapidly become a focal point for educational researchers and practitioners to explore future educational models. Its potential applications and profound impact have turned it into a hot topic in academics.

Proponents argue that GenAI has the potential to drive transformative changes in higher education. It can not only enable personalized learning paths and foster the innovative upgrading of educational evaluation systems but also enhance the efficiency of teaching and research processes. More importantly, it has the capacity to stimulate students’ critical thinking and creativity (Rudolph et al., 2023; Ruiz-Rojas et al., 2024; Urban et al., 2024). Several world-class universities, such as Harvard University, the California Institute of Technology, and the University of Michigan, have already begun to explore the integration of GenAI tools into daily teaching and research activities. However, many scholars have raised concerns about the widespread adoption of GenAI in education. This includes the risks of disseminating misleading information, fostering students’ over-reliance on technology, compromising data privacy and security, and exacerbating the digital divide (Hasanein & Sobaih, 2023; Kasneci et al., 2023). Such concerns have led some universities to adopt a cautious approach, with some even explicitly prohibiting the use of GenAI by teachers and students in specific contexts.

The question of whether to adopt an “open embrace” or enforce “strict restrictions” regarding GenAI tools is at the core of these debates. While existing literature predominantly focuses on theoretical perspectives and small-scale empirical studies, there remains a significant gap in large-scale, systematic research to understand the college students’ acceptance, actual usage, and expectations of GenAI tools in higher education.

As GenAI becomes increasingly integrated into higher education, understanding college students’ acceptance, usage behaviors, and expectations towards GenAI tools is critical. Such analysis not only provides valuable empirical support for educators and institutions to develop scientific, practical guidelines for GenAI usage guidelines and foster students’ artificial intelligence literacy but also informs policymakers in crafting balanced and effective regulations.

To address this gap, this study conducted a large-scale survey involving 12,678 students from 20 universities across China. Through in-depth empirical research, this study aims to answer the following research questions:

(1) What is the level of acceptance of GenAI tools among college students?

(2) How do college students use GenAI tools?

(3) What are college students’ expectations for the future development and application of GenAI tools?

Although several studies have explored students’ attitudes towards GenAI and its usage, previous research has notable shortcomings: First, most prior studies are based on small sample sizes and lack large-scale national survey data to provide comprehensive evidence. Second, these studies often fail to account for the significant variations in technology acceptance among students from universities of different levels, resulting in limited representativeness in sample selection. Third, existing research on students’ attitudes towards GenAI tools predominantly focuses on foreign universities, with little empirical investigation targeting Chinese college students. Moreover, most studies have emphasized whether students use GenAI tools and the contexts in which they are used, but they lack a detailed understanding of kay aspects such as time and financial investment. These factors are crucial for measuring students’ usage behaviors more precisely. Finally, while some studies have examined the influence of factors like gender, grade, and major on GenAI usage, other critical factors-such as family economic background and students' academic performance-have been overlooked, leaving gaps in understanding the broader determinants of GenAI adoption.

The dataset for this study was derived from a nationwide survey conducted by the research team in June 2024. The survey encompassed 20 higher education institutions across the country, including 11 elite universities and 9 general undergraduate institutions, selected through a non-proportional stratified sampling strategy. Following rigorous data cleaning to eliminate invalid or low-quality responses, this study retained a total of 12,678 valid responses from the 20 universities. This study examines students’ acceptance of GenAI tools, their actual usage behaviors, and their expectations for future applications. To assess college students’ attitudes towards GenAI tools, this study adopts the Technology Acceptance Model (TAM) framework, analyzing six core dimensions: Perceived Usefulness, Perceived Ease of Use, Perceived Credibility, Perceived Social Influence, Personal Motivation, and Overall Attitude. A four-point Likert scale is employed to measure college students’ acceptance levels, combining their responses to both affirmative and negative questions regarding their usage experiences. Regarding the analysis of usage behaviors, this study begins by exploring students’ usage frequency and associated expenses to establish a baseline understanding of their interactions with GenAI tools. To further analyze specific usage scenarios, the study draws on prior literature and insights from in-depth interviews, categorizing GenAI application scenarios for college students into seven categories: idea generation, plan design, literature work, programming tasks, data analysis, thesis writing, and text translation. The likelihood and preferences for using GenAI tools across different scenarios are assessed using a six-point Likert scale. Regarding college students’ expectations for the future development of GenAI tools, a four-point Likert scale is also used to investigate two core issues: (1) whether teachers and students should actively learn and adopt GenAI tools, (2) whether schools should establish regulations to govern the use of GenAI tools by teachers and students This study employs both the Logistic regression model and the Ordinary Least Squares (OLS) linear regression model to estimate the relationship between college student characteristics and their usage behavior of GenAI tools. The models are as follows: Yi = β0+β1* Xi +η+εi When Yi represents “usage or not,” which is an ordinal categorical variable, the ordered Logistic regression model is applied; when Yi represents “usage duration” and “usage cost,” which are continuous variables, the OLS regression model is used. The core independent variables Xi include factors such as students’ gender, grade level, major, family background, school information, and GPA ranking. εi is the random error term.

Students exhibit a high level of acceptance for GenAI tools but have concerns about data security and personal skill development. Firstly, students generally perceive GenAI tools as effective and efficient aids for completing tasks. However, many students are also acutely aware of the associated risks. A significant number expressed concerns about data security and privacy protection, alongside apprehensions about the potential negative impact of GenAI tools on their personal ability development. Students primarily use GenAI tools for text processing tasks. In this study’s sample, the average daily interaction time with GenAI tools was 1.22 hours, with students mainly engaging in activities such as text translation, literature processing, and thesis writing. While these tools help improve writing efficiency, their use as proxies for tasks like thesis writing presents new challenges to academic integrity and critical thinking skills. There are significant differences in the usage of GenAI tools among students with varying academic backgrounds. First, compared to students from general universities, those from elite universities demonstrate longer engagement duration and incur higher economic cost in using GenAI tools. In addition, senior students invest more time and money in GenAI tools compared to freshmen. In terms of majors, engineering students spend more time and incur greater costs in using GenAI tools compared to students from other disciplines. Lastly, students with stronger academic foundations also tend to dedicate more time to engaging with GenAI tools. There are significant variations in GenAI tools usage among students with different backgrounds. Firstly, male students, on average, exhibit a stronger tendency to use GenAI tools, both in terms of time and expenditure. Furthermore, students from families based in provincial capitals, municipalities, as well as those whose fathers have higher levels of education, are more likely to invest time and financial resources in GenAI tools.

Hasanein, A. M., & Sobaih, A. E. E. (2023). Drivers and Consequences of ChatGPT Use in Higher Education: Key Stakeholder Perspectives. European Journal of Investigation in Health, Psychology and Education, 13(11), Article 11. https://doi.org/10.3390/ejihpe13110181 Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., … Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. https://doi.org/10.1016/j.lindif.2023.102274 Rudolph, J., Tan, S., & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of traditional assessments in higher education? Journal of Applied Learning and Teaching, 6(1), Article 1. https://doi.org/10.37074/jalt.2023.6.1.9 Ruiz-Rojas, L. I., Salvador-Ullauri, L., & Acosta-Vargas, P. (2024). Collaborative Working and Critical Thinking: Adoption of Generative Artificial Intelligence Tools in Higher Education. Sustainability, 16(13), Article 13. https://doi.org/10.3390/su16135367 Urban, M., Děchtěrenko, F., Lukavský, J., Hrabalová, V., Svacha, F., Brom, C., & Urban, K. (2024). ChatGPT improves creative problem-solving performance in university students: An experimental study. Computers & Education, 215, 105031. https://doi.org/10.1016/j.compedu.2024.105031