Guided by the SEVT, this study explored the relationships among supportive environments, expectancy–value beliefs, and intentions to learn AI. Overall, both the variable-centered (Study 1) and person-centered approach (Study 2) showed the critical role of supportive environments and expectancy–value beliefs in fostering students’ intentions to learn AI. Study 1 indicated that expectancy–value beliefs mediated the effect of supportive environments on intentions to learn AI. Study 2 identified three subgroups of students with high, medium, and low supportive environments and expectancy–value beliefs. Two studies indicated that students with supportive environments and high levels of expectancy–value beliefs showed the strongest intentions to learn AI. Additionally, we found that demographic factors had a weak effect on supportive environments, expectancy–value beliefs, and intentions to learn AI and their relationships.
The importance of supportive environments and expectancy–value beliefs
Our study provides evidence that both supportive environments and expectancy–value beliefs are crucial factors in fostering students’ intentions to learn AI, which is consistent with previous research (Kelly et al.,
2023; Teo & Zhou,
2014; Venkatesh et al.,
2003). First, our findings highlight the importance of creating a supportive environment that emphasizes the importance of AI and offers adequate technology resources and technical assistance to foster students’ intentions to learn AI ( Teo,
2010). Second, our study reveals that students who perceive higher levels of efficacy and usefulness in learning AI are more likely to learn AI. This finding is consistent with previous research, as students tend to devote more time and energy to learning activities when they have confidence in mastering and perceive important (Huang et al.,
2020; Park,
2009). Third, the mediation effect of expectancy and value beliefs between supportive environments and intentions to learn AI suggests that supportive environments can enhance students’ efficacy and perceived usefulness, thereby fostering their intentions to learn AI. This finding aligns with SEVT (Eccles & Wigfield,
2020) and informs universities to simultaneously focus on creating supportive environments and promoting expectancy–value beliefs to foster students’ intentions to learn AI.
Our study complements previous variable-centered research by revealing individual differences in students’ perceptions of supportive environments and expectancy–value beliefs in learning AI. We identified three subgroups of students that are not visible in variable-centered research (e.g., Chai et al.,
2020,
2022b; Chatterjee & Bhattacharjee,
2020). The identification of low, medium, and high profiles provides a more nuanced understanding of individual differences in students’ perceptions of supportive environments and expectancy–value beliefs, which can inform stakeholders to tailor interventions for the subgroups of students based on their specific needs. The majority of students in our sample (63%) fell into the low and medium profiles of supportive environments and expectancy–value beliefs, indicating that there is a huge room for improvement, especially for universities to optimize students’ learning environments and foster their expectancy–value beliefs in learning AI. Specifically, the 9% of students in the low supportive environments and expectancy–value beliefs profile require particular attention to ensure that they have adequate support to develop their intentions to learn AI.
The role of demographic factors in fostering students’ intentions to learn AI
Our study suggests that gender differences exist in learning AI. Both latent mean differences and LPA analysis demonstrated that male students perceive higher levels of supportive environments and expectancy–value beliefs in learning AI (Papastergiou & Solomonidou,
2005; Qazi et al.,
2022; Terzis & Economides,
2011). This finding highlights the need to pay greater attention to the supportive environments and expectancy–value beliefs of female students to enhance their learning in AI. The male advantage in AI seems to corroborate prior research showing a male advantage in STEM-related subjects (e.g., Wang & Degol,
2017), but seems go against prior research showing male disadvantage in other school subjects (e.g., King & Ganotice,
2014; King,
2016; Korpershoek et al.,
2021; Voyer & Voyer,
2014).
Our study shows that, compared to their senior peers, junior students were more likely to fall into the profile of high support and expectancy–value belief and were more influenced by supportive social norms (Hauk et al.,
2018; Venkatesh & Davis,
2000). As individuals gain more experience with technology, the influence of significant others is expected to diminish, with individuals relying more on their own past experiences to shape their perceptions of technology, rather than the opinions of others (Venkatesh & Davis,
2000; Venkatesh & Morris,
2000). Therefore, junior students might be more sensitive to the opinions of their teachers, peers, and parents. This finding suggests that universities should focus on providing support specifically for junior students to increase their awareness of the importance of AI and foster their intentions to learn it.
Practical implications
This study has practical implications for improving students’ intentions to learn AI. First, our study highlights the significance of supportive norms in promoting students’ self-efficacy and intentions to learn AI. From a practical perspective, one way to create supportive norms is to offer AI-related enrichment lessons (Chai et al.,
2020; Kong et al.,
2021). Another route is through positive encouragement from parents, teachers, and peers (Sohn & Kwon,
2020). This finding emphasizes the importance of creating an environment that supports and encourages AI learning. Universities could consider incorporating AI-related content into the existing curriculum system (e.g., general elective courses) and offering more AI-related workshops (e.g., artificial intelligence and data science) to provide opportunities for students to enhance their experience with AI learning. Furthermore, policymakers and university leaders might need to develop clear initiatives to address the increasing importance of AI in the twenty-first century and enhance graduates’ awareness of the crucial need for AI. For example, the government could publicize job reports to highlight how AI can improve employability. The university could also invite industry leaders as guest speakers to discuss the implications of AI developments in the future workplace. This is especially important in the midst of worries about the negative effects of generative AI, which could translate into restrictive policies instead of enabling ones.
Second, facilitating conditions were operationalized as the accessibility and availability of support, including technology resources and technical assistance when needed (Venkatesh et al.,
2003). Our study highlights the critical role that infrastructure and resources play in supporting students’ intentions to learn AI. In this context, universities could prioritize the development of AI-compatible infrastructure and provide students with access to the necessary technological resources (e.g., AI-empowered software and tools) and technical assistance. Centers for teaching and learning or academic development centers housed within universities need to upgrade the technical proficiencies of support staff to provide relevant support to university educators and students.
Third, we found that expectancy and value beliefs mediated the relationship between supportive environments and intentions to learn AI. Previous research has indicated that these beliefs are malleable and can be nurtured through appropriate interventions, such as connecting tasks to valued identities and helping students reflect on what prepares them for future success (Rosenzweig et al.,
2022). To promote knowledge creation and teaching utilizing AI, higher education educators may need to intentionally learn how to apply AI into their respective specializations. This may require dedicated funding for projects that explore how to integrate AI into subject matter research, teaching, and evaluation, as well as the possible impacts of such integration. When subject matter experts consider AI as a knowledge creation and teaching tool (e.g., in the field of medicine), the integration of AI tools into the curriculum for undergraduates would become more common, establishing new norms for subject matter expertise. However, despite the potential benefits, few cross-disciplinary studies have been reported. Our study emphasizes the importance of creating a supportive learning context to foster students’ motivation and expectancy beliefs in learning AI. Universities can provide students with opportunities to engage in AI-related activities that align with their values and promote confidence in their abilities by offering personalized feedback and support.
Fourth, our study indicates that female students receive less support and have lower expectancy–value beliefs than their male peers. Educators might need to create a more supportive and inclusive learning environment for female students by considering their specific needs and experiences. Meanwhile, we also found that junior students are more sensitive to social norms than their senior peers, which suggests that universities might need to set up more AI-related elective courses for junior students.
Limitations and future directions
Several limitations should be considered when interpreting the findings. First, the data used in this study were collected from self-report surveys, which might not capture the full breadth and complexity of the constructs we are measuring. We encourage future studies to collect qualitative data (e.g., interviews) to cross-validate our research findings.
Second, our data are cross-sectional, which prevents us from drawing causal and reciprocal relationships among variables. Future studies might need to gather longitudinal data to provide a more dynamic picture of how the environment and beliefs interact to promote students’ intentions to learn AI.
Third, the sample of this study was drawn from Chinese mainland universities, which may limit the cross-cultural generalizability of the findings. While studies conducted in other countries, such as the UK and Belgium (Udeozor et al.,
2023), Thailand (Ngampornchai & Adams,
2016), and Lebanon (Tarhini et al.,
2015), have demonstrated the importance of supportive environments in influencing students’ intentions to learn technologies, most of them did not specifically focus on AI. Therefore, we encourage researchers to cross-validate the generalizability of our findings in other cultural contexts. Furthermore, our study only examined university students, and future research could investigate the impact of supportive environments on AI learning among other age groups, such as primary and secondary school students.
Fourth, this study was quantitative in nature, focusing on exploring the relationship between variables rather than providing answers to “how-to” questions. Although this research paradigm has been widely used in existing technology research (e.g., Chatterjee & Bhattacharjee,
2020; Staddon,
2020), it limited our ability to identify best practices that could guide future practice. Therefore, we encourage future studies to adopt qualitative research methods, such as conducting interviews with relevant stakeholders, including government officials, teachers, and students, to better understand how these supportive environments influence students’ intentions to learn AI and to identify effective strategies for promoting students’ intentions to learn AI.
Fifth, although this quantitative study empirically revealed the relationship among supportive environments, expectancy and value beliefs, and students’ intentions to learn AI based on the SEVT, one limitation was that all variables were predetermined based on the theoretical framework using a top-down paradigm. This approach limited our ability to gain a deeper understanding of how supportive environments influence students’ intentions to learn AI, as well as explore how other elements of supportive environments influence students’ intentions to learn AI. Therefore, other bottom-up exploratory methods, such as qualitative or mixed-methods designs, can be used in future studies to deepen our understanding of the relationship between supportive environments, expectancy and value beliefs, and students’ intentions to learn AI using an exploratory research paradigm.
Sixth, although we provided a brief explanation of the difference between AI and non-AI and gave some examples to students before they filled out the questionnaire, it is possible that students’ perceptions of supportive environments and expectancy–value beliefs may vary depending on their exposure to AI-related programs or courses. Therefore, future studies could provide a more nuanced understanding by examining the impact of different levels of exposure to AI on students’ expectancy and value beliefs in learning AI, as well as the role of supportive environments in shaping these beliefs.
Last, we only examined the direct effect of supportive environments on students’ expectancy and value beliefs. However, previous research by Eccles and Wigfield (
2020) has suggested that learning environments may also influence these beliefs indirectly through social cognitive factors, such as personal and social identities, short-term and long-term learning goals, and affective reactions. Therefore, to gain a more comprehensive understanding of the relationship between supportive environments and expectancy–value beliefs in learning AI, future studies could explore the potential role of these mediators.