Enhancing Global Artificial Intelligence Literacy,Innovation and Entrepreneurship Through an Exploratory Cross-Cultural Study: Interactions With Pedagogical Agents in Immersive Virtual Reality Environments

Bloom’s Taxonomy for AI-Supported Learning

Bloom’s taxonomy consists of three domains of learning, i.e., cognitive, affective, and psychomotor (Crawley et al., 2007; Sharunova et al., 2018). It includes six levels. The remember level refers to recognizing or recalling knowledge from memory. Remembering is when memory is used to produce or retrieve definitions, facts, or lists, or to recite previously learned information. The understand level represents constructing meaning from different types of functions, be they written or graphic messages or activities. The apply level consists of carrying out or using a procedure by executing or implementing it. The analyze level involves breaking materials or concepts into parts, determining how the parts relate to one another or how they interrelate, or determining how the parts relate to an overall structure or purpose. The evaluate level represents making judgments based on criteria and standards through checking and critiquing. The create level involves assembling elements to form a coherent whole or reorganizing them into a new structure or pattern.

Bloom’s taxonomy, as a well-established framework in the field of education, has been widely applied in AI-supported learning contexts. For instance, Hutson (2025) advocate for a phased introduction of AI into instruction, whereby students must demonstrate cognitive mastery at lower levels of Bloom’s hierarchy before leveraging AI tools to enhance higher-order skills such as synthesis and evaluation. Empirical studies exploring the integration of generative AI with Bloom’s taxonomy reveal nuanced dynamics in learners’ cognitive processes. For example, research involving primary students interacting with conversational agents showed frequent engagement in creating and evaluating, yet surprisingly low proficiency in applying AI-generated knowledge to new contexts (Hui, 2025). This suggests a critical need to reinforce middle-tier cognitive skills to bridge the gap between comprehension and innovation. Moreover, scholars have argued that the growing reliance on large language models (LLMs) necessitates a reconceptualization of information literacy frameworks. A proposed LLM-driven taxonomy maps cognitive behaviors such as perceiving, reasoning, and curating onto Bloom’s hierarchy, offering a structured pathway to foster metacognitive awareness and critical AI fluency (Luo et al., 2025).

Collectively, these studies underscore the relevance of Bloom’s taxonomy not only as a curriculum design tool but also as a cognitive framework for evaluating the depth and quality of AI-supported learning. Despite these advancements, there remains a notable research gap concerning how Bloom’s taxonomy can be holistically applied to assess students’ cognitive development in areas such as global AI literacy within immersive and cross-cultural learning environments.

VR in Cultural Learning

Virtual reality (VR) immerses users in highly engaging, virtual environments where their interaction with the physical world is either minimal or absent. To achieve a profound sense of immersion, VR creates expansive, inclusive, and vivid simulations of reality (Slater & Wilbur, 1997). Recognized for its significant potential, VR is widely regarded as a powerful tool for fostering meaningful, context-rich learning experiences, making it particularly beneficial in cultural learning contexts.

Previous studies on the application of VR in cultural learning have highlighted its effectiveness in enhancing students’ cross-cultural performance, including improvements in cross-cultural sensitivity (Gao et al., 2021), cross-cultural knowledge (Georgiev & Nikolova, 2021; Lucet, 2009), and cultural communication skills (Li et al., 2022; Pantano & Tavernise, 2009). Additionally, research has explored learners’ perceptions of VR and their overall experiences with cultural learning activities (e.g., Dahya et al., 2021; Shadiev et al., 2021).

However, with the rapid advancement of AI technologies (e.g., large language models, PAs), it remains unclear whether these innovations can seamlessly extend VR environments to further enhance cross-cultural learning.

The Role of AI in Innovation and Entrepreneurship Education

Innovation is commonly understood as a process that unfolds in three stages: the generation of new ideas, their evaluation, and ultimately, their practical implementation (Carayannis et al., 2015; Kahn, 2018). Innovation emphasizes the introduction of novel concepts and is driven by existing frameworks to propose alternatives that challenge conventional thinking. As a key component of effective entrepreneurship, innovation plays a vital role in the creation of sustainable, profitable, and scalable enterprises. Entrepreneurship is defined as the creation of viable, profitable, and scalable businesses, which, through innovation, form self-replicating and mutually reinforcing networks and knowledge clusters (innovation ecosystems) that contribute to robust competitive advantages (Carayannis & Campbell, 2014; Lans et al., 2010).

Recent advancements in AI, particularly in generative AI technologies, have significantly influenced the landscape of innovation and entrepreneurship education. Scholars have emphasized the urgent need to integrate these technologies into pedagogical design and practice to better prepare students for future entrepreneurial opportunities (Bell & Bell, 2023). Recent studies have also provided a comprehensive perspective on the role of artificial intelligence in innovation and entrepreneurship, highlighting how AI can help regions capitalize on opportunities arising from technological advancements and emphasizing its significance in innovation ecosystems (Carayannis & Grigoroudis, 2023). In parallel, simulated environments and adaptive platforms provide experiential learning opportunities that mirror entrepreneurial decision-making scenarios, helping students build confidence and autonomy, two traits essential for innovation (Kangiwa et al., 2024).

Although some studies have explored students’ attitudes, perceptions, reflections, and intentions related to innovation and entrepreneurship (e.g., Haddad et al., 2021; Rahim & Mukhtar, 2021; Täks et al., 2014), there is limited research that investigates the potential of AI-assisted cross-cultural learning activities to enhance innovation and entrepreneurship.

Global AI Literacy

AI literacy encompasses the recognition and understanding of AI technologies in practical contexts, the proficient application of these technologies to achieve tasks, and the critical analysis of the data and insights they generate (Chung et al., 2024; Walter, 2024). Generally, AI literacy includes three key components: AI knowledge, AI skills, and AI attitudes. AI knowledge covers essential areas such as knowledge representation, data learning, machine learning, deep learning, and ethical issues related to AI (Ahn & Oh, 2024). AI skills enable learners to identify and apply suitable AI tools to achieve learning objectives in diverse contexts. AI attitudes involve understanding and evaluating students’ perceptions and emotional responses to artificial intelligence. Furthermore, based on the iterative co-design framework for AI literacy (Chiu et al., 2024), AI ethics is also considered a key dimension. Recognizing AI ethics, such as data security risks, privacy concerns, and algorithmic biases, helps learners harness the benefits of AI while mitigating potential harms (Alnahhas & Yousef, 2024; Heilinger, 2022).

Developing AI literacy from both local and global perspectives is crucial. A local perspective enables individuals to comprehend the specific cultural, social, and economic impacts of AI within their own communities, facilitating more relevant and effective applications (Galmar, 2024; Perchik et al., 2023). In contrast, a global perspective expands awareness of diverse AI innovations and practices, encouraging collaboration and preparing individuals to navigate the interconnected nature of today’s world (Korte et al., 2024; Ma et al., 2024).

While the global importance of AI literacy is widely acknowledged, most studies on its development tend to focus predominantly on local perspectives, with limited attention given to its global dimension.

Bloom’s Taxonomy, AI and VR

Bloom’s taxonomy provides a well-established framework for understanding cognitive development in education, progressing from lower-order thinking skills to higher-order cognitive abilities. However, traditional instructional methods often struggle to effectively support students in advancing through these cognitive stages, particularly in fostering the higher-order thinking skills essential for innovation and entrepreneurship. Emerging technologies such as VR and AI-driven PAs offer transformative potential in this regard. VR immerses learners in authentic, interactive environments that facilitate experiential learning, allowing students to engage in problem-solving tasks, collaborative projects, and real-world simulations that align with Bloom’s higher-order cognitive domains (Chen, 2016; Huang et al., 2024). Meanwhile, PAs dynamically adapt instructional strategies to learners’ cognitive needs, providing personalized scaffolding, real-time feedback, and adaptive challenges that help students progress through the cognitive hierarchy of Bloom’s taxonomy (Gonsalves, 2024; Hui, 2025).

However, a disconcerting reality persists that the majority of virtual learning environments may fall short of established educational standards and pedagogical principles (Rezaei-Zadeh et al., 2013). Barari et al. (2022) developed pedagogical standards based on Bloom’s taxonomy for virtual learning environments. Additionally, some researchers have leveraged Bloom’s framework to enhance questioning strategies and reflective thinking through AI-powered tools in language learning. Despite its significant potential, the integration of VR and PAs within Bloom’s taxonomy remains largely unexplored, particularly in the context of advancing global AI literacy, innovation, and entrepreneurship education.

At the foundational levels of Bloom’s cognitive hierarchy, VR offers an immersive and engaging platform that enables students to grasp essential AI concepts, while PAs enhance understanding by delivering context-sensitive explanations and interactive assessments. As learners ascend to higher-order cognitive domains, VR supports experiential learning through scenarios such as cross-cultural simulations, where students can apply AI tools and devise innovative projects within diverse global contexts. Concurrently, PAs provide personalized cognitive scaffolding, guiding learners through intricate problem-solving processes and cultivating metacognitive awareness. By aligning Bloom’s cognitive stages with immersive and adaptive technologies, the synergy of VR and PAs fosters a dynamic, tailored, and intellectually enriching educational experience.

Research Model for the Present Study

AI literacy is closely linked to the success of human-AI interactions, particularly when AI is embodied through PAs in learning environments. In educational contexts, PAs may play a critical role in fostering students’ innovation and entrepreneurship. Similarly, in VR-supported cross-cultural learning activities, students’ global AI literacy may be enhanced through their interactions with AI PAs.

To conceptualize how these elements interrelate, this study introduces a theoretical model (Figure 1) grounded in Bloom’s taxonomy. In this model, VR-supported cross-cultural learning activities mediated by AI-driven PAs serve as the pedagogical framework through which students’ progress from “remembering” and “understanding” AI-related knowledge to “applying” it in cross-cultural contexts. This structured cognitive development supports not only the growth of global AI literacy but also the cultivation of creative and strategic competencies that underpin innovation and entrepreneurship.

OPEN IN VIEWER

Figure 1.

A model of VR and PA-supported global AI literacy, innovation, and entrepreneurship based on the remembering, understanding, and applying levels of Bloom’s taxonomy. Note: Analyze, Evaluate, and Create were not measured in the present study

This model, grounded in Bloom’s taxonomy, also emphasizes the interactive relationship between intelligent technologies (i.e., PAs and VR), teachers, students, and learning content related to AI and its application in cultural contexts.

Specifically, teachers, following the framework of Bloom’s taxonomy, design cross-cultural learning activities and content integrated with PAs within VR environments, guiding students throughout the learning process. For instance, teachers assign cross-cultural tasks and provide instructional manuals that outline how to interact with PAs and create 360-degree videos. The overarching aim is to enhance students’ global AI literacy, as well as their innovation and entrepreneurship capabilities.

PAs, serving as personalized guides, offer step-by-step support to students in completing cross-cultural learning tasks. Through interactive dialogue, PAs assist students in summarizing and refining video scripts, addressing cultural questions, and expanding their understanding of foreign cultures and AI application in cultural contexts in real-time. The diverse sources of information provide learners with novel business opportunities and solutions.

VR technology immerses learners in a virtual environment, facilitating deeper engagement with foreign cultures. By simulating real-world scenarios, VR allows students to experience cultural customs and traditions first-hand, such as traditional clothing and culinary practices. This array of information sources helps stimulate innovative thinking among aspiring entrepreneurs. Moreover, the immersive nature of VR enables students to explore AI applications within foreign cultural contexts, such as robot-assisted food delivery, thereby helping entrepreneurs recognize global trends.

Through the creation and viewing of 360-degree videos, learners gain enriched content on AI applications, their use within cultural environments, and engagement with cultural practices. These immersive experiences significantly contribute to enhancing global AI literacy.

Based on Bloom’s taxonomy, the design of the cross-cultural learning activities aligns with the lower cognitive levels: remembering, understanding, and applying. In these activities, students benefit from the guidance provided by instructors (e.g., example videos and scripts), the support of PAs in script writing, the immersive VR environment, and the application of AI in various cultural scenarios, all of which collectively foster the development of global AI literacy, innovation, and entrepreneurship.

Participants

It is essential to select students with cross-cultural needs for experimental studies aimed at enhancing cross-cultural educational awareness (Chen, 2024). For this study, convenience sampling was used to recruit 68 students from three courses, representing three countries. Among them, 24 students were from China, 20 students from Northern Cyprus, and the remaining 24 students from Kazakhstan. Student demographics are shown in Table 1.

OPEN IN VIEWER

Table 1.

Distribution of students

Variables		Frequency	Variables		Frequency
Gender	Male	19	Education level	Bachelor	31
	Female	49		Bachelor	31
Major	Education	35		Doctor	6
	Language learning	24	Age	19-21	40
	Management	4		22-25	18
	Gastronomy	3		>25	10
	Science and technology	2

Pedagogical Agent Context and Design

Pedagogical agents (PAs) are computer-controlled, on screen characters designed to facilitate instruction, and have been used for presenting information, increasing motivational support, and providing conversation with the learner (Schroeder & Craig, 2021; Siegle et al., 2023). Through text- and voice-based interactions, these agents actively encourage learners to engage in educational tasks while fulfilling core instructional functions such as guiding learning processes, evaluating student outputs, and delivering formative feedback (Lan & Chen, 2024).

Although PAs have great potential to enhance teaching and learning, particularly in fostering AI literacy, innovation, and entrepreneurship education, existing research rarely examines their role in supporting learning. Moreover, research on the integration of PAs into immersive virtual reality and their potential to enhance the learning process remains highly limited or almost non-existent. Therefore, this study aimed to address this gap. We employed PAs to facilitate learning, evaluate students’ outputs, and provide real-time feedback (Lan & Chen, 2024). To support cross-cultural learning, we utilized ChatGPT 4o in this study. Within this version, we employed the Generative Pre-trained Transformer (GPT) functions, which allow users to create customized GPTs for specific purposes (Figure 2). Building on these functions, we developed three tailored PAs, each configured with structured project background information and clearly defined objectives to minimize AI hallucinations and enhance credibility. These PAs were designed respectively for script improvement, oral interaction, and reflection, enabling them to provide dynamic and context-sensitive responses throughout the learning activities.

OPEN IN VIEWER

Figure 2.

Customized configuration of PAs for cross-cultural learning

Furthermore, the PICO VR (https://www.picoxr.com/cn) tool used in this study was connected to the internet to enable the PAs to operate within the VR environment and to support oral interaction through ChatGPT 4o′s real-time voice conversation function. Within the VR environment, students interacted with the PA through voice, asking questions related to their partners’ cultures and receiving immediate responses. This integration supported multimodal communication, which enhanced engagement and accessibility in the cross-cultural learning activities.

These PAs, implemented as 1) a facilitator for guiding script writing, 2) an agent for oral interaction in VR environments, and 3) a feedback provider for delivering real-time responses to reflection writing, are integrated into a VR environment as software-based agents without a physical or rendered avatar. The PA types and their corresponding functions are summarized in Table 2. These agents underwent rigorous testing by two researchers, who provided iterative feedback to refine and enhance their functionality and effectiveness.

OPEN IN VIEWER

Table 2.

Types and Functions of Pedagogical Agents

Type of PA	Functions
Script assistance	1) Role personalization; 2) structured writing guidance; 3) feedback and revision support; 4) content consolidation
Oral interaction	1) Role personalization; 2) analysis for script, picture and video; 3) voice-based interaction; 4) real-time cultural feedback
Reflection assistance	1) Role personalization; 2) structured reflection guidance; 3) feedback and revision support; 4) content consolidation

Our choice to utilize PAs in this study was based on their affordances. These agents provide precise, individualized guidance by leveraging AI technologies, addressing challenges such as the limited availability of teachers, who must manage 68 participants from three countries in this study. Unlike teachers, who are constrained by class time and the large number of students, PAs are available 24/7 and can provide one-on-one assistance. Additionally, PAs offer adaptive feedback, track student progress, provide multilingual support, and ensure consistent instruction, enhancing engagement and personalized learning experiences. In the interactions between students and PAs, all communication, whether text-based or voice-based, was conducted in English.

PA support was primarily provided for script creation (e.g., self-introductions, AI applications in cultural contexts, innovation and entrepreneurship, and reflections) and for facilitating the learning process during the viewing of immersive cultural content. For script creation, first, instructional guidance was provided to the agents, enabling them to assist students step by step in developing their scripts for cultural introduction videos. The agents offered clear and precise directions, eliminating the need for teachers to manually guide every student. After scripts were created, PAs performed content checks, ensuring the accuracy and completeness of the information presented in student scripts. For instance, the agents evaluated whether scripts adhered to cultural themes, included relevant details, and met word count requirements. In addition, PAs assessed the quality of cultural content presented in scripts, providing feedback on clarity and informativeness, and language use, particularly in ensuring accuracy and fluency as students wrote in a foreign language. Through scaffolding, PAs inspired students with ideas for script creation and provided links to supplementary information, such as more specific and meaningful details when introducing cultural festivals (e.g., historical significance, unique traditions) or highlighting AI applications in specific cultural contexts (e.g., AI-driven translation in multilingual communities, AI-assisted heritage preservation).

During the viewing of immersive cultural content, PAs facilitated real-time student learning. They were instructed to guide students by analyzing immersive content and scripts, summarizing key information to enhance understanding of the background and central concepts presented in the 360-degree videos. PAs also helped students grasp new and unfamiliar cultural and technological themes by offering clear explanations and contextualizing complex ideas. Additionally, they answered students’ questions, fostering interactive engagement and providing immediate, personalized feedback to reinforce learning.

Cross-Cultural Learning Activity Design

This study was conducted over the course of a semester as part of the AI in Education course. Professors specializing in educational technology shared knowledge on AI and its ethical implications, including topics such as robotics and computer vision technologies. During the research period, participants from the three countries were supported by two instructors through a combination of online and offline interactions. The instructors facilitated online discussions and ensured active student participation in cultural information exchange activities. Additionally, a project manager oversaw the implementation and progress of the activities, encouraged online collaboration among teachers from the three countries, and provided constructive feedback and recommendations. All instructors focused on supporting student engagement in learning activities related to AI, its applications within cultural contexts, and cultural information exchange.

The research process is illustrated in Figure 3. Prior to the learning activity, researchers recruited a total of 68 participants. During the first two weeks, the researchers introduced the technologies employed in the study, including Insta 360 cameras, PAs, and VR headsets. They explained the functionality of these tools in detail and demonstrated how they supported the learning activities. Following this, researchers administered questionnaires and conducted interviews with all participants.

OPEN IN VIEWER

Figure 3.

Cross-cultural learning activity design

In the third and fourth weeks, participants interacted with the PA for script writing in English to create video scripts, which included two sections: self-introduction and local culture with AI applications. For the self-introduction, participants provided details such as their name, nationality, city, university, major, hobbies, interests, and daily life. Subsequently, students recorded 360-degree videos of their self-introductions based on the script content.

In the fifth and sixth weeks, participants selected one of ten cultural themes they were interested in (e.g., festival, tradition, historical building, handicraft, etc.). Under the guidance of the PA for script writing, each participant prepared a detailed description of the selected cultural topic and discussed the application of AI within that cultural context. Following their interactions, the PA consolidated a script for each participant based on their discussions. Using the developed script, participants recorded culture-related 360-degree videos with Insta 360 devices at locations reflective of their cultural backgrounds.

In the seventh week, participants exchanged all videos through a cloud platform. Priority was given to pairing videos on similar themes from two countries for exchange and viewing. Participants could also choose videos that matched their personal interests. This ensured that all ten cultural themes from each country were viewed by international peers. For instance, a Chinese participant (ID 5) created a video about the qipao (traditional clothing), while their Northern Cypriot partner introduced traditional clothing from their country. These videos were paired and exchanged for mutual viewing.

In the eighth week, prior to viewing their partner’s videos, participants submitted their scripts and video screenshots to the PA for oral interaction. The PA analyzed visual elements (e.g., the environment) and summarized the video scripts to prepare students for viewing. Participants then used VR headsets to watch videos created by their foreign peers. During this process, the PA for oral interaction was integrated into the VR environment to assist with cross-cultural learning by providing real-time feedback to enhance participants’ understanding, as illustrated in Figure 4. For example, when Chinese students watched a video created by Northern Cypriot students and noticed the sea, they asked the PA for additional information about Northern Cyprus’s geographical location.

OPEN IN VIEWER

Figure 4.

Interactions between students and PA in VR environments

In the ninth week, under the guidance of the PA for reflection, students reflected on their cross-cultural learning activities, critically evaluating what they had learned. For example, participants considered whether their global AI literacy had improved and whether exposure to foreign cultures had inspired innovative ideas. Upon the completion of the cross-cultural learning activities, researchers conducted questionnaires and interviews with all participants to gather comprehensive feedback and insights.

Data Collection and Analysis

A mixed-methods approach was employed to collect data, incorporating questionnaires, interviews, and reflections. Cultural concepts are often abstract and general, making it challenging to design effective evaluation methods based solely on qualitative interviews and observations (Deardorff, 2006; Gao et al., 2021). As such, a combination of quantitative and qualitative methods is considered the most effective approach to assess cultural learning outcomes.

Improving Global AI Literacy through PA- and VR-Supported Cross-Cultural Activities

To address RQ1, the results of a Wilcoxon signed-rank test demonstrated statistically significant gains in global AI literacy among participants from three countries following cross-cultural learning activities supported by PAs in VR environments, as shown in Table 3.

OPEN IN VIEWER

Table 3.

The Difference Between students’ Global AI Literacy on Pre- and Post-test

Dimension	Pre-test		Post-test		Wilcoxon signed-rank test
Global AI literacy	Mean	SD	Mean	SD	z	Sig.
	1.023	0.743	2.598	0.471	−4.390	<0.01

The cognitive process level in the post-test (M = 2.598, SD = 0.471) was notably higher than that in the pre-test (M = 1.023, SD = 0.743). Statistical analysis revealed a significant difference, z = −4.390, p < 0.01. The effect size calculation further demonstrated a substantial difference in cognitive process levels between the pre- and post-tests, r = 0.79. These results suggest that integrating interactive AI PAs in VR environments to support cross-cultural learning activities is highly effective in enhancing students’ global AI literacy across all dimensions.

Our qualitative data from 68 students were further analyzed using thematic coding based on Bloom’s taxonomy domains, revealing recurring themes related to the integration of PAs and cultural understanding (Table 4). Specifically, advancements were observed across all dimensions, including AI knowledge, AI understanding, and AI application in various cultural contexts. Many participants emphasized how AI technologies, particularly PA-powered translation and communication tools, helped bridge cultural gaps and foster more effective collaboration.

OPEN IN VIEWER

Table 4.

Key Themes of Global AI Literacy (RQ1) From Qualitative Analysis

Main theme	Sub-theme	Frequency	Source
1. AI knowledge	1.1 understanding AI technologies	56	ID 8: Before the learning activities, I had never used an AI tool. Now, I understand how machine learning algorithms work and can explain them to my peers.
	1.2 familiarity with AI applications	49	ID 14: I Always thought AI was just about robots, but now I realize it’s used in daily life and cultural contexts.
	1.3 perceptions of AI’s impact on society and daily life	39	ID 37: AI is transforming society. I See it helping us in education and culture, but I worry about its effect on jobs.
2. AI understanding	2.1 from limited awareness to diversified understanding	43	ID 8: When we started, I had heard ChatGPT but only name. Now I have started to think about how I will use museum educational programming with my students.
	2.2 AI as a resource for cross-cultural exchange	52	ID 47: I Realized that we can also use AI technologies to introduce our culture to the people from all around the world.
	2.3 AI’s societal value and potential	61	ID 56: I’ve learned how AI can enhance language learning and help bridge cultural gaps through translation tools.
3. AI application in various cultural contexts	3.1 AI for preserving and digitizing cultural heritage	59	ID 21: AI can be used to digitize images, videos, historical records, interviews with craftsmen, and 3D models of tools and performance scenes.
	3.2 integration of AI into daily life and traditional practices	57	ID 47: AI-driven platforms have transformed social interactions and commerce, blending seamlessly into cultural norms of convenience and connectivity.
	3.3 AI as a bridge between cultural preservation and innovation	47	ID 64: AI is enhancing culinary traditions, with robots helping to prepare foods, improving efficiency while keeping cultural practices alive.

Improving Innovation and Entrepreneurship through PA- and VR-Supported Cross-Cultural Activities

To address RQ2, the results for innovation and entrepreneurship demonstrated statistically significant improvements following the intervention (Table 5). Specifically, in the dimension of innovation, post-test scores (M = 4.245, SD = 0.425) were significantly higher than pre-test scores (M = 3.268, SD = 0.433), with a paired-sample t-test result of t = −12.902, p < .01. Similarly, for entrepreneurship, post-test scores (M = 4.190, SD = 0.296) significantly exceeded pre-test scores (M = 3.077, SD = 0.621), with a t-value of t = −13.600, p < .01. These findings suggest that the PA- and VR-supported cross-cultural activities may have contributed to fostering both innovation and entrepreneurship among participants from three countries.

OPEN IN VIEWER

Table 5.

Differences in innovation and entrepreneurship

	Pre-test		Post-test		95% CI	t	Sig.
Dimension	Mean	SD	Mean	SD
Innovation	3.268	0.433	4.245	0.425		−12.902	<0.01
China	3.330	0.301	4.194	0.215	[-1.025, −0.705]	−11.177	<0.01
Northern Cyprus	3.071	0.610	4.304	0.522	[-1.603, −0.863]	−6.974	<0.01
Kazakhstan	3.372	0.415	4.247	0.498	[-1.140, −0.610]	−6.832	<0.01
Entrepreneurship	3.077	0.621	4.190	0.296		−13.600	<0.01
China	2.933	0.640	4.131	0.243	[-1.454, −0.941]	−9.652	<0.01
Northern Cyprus	3.037	0.453	4.177	0.401	[-1.410, −0.870]	−8.847	<0.01
Kazakhstan	3.256	0.698	4.261	0.313	[-1.349, −0.662]	−6.053	<0.01

In both the interviews and reflections, almost all students confirmed the development of their innovation and entrepreneurship skills, aligning with the results observed in the quantitative data. The thematic analysis of interviews and reflections for RQ2 (Table 6) identified two main themes. First, participants reported enhanced confidence and communication, including increased cross-cultural interaction and greater ability to present and promote ideas. Second, students emphasized the development of creativity and innovation, particularly by integrating traditional culture with AI and designing culturally adaptive solutions. These skills highlight the role of cross-cultural collaboration in expanding students’ cognitive flexibility, enabling them to integrate diverse viewpoints into their problem-solving strategies. The immersive nature of the learning activities provided participants with a unique opportunity to apply PAs within intercultural contexts, prompting them to think creatively and innovate.

OPEN IN VIEWER

Table 6.

Key Themes of innovation and entrepreneurship (RQ2) From Qualitative Analysis

Main theme	Sub-theme	Frequency	Source
1. Enhanced confidence and communication as entrepreneurial skills	1.1 increased confidence in cross-cultural communication	56	ID 5: I Feel more confident communicating and interacting with people from our countries and cultures.
	1.2 improved ability to present and promote ideas	46	ID 19: I Feel more confident to advertise and explain technologies like this peer agent as a good solution for cultural context.
2. Development of creativity and innovation in educational and cultural contexts	2.1 generating creative ideas by integrating traditional culture with AI	52	ID 33: I Learned to integrate traditional narratives with innovative technology… which allowed the project to be successful and resonate with diverse audiences.
	2.2 designing culturally adaptive and inclusive solutions	49	ID 42: I Recognized the importance of integrating traditional cultural elements into modern business models. This led me to innovate a product line.

Positive Perceptions of Technologies Used in Cross-Cultural Activities

To address RQ3, the results presented in Table 7 showed that all measured dimensions of technology perception, credibility of AI, involvement, sensory fidelity, and sense of immersion, exhibited high mean scores, indicating overall positive perceptions among participants.

OPEN IN VIEWER

Table 7.

Results of technology Perception

Dimension	Neutral	Agree	Strongly agree	Max	Min	Mean	SD
Credibility of AI				5	3	4.248	0.390
China	20.833%	44.444%	34.722%	5	3	4.132	0.344
Northern Cyprus	8.333%	48.333%	43.333%	5	3	4.350	0.432
Kazakhstan	6.250%	59.722%	34.028%	5	3	4.278	0.383
Involvement				5	3	4.231	0.430
China	18.452%	37.500%	44.048%	5	3	4.256	0.292
Northern Cyprus	17.143%	44.286%	38.571%	5	3	4.214	0.521
Kazakhstan	16.072%	43.452%	40.476%	5	2	4.220	0.478
Sensory fidelity				5	3	4.203	0.441
China	20.000%	55.000%	25.000%	5	3	4.050	0.388
Northern Cyprus	10.000%	48.000%	42.000%	5	3	4.320	0.456
Kazakhstan	10.000%	54.167%	35.833%	5	3	4.258	0.451
Sense of immersion				5	3	4.306	0.423
China	9.167%	47.500%	43.333%	5	3	4.342	0.267
Northern Cyprus	8.000%	52.000%	40.000%	5	3	4.320	0.397
Kazakhstan	12.500%	48.333%	39.167%	5	3	4.258	0.561

The credibility of AI dimension showed a mean score of 4.248 (SD = 0.390), suggesting a relatively high level of confidence in the PAs used in the study. Notably, over 80% of participants across all three countries rated the credibility of PAs as high. In the involvement dimension, participants reported a mean score of 4.231 (SD = 0.430), suggesting active engagement during the learning activities. More than 80% of participants perceived their level of involvement in the interactive VR environment as above average. Similarly, the sensory fidelity dimension, which measures the realism of sensory input, achieved a mean score of 4.203 (SD = 0.441), highlighting the immersive and realistic nature of the virtual reality environment. A majority of participants expressed agreement or strong agreement regarding the high sensory fidelity of the system. Lastly, the sense of immersion dimension recorded the highest mean score at 4.306 (SD = 0.423), indicating that participants experienced a highly immersive environment during the activities. Only about 10% of students held a neutral stance toward their sense of immersion, while the vast majority rated it positively.

The qualitative findings, based on contributions from all 68 participants, were generated through thematic analysis of 62 interviews and 59 reflections, as not all participants completed both qualitative components. The results further highlighted students’ positive experiences with the use of PAs in VR environments (Table 8). The themes of trustworthiness and reliability reflected participants’ perceptions of AI’s accuracy, neutrality, and safety. Involvement was demonstrated through active engagement, particularly through 360° videos, while sensory experience emphasized the audio-visual realism in VR. Finally, depth of immersion was reflected in participants’ emotional engagement with cultural events. This feedback underscores the immersive nature of the VR environment and the engaging role that PAs played in enhancing students’ learning experiences.

OPEN IN VIEWER

Table 8.

Key Themes of technology Perceptions (RQ3) From Qualitative Analysis

Main theme	Sub-theme	Frequency	Source
1. Trustworthiness and reliability	1.1 perceived reliability and accuracy	53	ID 41: When I asked about the significance of the festival, the AI agent offered accurate and detailed explanations.
	1.2 trust shaped by neutrality and consistency	48	ID 12: The AI agent is reliable. Because this application has no emotions, it does not judge us on any subject.
	1.3 safety, confidence, and confidentiality	49	ID 44: I Generally felt safe and confident using the AI agent. However, I had some concerns about confidentiality, as I was unsure how my data was being handled.
2. Involvement	2.1 active task engagement in 360° video	61	ID 38: I Felt truly engaged as if I had personally visited a cultural site, where every detail of the surroundings became visible in the recording.
3. Sensory experience	3.1 audio-visual realism in VR scenes	64	ID 29: In the virtual market, I could hear the bustling sounds of vendors and see the vibrant colors, making it feel like I was really there
4. Depth of immersion	4.1 Event-based deep immersion	63	ID 55: During the virtual Chinese new Year parade, I was completely captivated by the lanterns, dragon dances, and drumming, feeling as though I was truly at the scene.

Understanding the Effect of Cross-Cultural Learning on Global AI Literacy

Concerning RQ1, our findings suggest that participation in PAs and VR-supported cross-cultural learning activities was associated with observed gains in global AI literacy. This result aligns with the studies of Li et al. (2024) and Tafazoli (2024b), which demonstrated the benefits of large language models and VR in cross-cultural learning contexts. However, our study differs in that it encouraged students to utilize PAs for cross-cultural information exchange and integrated PAs into VR environments to potentially facilitate deeper engagement with foreign cultures, which may have contributed to enhancing global AI literacy.

The observed patterns may be related to the instructional design grounded in Bloom’s cognitive taxonomy. During the study, students gained foundational knowledge of AI, such as the typical applications of computer vision technologies in their local context. Expert-led lectures provided students with a deeper understanding of the principles and mechanisms behind AI technologies. Building on this foundation, students were able to analyze and synthesize information generated through interactions with PAs to complete cross-cultural learning activities. This process may have supported students’ understanding of AI and its applications in cultural contexts, particularly in relation to applying PAs to cultural learning tasks. After viewing 360-degree videos created by peers from other countries, students engaged in reflective interactions with PAs to evaluate the development of their global AI literacy. This reflective practice not only deepened their understanding of AI applications but also allowed them to critically assess their progress in leveraging AI within cultural contexts.

Specifically, AI knowledge demonstrated significant gains following the PA- and VR-supported cross-cultural learning activities. This finding aligns with the principles of constructivist learning theory, which emphasizes that learners construct knowledge more effectively when they engage in authentic, context-rich environments (Hwang et al., 2026; Narayan et al., 2013; Wang et al., 2025). In this study, the VR environment provided students with immersive and experiential opportunities to explore AI applications in various cultural contexts. By interacting with real-world examples, such as the use of computer vision or robotics in different countries, students were able to connect theoretical AI concepts to their practical applications. Through these activities, students bridged the gap between theoretical knowledge and practical implementation, thereby reinforcing this dimension of their global AI literacy.

AI skills demonstrated statistically significant growth following participation in the learning activities. Previous research has also highlighted the effectiveness of using AI tools or AI agents in classrooms to enhance students’ AI skills (Morandini et al., 2023; Ng, Lee, et al., 2023). However, unlike these prior studies, our study engaged students with PAs designed for specific functions, enabling them to complete tasks such as writing cultural video scripts. Furthermore, students interacted with PAs integrated into VR environments, which not only provided deeper insights into foreign cultures but also enriched their understanding of AI concepts and their practical applications in diverse cultural contexts. Across these learning activities, students appeared to develop greater proficiency in applying AI tools to solve problems encountered during the learning process.

In the cross-cultural learning experiences, students reported statistically significant gains in their AI attitudes. They also expressed a keen interest in experimenting with various types of AI systems. Previous research has demonstrated the positive learning experiences provided by VR environments, such as in gamified learning (Derakhshan et al., 2024) and language learning (Lin & Lan, 2015). Integrating PAs into VR environments appears to be associated with students’ positive attitudes toward AI usage. During interviews and reflections, students highlighted that this was their first time participating in such an innovative cross-cultural learning activity. They expressed openness to using AI to enhance cultural learning experiences, further reflecting students’ positive perceptions of this integration.

AI ethics, another critical dimension of AI learning, also showed statistically significant improvement over the study period. Previous studies suggest that ethics is closely related to the safe and responsible application of AI systems (Cheng et al., 2021; Murugan et al., 2025; Peters et al., 2020). In our study, students emphasized the importance of AI systems performing reliably and safely. Throughout the cross-cultural learning activities, students interacted with PAs to assist in script writing, rather than relying on the PAs to fabricate content. This specific configuration of the PAs was intended to reduce the likelihood of hallucinated content during interactions, reinforcing students’ understanding of ethical AI usage.

The results indicate a significant improvement in students’ ability to apply AI in cultural contexts, which may be related to two key factors. First, students interacted with PAs, which provided detailed, context-specific information on how AI is utilized within their own local cultures. Second, by watching 360-degree videos, students explored the applications of AI in foreign cultural contexts. These immersive experiences, combined with interactions with PAs in the VR environment, allowed students to acquire authentic and reliable information about the cross-cultural use of AI both domestically and internationally. While prior research has highlighted the role of AI in preserving and promoting intangible cultural heritage (Zhang & Jing, 2022), there has been limited exploration of how AI can be effectively integrated into immersive, cross-cultural learning environments to foster practical applications of AI. This study bridges the gap by integrating theoretical knowledge with practical applications, highlighting how students engaged with real-world problem-solving scenarios using AI. For instance, students learned how computer vision technology can be applied in systems like food delivery robots.

In summary, the findings of this study make a significant contribution to the field of global AI literacy education. By integrating PAs into cross-cultural learning environments and utilizing immersive VR technologies, this study demonstrates how AI literacy can be effectively cultivated in global contexts. Compared with previous studies that focused primarily on AI literacy in general, the use of PAs in this study not only facilitated students’ engagement with AI but also enabled them to apply AI concepts and skills in culturally diverse settings. This integration of technological proficiency with intercultural competence highlights the development of global AI literacy, which extends beyond technical knowledge to encompass the capacity to leverage AI responsibly and effectively across cultural boundaries.

Understanding the Effect of Cross-Cultural Learning on Innovation and Entrepreneurship

With respect to RQ2, participants engaged in verbal interactions with PAs to gain deeper insights into cultural differences, which may have inspired innovative ideas and been associated with the enhancement of their innovation-related thinking and entrepreneurship-oriented motivation. Our findings align with previous research, which highlights how technology and digitalization are reshaping entrepreneurial and innovation practices (Si et al., 2023). Both short-term and long-term courses utilizing AR and VR have been identified as new-age learning mechanisms that effectively prepare youth for innovation and entrepreneurship (Rath et al., 2019). For instance, education based on VR technology, guided by the theory of planned behavior, has been shown to significantly enhance students’ willingness to engage in innovation and entrepreneurship (Dong & Tu, 2021). This cross-cultural learning activity may have supported entrepreneurship-related outcomes primarily by strengthening motivational and cognitive precursors of entrepreneurial intention, rather than by directly developing entrepreneurial skills. Specifically, the combination of cross-cultural tasks, PA-supported guidance, and immersive VR experiences may have enhanced students’ perceived autonomy, competence, and confidence in applying ideas across unfamiliar contexts, which are core components emphasized in self-determination theory (Deci et al., 2017).

Our carefully designed learning activities introduced students to novel cultural contexts and the application of AI within these settings. This experiential learning approach, delivered through VR environments, illustrated how technology-supported methods may support the generation of creative ideas among students. On one hand, students were able to recognize the diverse applications of AI across different cultural scenarios from multiple perspectives. On the other hand, the cross-cultural learning framework sparked innovative thinking, such as the conceptualization of multi-technology integration applications.

The findings also highlight participants’ increased confidence and motivation toward entrepreneurship, with many expressing satisfaction in pursuing entrepreneurial endeavours. This can be understood through the lens of self-determination theory, which emphasizes the importance of intrinsic motivation in achieving personal growth and competence. The cross-cultural learning activities, supported by PAs in VR environments, were associated with students’ active engagement with real-world problems, and their reported sense of autonomy and accomplishment (Shadiev et al., 2026). For instance, by identifying innovative ideas and applying them within cultural and AI-related contexts, students perceived entrepreneurship as a fulfilling and achievable goal.

Overall, the findings suggest that the integration of PAs within VR-supported cross-cultural learning activities was associated with authentic opportunities for students to generate creative ideas, strengthen entrepreneurial confidence, and explore innovative solutions in culturally diverse settings. The carefully structured design of the learning activities, grounded in Bloom’s frameworks, ensured that students could actively connect AI applications with real-world cultural contexts.

Understanding the Positive Perception of Technologies Used in Cross-Cultural Activities

For RQ3, students expressed positive perceptions of the use of PAs and VR in the learning activities. Previous research has also highlighted predominantly favorable perceptions of VR and large language models when applied in educational settings (Jeon et al., 2024; Khukalenko et al., 2022; Lim et al., 2021). This study found that embedding PAs into VR environments was associated with more positive perceptions of technology use. The immersive and realistic experience provided by VR allowed students to feel as though they were truly present in a foreign country, engaging with its culture. Moreover, for almost all students, this was their first experience of engaging in cultural learning within a technologically rich virtual environment. The constant guidance and support provided by PAs within the VR environment contributed to their appreciation of the convenience and effectiveness of intelligent technologies.

Another important concern of this study is the creation of credible PAs. During configuration, clear boundaries and contextual backgrounds were established, tailored to the cross-cultural learning activities involving students from three countries: Kazakhstan, Northern Cyprus, and China. This approach significantly reduced the likelihood of PAs generating hallucinated content during interactions. As a result, students trusted the cultural knowledge provided by the PAs. Through ongoing dialogue, they were able not only to acquire factual knowledge but also to develop deeper understanding, analysis, and insights into foreign cultures.

Taken together, these findings provide a strong indication that students’ positive perceptions were not incidental but closely tied to the structured design of the intervention. The systematic integration of PAs and VR ensured consistency, credibility, and meaningful engagement, which enhances confidence in the robustness of the results. While further studies are needed to broaden the scope, the present evidence offers a reliable basis for understanding how intelligent technologies can shape learners’ perceptions in cross-cultural contexts.

Despite these encouraging findings, several alternative explanations should be considered when interpreting the results. First, although participants were drawn from three countries, formal subgroup comparisons were not conducted due to the limited sample size within each group. Second, as the intervention was embedded within a semester-long “AI in Education” course and no follow-up measures were collected, the persistence of the observed outcomes over time remains unclear (Shi et al., 2025). Finally, while the VR- and PA-supported activities constituted a distinct, task-centered component of the course, their effects cannot be fully disentangled from other instructional elements.