Despite recognizing AI ethics as important for student development, Chinese secondary school teachers rarely implement it in classrooms due to overwhelming barriers including insufficient content knowledge, inadequate professional development, resource scarcity, and social norms prioritizing technical skills over ethical reasoning.
Objective: This study investigates the factors influencing secondary information technology teachers' intentions to teach AI ethics within China's national AI curriculum reform context, specifically examining what prevents teachers from implementing mandated AI ethics education despite policy requirements. The research seeks to understand the key factors shaping teachers' beliefs about teaching AI ethics and how these beliefs interact to influence their classroom intentions and practices within the framework of China's Vision 2040 educational goals.
Methods: The study employed a qualitative systematic approach grounded in the Theory of Planned Behaviour (TPB), conducting semi-structured interviews with 14 in-service secondary IT teachers in China's Greater Bay Area (Guangzhou, Zhuhai, and Shenzhen) between September 2024 and February 2025. Participants were purposefully sampled to include diverse professional roles, from general IT teachers to Expert Teachers (Mingshi), curriculum coordinators, and school principals, all with experience teaching AI-related content. Interviews ranged from 12 to 58 minutes and were analyzed using a hybrid qualitative analysis approach combining inductive and deductive coding. The researchers utilized AI-assisted analysis with the DeepSeek-R1 model and ATLAS.ti software, achieving strong inter-coder reliability (κ=0.861 for deductive codes; κ=0.792 for inductive codes). The TPB framework guided the identification of behavioural beliefs, normative beliefs, and control beliefs that shape teachers' attitudes, subjective norms, and perceived behavioural control regarding AI ethics instruction.
Key Findings:
The study identified critical factors across three belief categories:
Behavioural Beliefs: Teachers recognized significant benefits of teaching AI ethics, including fostering critical thinking, preparing students for technology-saturated futures, and aligning with personal moral values. However, they perceived substantial challenges: feeling overwhelmed by AI's rapid evolution, difficulty keeping content current, struggles to adapt Western-centric ethical frameworks to Chinese sociopolitical contexts, and concerns about crossing ideological boundaries. Some teachers dismissed certain ethical issues (like algorithmic bias) as irrelevant to China, revealing a cultural exceptionalism that risks overlooking local forms of discrimination.
Normative Beliefs: A significant policy-practice gap emerged. While teachers were aware of the 2022 national curriculum standards mandating AI ethics education, they found guidance too abstract and lacking locally relevant case studies. School leadership expectations prioritized exam-oriented subjects and technical competition wins over ethical literacy, creating social environments unsupportive of ethics instruction. Peer collaboration on AI ethics was minimal. Student interest in ethical issues (privacy, deepfakes, data security) motivated some teachers, but ineffective pedagogical approaches (direct lectures) often failed to engage students. Confusion about integrating emerging technologies like Generative AI into ethics curriculum was widespread, even among curriculum coordinators.
Control Beliefs: This emerged as the most pronounced barrier. Teachers lacked fundamental content knowledge in AI ethics—most had educational technology backgrounds without AI-related coursework, and even those with computer science degrees received no ethics training in college. Professional development programs focused almost exclusively on technical skills rather than pedagogical strategies for ethical discussions, leading to teacher disengagement. Significant resource disparities existed between tech hub cities (Shenzhen with corporate-supported materials) and other areas. Classroom constraints (45-minute periods, 30-40 student classes) made in-depth ethical discussions logistically prohibitive. Teachers feared ethical discussions might devolve into chaos or breach politically sensitive topics. Textbooks mentioned ethics only briefly, and AI educational platforms provided no ethics-related resources.
These factors interacted to shape teachers' intentions: those with positive attitudes and higher perceived control showed stronger implementation intentions, while most teachers maintained neutral or negative attitudes due to low confidence. Social norms prioritizing technical aspects and low perceived behavioural control predominantly constrained implementation, resulting in AI ethics receiving only 5.1% of classroom instructional time despite policy mandates.
Implications: This research makes several important contributions to AI ethics education. It provides the first systematic examination of teachers' beliefs about AI ethics instruction within a top-down curriculum reform context, revealing that policy mandates alone are insufficient without addressing implementation barriers. The findings demonstrate that current professional development models emphasizing technical skills are inadequate for supporting ethics education, which requires fundamentally different pedagogical approaches centered on facilitating ethical reasoning and discussion rather than transmitting technical knowledge. The study highlights the critical need for culturally situated pedagogical resources that adapt global ethical principles to local Chinese contexts while addressing genuine local ethical concerns (gender, age, regional discrimination) rather than dismissing ethics as Western-centric. The research reveals systemic inequities in resource access between regions and identifies the tension between exam-oriented educational priorities and ethics instruction. Successfully implemented practices, such as integrating AI ethics into Moral and Political Education classes and leveraging student curiosity about real-world AI issues (deepfakes, privacy), provide actionable models for adaptation. The study exposes how prioritizing technical skills and coding creates an "exclusionary pedagogy" that becomes critically inadequate in the generative AI era, where ethical reasoning is increasingly essential.
Limitations: The study acknowledges several limitations. The relatively small sample size of 14 teachers, while achieving thematic saturation, was limited by the specific requirement for IT teachers with AI teaching experience in the Greater Bay Area. The study captured the educational context before a November 2024 policy document that further emphasized AI ethics, meaning findings reflect earlier implementation conditions. The reliance on self-reported interview data from a specific geographic region (Greater Bay Area) may limit generalizability to other Chinese contexts with different resources and development levels. The study did not include direct classroom observations to validate teachers' reported practices, relying instead on their accounts of teaching behaviors and intentions.
Future Directions: The researchers propose several critical directions for future work. There is an urgent need to redesign professional development programs away from one-off technical training toward sustained collaborative professional learning communities where teachers develop AI ethics competencies through group moral reasoning engaging with core ethical principles. This would enable teachers to reflect on pedagogical designs for creating active learning activities rather than lectures. Future research should explore developing culturally situated pedagogical resources and case studies relevant to Chinese contexts that address local ethical concerns. Investigating effective pedagogical approaches for facilitating ethical discussions in large classrooms with time constraints is essential, as is studying how to integrate ethics across different subject areas beyond IT classes. Research should examine the long-term impacts of different professional development models on teacher confidence and classroom implementation. Studies could investigate how to address regional resource inequities and support teachers in less-resourced areas. There is also a need to develop appropriate assessment methods for AI ethics learning that could potentially influence school leadership priorities. Finally, research should explore how emerging technologies like Generative AI can be incorporated into ethics curriculum in ways that help students develop critical evaluation skills rather than simply technical proficiency.
Title and Authors: "Why Don't Teachers Teach AI Ethics? Understanding Teachers' Beliefs and Intentions in Chinese AI Curriculum Implementation Through the Theory of Planned Behaviour" by Ming Ma, Davy Tsz Kit Ng, Zhichun Liu, Jionghao Lin, and Gary K.W. Wong.
Published On: Accepted November 29, 2025 (Journal Pre-proof version)
Published By: Computers and Education: Artificial Intelligence (Elsevier Ltd.)