Exploring the effects of AI literacy in teacher learning: an empirical study

The integration of Artificial Intelligence (AI) in education presents significant opportunities, but many teachers lack the necessary AI knowledge and skills to effectively utilize it. This study addresses the scarcity of research on K-12 teacher AI education, particularly in China, where previous research indicated teacher anxiety and reluctance towards AI learning. The research aims to explore the antecedents of K-12 teachers' intentions to learn AI by proposing and validating a model that integrates AI literacy and awareness of AI ethics with the Theory of Reasoned Action (TRA) and the Theory of Planned Behavior (TPB). Understanding these determinants is crucial for promoting effective AI learning among teachers, which is a prerequisite for effective AI-enhanced teaching practices. The study builds upon existing models like TRA, TPB, and TAM, but addresses their limitations by explicitly including the often-overlooked perspectives of AI literacy and ethical awareness. AI literacy is considered an essential skill in the AI-powered world, influencing perceptions of AI and self-efficacy. Awareness of AI ethics is crucial due to the potential risks associated with AI, impacting trust and willingness to learn. The integration of these perspectives is novel and aims to provide a more comprehensive understanding of the teachers' behavioral intentions.

The study draws upon the Theory of Reasoned Action (TRA), the Theory of Planned Behavior (TPB), and the Technology Acceptance Model (TAM) to predict behavioral intentions. However, these models have seldom been applied to AI education, and the construct of intentions to learn AI, especially among teachers, has been scarcely explored. The study incorporates two crucial aspects often neglected in traditional models: AI literacy and ethical awareness. AI literacy is defined as the ability to understand AI concepts and applications, directly impacting attitudes towards AI learning and self-efficacy. Ethical awareness, concerning the risks and responsibilities of AI, influences attitudes towards AI. The study hypothesizes that teachers' perceptions of AI's social benefit, self-efficacy in learning AI, AI literacy, and awareness of AI ethics directly and indirectly influence their intentions to learn AI.

The study employed a quantitative approach, collecting survey data from 318 K-12 teachers across sixteen provinces in China. The sampling method involved randomly selecting partner schools and teachers within those schools, aiming for a sample size adequate for structural equation modeling (SEM). A seven-point Likert-type questionnaire was used, incorporating scales adapted and translated from existing literature to measure AI literacy, self-efficacy in learning AI, perceptions of AI's use for social good (PAIS), awareness of AI ethics (AAIE), and behavioral intentions to learn AI. The questionnaire's validity and reliability were established through confirmatory factor analysis (CFA). A two-step structural equation modeling approach was employed, first conducting CFA to validate the measurement model, and then estimating the structural model to test the hypotheses and analyze direct and indirect effects. AMOS 21 software was used for data analysis. The study also considered control variables such as gender, school stage, age, school district, education background, and major to assess their influence on the relationships between the constructs.

Confirmatory factor analysis (CFA) validated the measurement model, demonstrating the validity and reliability of the constructs. The structural equation modeling (SEM) analysis revealed that all eleven hypotheses were supported. Specifically, perceptions of the use of AI for social good (PAIS) and self-efficacy in learning AI directly and positively influenced behavioral intentions to learn AI. AI literacy and awareness of AI ethics indirectly influenced behavioral intentions, mediated by PAIS and self-efficacy. AI literacy directly influenced PAIS, self-efficacy, and awareness of AI ethics. Awareness of AI ethics directly influenced PAIS. Control variables analysis showed that teacher age and major significantly impacted awareness of AI ethics, while school district significantly impacted PAIS. The model explained 75% of the variance in behavioral intentions to learn AI. The Sobel test confirmed the indirect effects of AI literacy on behavioral intentions mediated by PAIS and self-efficacy and the indirect effect of AI literacy on PAIS mediated by awareness of AI ethics. Finally, the awareness of AI ethics had an indirect influence on behavioral intentions mediated by PAIS.

The findings support the integration of TRA and TPB in the context of AI education, confirming the importance of attitudes (PAIS) and self-efficacy in predicting behavioral intentions to learn AI. The study's key contribution lies in incorporating AI literacy and ethical awareness, demonstrating their significant, albeit indirect, influence on teachers' intentions. AI literacy's impact on various constructs underscores its crucial role in promoting AI learning. The influence of teacher age, major, and school district on different constructs reveals contextual factors that should be considered when designing AI professional development programs. The strong explanatory power of the model highlights the comprehensive nature of the proposed framework.

This study provides a comprehensive model for understanding the factors influencing K-12 teachers' intentions to learn AI, emphasizing the crucial role of AI literacy and ethical awareness. The findings highlight the need for professional development programs that focus on building teachers' AI literacy, self-efficacy, and understanding of AI's social benefits and ethical considerations. Future research could explore the effectiveness of such programs and expand the model by incorporating additional factors like subjective norms or exploring regional variations in teachers' AI learning intentions.

The study's limitations include its reliance on self-reported data, which could introduce subjectivity. The geographical scope of the study might limit the generalizability of findings to other regions in China or globally. While the model demonstrates strong explanatory power, including additional factors could further improve it. Finally, the failure to validate three specific questionnaire items highlights a potential issue with item validity across cultural contexts that requires further exploration.