Students Lead AI: Understanding Student Agency in AI-supported Learning

• In Event: Highlighted Session: Perils and Promises of Generative AI in Higher Education: Comparative Insights from Mainland China, Hong Kong, Japan, Mongolia, and the U.S.

Proposal

Generative AI (GAI) tools like ChatGPT are increasingly adopted by university students, presenting both opportunities and challenges in higher education (HE). In this context, student agency—the capacity to take initiative and shape one’s educational experiences—has become crucial for the responsible use of GAI in academic settings. However, there remains limited empirical knowledge about the nature and manifestation of student agency in AI-supported learning (ASL) environments. The status quo is largely due to the inherent complexity and ambiguity of the concept, which complicates efforts to conceptualize, operationalize, and observe this phenomenon within real-world educational contexts.

To address this gap, this qualitative study examines how students exercised agency in authentic ASL environments. Informed by the effortful and authorial agency perspectives, we conceptualize student agency in ASL as a capability to actively shape, influence, and transform their educational experiences by engaging with AI tools in ways that transcend the given parameters of the tools and environments. Following a grounded theory approach, we analyzed 20 Hong Kong undergraduate students’ conversation records with GAI tools and conducted stimulus-recall interviews to explore their thought processes, decision-making strategies, and reflective practices.

Our analysis identified four key aspects of student agency in ASL:

1) Control and (re)direct involves students guiding and configuring AI tools to generate outputs in alignment with their learning needs and objectives.
2) Mindful adoption refers to the critical and deliberate engagement with AI outputs, ensuring a contextually relevant and appropriate adoption of the outputs.
3) External help-seeking reflects students’ recognition of AI's limitations and proactive efforts to supplement, validate, or enhance their learning by utilizing additional educational resources and human expertise.
4) Reflective learning encompasses ongoing self-examination and monitoring of their AI use and their cognitive processes, enabling them to actively and intentionally refine and improve their learning experiences in a self-directed manner.

Based on the grounded evidence, we propose a theoretical framework that captures the dynamic, iterative, and multifaceted nature of student agency in AI-supported learning (ASL) environments. This framework emphasizes the interconnectedness of the four identified aspects and demonstrates how they collectively contribute to students’ ability to effectively navigate and harness the potential of AI tools in their academic work. As such, this study fills a significant gap in the literature, offering a clear conceptualization of student agency that can inform future research and the development of measurement tools.

Practically, this study offers valuable insights for HE educators, instructional designers, and administrators. Training programs and interventions can be developed to specifically cultivate student agency in ASL environments. This framework can be also used to guide the development of AI systems that are more responsive to students' needs, fostering environments where students can effectively control, evaluate, and adapt AI-generated content to their learning objectives. Administrators can use these insights to advocate for curriculum changes that integrate AI literacy and agency-building activities, ensuring that students are equipped with the skills necessary to navigate and thrive in increasingly AI-driven educational contexts.

Authors

• Yun Dai

• Sichen Lai