Page 709 - ISC PROCEEDINGS 21.4

Perceived ease of use also significantly affects intention to use AI tools for learning
(Beta = 0.276). This is consistent with TAM and with Maheshwari [5], who found that ease
of use plays an important role in students’ intention to adopt ChatGPT in Vietnam. At the
same time, its weaker effect relative to perceived usefulness suggests that convenience
alone is not sufficient. Once students become familiar with AI tools, their judgments
increasingly depend on whether the tool is truly beneficial for learning rather than merely
easy to operate.
Trust in AI (Beta = 0.259) is another important determinant. This result is in line with
Duong [6] and Ngo et al. [10], both of whom highlight the relevance of trust in students’
AI-related behavior. In the context of learning, trust is especially salient because students
must decide whether AI outputs are credible enough to support academic tasks. The
present finding indicates that students’ willingness to use AI is strengthened when they
believe the technology can provide reasonably reliable and meaningful assistance.
Social influence is also significant, although it is the weakest predictor in the model
(Beta = 0.227). This does not mean that the social environment is unimportant; rather, it
suggests that peers, instructors, and prevailing academic norms help legitimize AI use but
are somewhat less decisive than perceived value and personal capability. This finding
remains consistent with UTAUT and with recent work on AI adoption in higher education
[9,10]. In other words, students are affected by the views of important others, but their
own evaluation of usefulness, confidence, and trust still plays a more direct role.
Taken together, these results clarify the study’s contribution. First, unlike studies
that focus narrowly on ChatGPT, the present article addresses the broader category of AI
tools for learning, which better reflects current educational practice. Second, by
integrating TAM, UTAUT, and TPB-related constructs in one direct-effects model, the
study shows that technological, social, and learner-related factors all matter
simultaneously. Third, the high explanatory power of the model (adjusted R² = 0.805)
indicates that this integrated framework provides a robust baseline for understanding AI
adoption in Vietnamese higher education, even though more advanced approaches such
as CFA or SEM could extend the analysis in future work.
5.2. Practical implications
The findings have several implications for higher education institutions in Vietnam.
First, universities should not promote AI use in learning through general encouragement
alone. Because perceived usefulness is the strongest predictor, institutions need to
demonstrate concrete academic value through discipline-specific examples, practical
workshops, and classroom tasks that show how AI can support learning without replacing
critical thinking.
Second, universities should invest in building students’ AI self-efficacy. Training
should focus on practical skills such as writing effective prompts, evaluating outputs,
verifying information, and using AI ethically in different academic activities. This is
especially important because confidence in using AI significantly strengthens intention to
use it.
Third, institutions should establish transparent and responsible guidelines for AI-
supported learning. The significance of trust in the model suggests that students’
willingness to use AI depends partly on whether they see the technology as sufficiently
reliable and legitimate in academic contexts. Clear rules on verification, citation,
acceptable use, and academic integrity can therefore support adoption while reducing
misuse.

708