Three-Dimensional Modeling and AI-Assisted Contextual Narratives in Digital Heritage Education: Course for Enhancing Design Skill, Cultural Awareness, and User Experience

Abstract

This study introduces an educational framework that merges 3D modeling with AI-assisted narrative interaction to apply digital technology in cultural heritage education, exemplified by an ancient carriage culture. Through immersive tasks and contextual narratives, the course notably improved learners’ professional skills and cultural awareness. Experimental results revealed significant knowledge acquisition among participants post-engagement. Additionally, the user experience improved, with increased satisfaction in the narrative interaction design course. These enhancements led to heightened interest in cultural heritage and deeper knowledge acquisition. Utilizing Norman’s three-layer interaction model, Ryan’s contextual narrative theory, and Falk and Dierking’s museum learning experience model, the study developed a systematic course for multi-sensory design and contextual interaction, confirming the positive impact of multimodal interaction on learning outcomes. This research provides theoretical support for the digital transformation of cultural education and practical examples for educational practitioners and cultural institutions to implement in virtual presentations and online learning.

1. Introduction

1.1. Research Background

The rapid development of digital technology has transformed the preservation and dissemination of cultural heritage. In recent years, virtual reality (VR), augmented reality (AR), and 3D modeling have provided new possibilities for the protection and communication of cultural assets [1,2,3]. Virtual museums and digital exhibitions have become important tools for cultural organizations to engage the public [4,5,6]. These approaches enhance the attractiveness of cultural content and enable users to interact with heritage in ways that foster deeper understanding of historical and cultural contexts [6,7,8,9].

However, the widespread adoption of digital technology in cultural communication has not been fully leveraged in the realm of education, as evidenced by several studies. Majda et al. highlighted the untapped potential of digital technology for personalized and cross-temporal cultural education [10]. But its impact is constrained by educators’ digital literacy and instructional design skills. The study underscores the need to translate abstract cultural concepts into engaging interactive learning experiences [10]. This research offers a practical framework for assessing the efficacy of cultural education and advocates for a more profound integration of technology with educational content. Ott, Pozzi et al. argue that conventional digital tools like 3D modeling and augmented reality often lack depth in exploring cultural nuances such as philosophical ideas and social values in heritage education. They propose the development of interactive learning resources through interdisciplinary collaboration involving history, education, and computer science to engage students in the reinterpretation of cultural heritage [11]. Also, Mayer highlighted the limitations of using mobile technology solely for information delivery, such as one-way video playback, without integrating cognitive strategy guidance, like problem-driven knowledge integration [12]. This approach hinders the enhancement of deep level learning outcomes, including knowledge transfer and metacognitive abilities. In the context of cultural education, the use of mobile technology may fail to effectively connect stimulating interest with enhancing cognition due to the absence of localized content, such as inadequate exploration of traditional cultural symbols, and insufficient interaction design, leading to a lack of immersive cultural experiences. This situation reflects the challenge of bridging the gap between the potential of technology and its educational impact [12]. These studies suggest that the mere incorporation of digital technology into cultural education is not viable. The effective utilization of these technologies to enhance learning engagement, elevate cultural cognitive abilities, and ultimately boost educational efficacy is a pivotal issue requiring immediate attention.

Traditional cultural education models often rely on unidirectional information transfer, such as teaching through text, pictures, or simplified explanations [13]. While this approach can deliver basic knowledge, it lacks interactivity and immersion, hindering deep engagement and emotional resonance in learners [14,15]. In contrast, digital learning environments can enhance the immersive and memorable effects of learning through multi-sensory experiences and dynamic interactions, offering new possibilities for cultural heritage education [16,17].

The emerging field of immersive learning and interactive narratives has garnered significant attention in academia and educational practice. Dede posits that immersive learning environments enhance learners’ knowledge internalization and interest through dynamic scenarios and multi-sensory experiences [18]. Participatory Design (PD) further enables learners to better understand complex cultural contexts by deeply involving users in the experience [19]. Additionally, Ryan’s theory of contextualized narratives emphasizes that immersion and motivation to learn can be significantly improved by making the user a participant and co-creator of the narrative [20]. Moreover, Norman’s three-layer model of interaction suggests that interactive experiences should be designed incrementally, addressing the user’s perceptual, operational, and cognitive needs [21]. These theories and methods provide a robust theoretical foundation for narrative interaction design and user experience research in cultural heritage education.

1.2. Research Questions

This study addresses the challenges of limited interactivity and immersion in cultural education by integrating immersive methods with interactive narrative design. It poses three core research questions:

RQ1: How can 3D modeling and narrative interaction design enhance learners’ knowledge acquisition? This is measured by changes in learners’ accuracy on cultural knowledge tests and strengthened cultural identity.

RQ2: How can these methods improve learners’ skill application? This is assessed through increased accuracy in model building, more efficient manipulation time, and improved narrative interaction design skills.

RQ3: How do 3D modeling and narrative interaction design enhance learners’ user experience? This involves examining the effects of immersive experiences on motivation, task completion rates, and overall learning satisfaction.

These questions are framed around three dimensions: technology application, cultural dissemination, and user experience, aiming to explore the comprehensive role of digital technology in cultural education.

1.3. Research Significances and Objectives

This study addresses existing research gaps by developing a formal curriculum for cultural heritage education, integrating 3D modeling and narrative interaction design to examine the role of digital technology, focusing on ancient carriage culture. The study aims to accomplish several goals: first, improve learners’ professional skills, such as 3D modeling accuracy and efficiency, and cultural awareness, including historical knowledge and cultural memory depth; second, enhance learner engagement and motivation through immersive tasks and dynamic narratives; third, explore the integration of digital technology with cultural education to offer an innovative model for cultural heritage preservation and educational dissemination. By designing immersive learning scenarios and multimodal interactive tasks, the study offers novel insights for improving cultural education and provides theoretical and practical guidance for educators and cultural institutions in digital transformation applications. Additionally, the study employs the research methods of one-way analysis of variance and semi-structured interviews to empirically validate the course design, addressing empirical gaps in cultural heritage education research.

2. Relevant Theories

2.1. Theoretical Framework

2.1.1. Norman’s Three-Layer Interaction Model

Norman in “Emotional Design” posits that interaction design should enhance user experience across three levels—visceral, behavioral, and reflective—addressing perceptual, operational, and cognitive needs [21]. This framework can be integrated into course design to refine its effectiveness.

2.1.2. Ryan’s Contextual Narrative Theory

The contextual narrative theory proposed by Marie-Laure Ryan emphasizes the importance of user selectivity and contextualized narratives in enhancing user immersion [20]. Ryan’s theory highlights the construction of virtual contexts in which stories take place, arguing that narrative creates multi-dimensional possible worlds through linguistic symbols, endowing stories with unique spatial, temporal, and social contexts. The theory posits that context is not merely a backdrop, but a key element that influences the logic of characters’ actions, the trajectory of plot development, and the reader’s cognitive and emotional engagement, which can drive the emergence and evolution of narrative events. Ryan proposes a dynamic interactive relationship between context, narrator, characters, and readers, whereby readers engage with the narrative context through cognitive reasoning to discern the deeper meaning of the story [20].

2.1.3. Falk and Dierking’s Museum Experience Model of Learning

This study expands the theoretical framework by introducing the Museum Experience Model of learning by Falk and Dierking, which posits that learning arises from the interplay of individual, social, and physical contexts [22]. This model challenges traditional views by framing museum learning as a dynamic interaction among individual, physical, social, and cultural environments, rather than a simple knowledge transfer.

2.2. Limitations of Related Research

Research indicates that immersive learning environments and interactive narrative design hold significant promise for cultural communication and education. The AR-based cultural heritage application “Street Museum” has effectively engaged numerous users in exploring Trondheim’s historical background [23]. Similarly, Time Explorer deepens users’ understanding of historical events through dynamic narratives and interactions [8]. A recent study by Yu and Phillips provides insights into system-building models for narrative interaction [24]. These studies highlight the benefits of immersive techniques in enhancing memory and cultural awareness. However, many of these approaches rely on static digital presentations and are not integrated with formal learning programs. Additionally, there is a tendency to prioritize cultural dissemination over assessing learner skill enhancement and learning outcomes.

Interactive narratives and multimodal design are pivotal in enhancing learning experiences. Zhou et al. demonstrated that multimodal interaction, integrating visual, tactile, and auditory elements, significantly boosts user immersion and memory retention [25]. Dede highlighted that multi-sensory design fosters deeper cultural understanding in task-driven environments [18]. Despite these insights, current research predominantly focuses on single-sensory or static narratives, lacking systematic approaches to dynamic, multi-sensory integration. Additionally, there is insufficient exploration of quantitative assessments of learning outcomes, such as skill enhancement, learning effectiveness, and satisfaction [26].

3. Methods

3.1. Design Heritage Selection: Ancient Chinese Carriage

The selection of the ancient Chinese carriage as a case study is motivated by two primary factors. Firstly, extensive research has underscored that ancient Chinese carriages exemplify the fusion of social technology and culture in ancient China. These carriages played pivotal roles in military operations, rituals, and transportation, profoundly shaping the social development of ancient China [27,28,29,30]. Despite the existing literature on digital design focusing on ancient carriages, there remains a dearth of studies in this area. Secondly, conventional methods of presenting traditional cultural artifacts with utilitarian functions, such as ancient Chinese carriages, tend to be static. Typically, these artifacts are showcased in display cases with explanations limited to their appearance, materials, and basic historical context. While such displays effectively convey fundamental information about the artifacts, they fall short in illustrating their functionality and practical applications [31,32]. For instance, static displays of ancient vehicles allow observers to appreciate their physical attributes and materials but fail to elucidate their practical uses in military, ritual, or transportation contexts. These static displays lack the ability to recreate dynamic scenarios, such as depicting how war chariots maneuvered on battlefields, navigated rugged terrains with stability and flexibility, or engaged in combat actions. In addition, static display cabinets fail to convey the dynamic scenes and cultural symbolism of ancient carriages used in ritual activities, thus impeding the audience’s comprehensive grasp of the historical context and cultural significance of these artifacts [2,33,34]. Another reason is that this is a research project of Jiangnan University; this study represents a pivotal initiative in Jiangsu Province, focusing on the digital innovative design of educational materials related to ancient vehicles as the core design approach.

3.2. Objective and Scope of Research Method

This study seeks to combine 3D modeling with narrative interaction to create educational materials for cultural heritage courses and to develop corresponding curricula. In the realm of 3D modeling, leveraging insights from scholarly works such as “Ancient Chinese Vehicles and Horse Equipment”, “Research on Vehicles and Their Design Thinking in the Qin and Han Dynasties”, and “Analysis of the Functional Characteristics of Ancient Chinese Vehicles”1, and validated through consultations with experts, a 3D model representing cultural heritage is constructed. Regarding narrative interaction, drawing on the frameworks proposed by Murray, Ryan, Falk, and Dierking, narrative scenarios revolving around cultural heritage are crafted, such as depictions of wheel assembly and ceremonial processions in the context of ancient vehicles. Additionally, Artificial Intelligence-Generated Content (AIGC) technology is integrated to automate text generation and story branching, facilitating real-time engagement, addressing students’ inquiries, thereby fostering classroom engagement, and reducing development expenses.

The curriculum design and implementation section emphasize multimodal interaction tasks, AI applications, 3D model interaction, and role-playing. Multimodal interaction tasks incorporate vision, hearing, and gestures. AI applications involve AI-generated narrative texts and plots, historical scene simulation, personalized learning paths, and cultural heritage immersion. Three-dimensional model interaction enables detailed exploration of virtual cultural heritage [35]. Role-playing activities are organized around the multi-line plot of transport vehicles across historical periods to enhance learners’ engagement with cultural and historical contexts. Finally, a prototype is developed to assess students’ skills, user satisfaction, and cultural dissemination efficiency. Classroom user experience experiments validate the effectiveness of the curriculum design and courseware.

3.3. Literature Research and Modeling

This study integrates literature analysis and expert validation to create a precise 3D model for historical authenticity. The literature analysis involves extensive collection of historical documents, archaeological reports, and museum collections to examine the structure, function, and cultural context of ancient carriages (Figure 1). For instance, design features of Han Dynasty and Warring States carriages were derived from classical texts like the Kaogongji and Han Shu2, alongside archaeological findings [29,36] (Research on Carriages and Their Design Thinking in Qin and Han Periods, n.d.3). Studies affirm that accurate documentation underpins virtual cultural reconstruction [37]. For modeling, software like Rhino or Blender was employed to develop detailed 3D models akin to techniques used in animation and game production and virtual museum exhibitions [38].

3.4. Three-Dimensional Modeling and Solutions to Technical Difficulties

The 3D modeling process comprises three main steps: literature research, verification, and technical implementation. Initially, researchers extract historical parameters from sources like “Kao Gong Ji”4 and “Ancient Vehicle, Horse Harness” [36]. These parameters are then compared with calibrated dimensions of ancient vehicles from historical records. For instance, the length of a single-shaft carriage yoke from the Han Dynasty was adjusted to 2.1 m based on findings from the “Mancheng Han Tombs”5. Subsequently, polygon modeling is performed using Rhino (software is developed by Robert McNeel & Associates.) and Blender (software is developed by Luxion. The headquarters of Luxion is located in Tustin, CA, USA) to construct the vehicle’s structure, with material rendering executed through Unreal Engine 5.4.1 Developed by Epic Games. The company is headquartered in Cary, NC, USA. For example, modeling a Han Dynasty covered carriage involves drafting the frame in CAD 2020 (refers to AutoCAD developed by Autodesk. Autodesk is headquartered in San Francisco, CA, USA), importing it into Blender (it is an open-source 3D creation software developed by the Blender Foundation. The Blender Foundation is based in The Netherlands.) to carve the mortise-and-tenon joints, and rendering the bronze decoration in Key Shot (software is developed by Luxion. The headquarters of Luxion is located in Tustin, CA, USA), setting the reflectivity to 0.8 to mimic gilding effects.

The modeling process is constrained by fragmented historical data and missing vehicle and carriage materials from certain dynasties. For instance, the vehicle manufacturing technology of the Le Lang Commandery during the Yuan Dynasty is documented only in incomplete records found in “History of Yuan · Records of Vehicles and Apparel” 6. As a result, the number of wheel spokes in the model must be estimated by experts due to insufficient physical evidence and is ultimately set at 30, aligning with the Han Dynasty tradition of “thirty spokes share one hub”. The accuracy of archaeological artifacts presents further challenges. The bronze carriages and horses from the Mausoleum of the First Qin Emperor exhibit oxidation and deformation, resulting in an 8% discrepancy between measured sizes and historical records. In modeling, a decision was required between “faithfulness to cultural relics” and “historical restoration”, and researchers recommended limiting the error to within 5%. Additionally, conflicts among multiple data sources are common. “Kao Gong Ji” states the single-shaft carriage length as six feet and six inches, yet the measured length of the bronze carriages’ shaft is 2.3 m. Due to debate over the Zhou Chi unit conversion (1 Zhou Chi7 ≈ 23.1 cm or 24.8 cm), the model’s length was adjusted between 2.1 and 2.3 m, finally adopting the archaeological report’s measured value as the standard.

Technical bottlenecks present significant challenges. Balancing detail and rendering efficiency is particularly difficult. If the Kui dragon pattern on bronzeware is carved to match the original, the number of polygon faces exceeds 50,000, dropping the rendering frame rate below 30 fps. Using topology optimization algorithms, such as the Blender plugin Easy Bake, the model is reduced to under 10,000 faces, converting the three-dimensional relief to a flat texture. Cross-software compatibility issues are notable, importing the mortise-and-tenon structure from Rhino into Blender results in errors due to parametric node differences, requiring manual correction of 12 key nodes and increasing the time by 30%. For mobile experiences, models must be compressed to under 5MB, necessitating the simplification of the bronze bells’ movable structure on ancient vehicle carriages into a static model, while maintaining authenticity.

When discrepancies arise between the historical literature and technical implementation, a decision is made by a panel of three experts in ancient manufacturing and two interaction designers, including the study’s researchers. For instance, with the Han Dynasty’s “double-eaved” carriage design, experts recommend preserving it for historical accuracy. However, the texture resolution must be reduced from 4 K to 2 K to optimize performance, aligning with documented records.

3.5. Narrative Interaction Design

The course constructs a multi-stage narrative framework tracing the historical evolution of ancient carriages, enabling learners to examine the intricate relationship between carriages and culture across different historical periods. For instance, the Han Dynasty carriages module presents a series of dynamic tasks set against the Silk Road, a historically significant trade route. In one task, learners use virtual tools to repair issues such as damaged axles and loose wheels, drawing on their understanding of Han Dynasty carriage structures. This task necessitates a detailed study of the construction features of Han Dynasty carriages and supply wagons, including the lightweight design of carriages for long-distance travel and the large capacity of supply wagons for trade carriage, fostering a deep practical understanding of these ancient means of transport.

When selecting a cargo plan, learners must evaluate the demands of various regions along the Silk Road, alongside the cargo’s value, weight, and the carriage’s capacity. For instance, when transporting goods to the Western Region, it is crucial to recognize that items like silk and porcelain are valuable yet fragile. These goods require careful placement and protection within the carriage, along with sufficient supplies to ensure the safety of the caravan. This task immerses learners in the experience of a Han Dynasty caravan, highlighting the critical role of transportation in trade and cultural exchange, thereby deepening their understanding of trade culture and logistics. The historical data was consulted to generate the narrative script using an AI-powered text generation platform—Hugging Face. The user, as the protagonist, experienced various events and made subjective choices, which the Large Language Model (Meta-llama-3.1-70B-Instruct) then utilized to automatically produce multifaceted, extended content in response (Figure 2).

The development software, key features, functional modules, and design purpose can be viewed at Figure 3. In multi-dimensional interactive experience design, digital models and visual content are integrated using UI mapping software and 3D rendering engines like Unity 2021.3.21f1c1 or Unreal Engine 5.4. Cross-scene interactive pages facilitate user engagement in cultural interpretation through perspective switching, operation simulation, and plot selection. Additionally, 3D and AI models are linked to integrate sustainable design systems. By employing a modular architecture and a lightweight data management strategy, the reusability of digital assets and low-energy dissemination are enhanced. This approach constructs a virtual museum framework that supports continuous expansion, forming a closed loop for cultural heritage revitalization.

3.6. Application of Theoretical Models and Association with Research Questions

This study optimized curriculum design by integrating Norman’s three-layer interaction model, Ryan’s situational narrative theory, and Falk and Dierking’s museum learning experience model, focusing on interaction design, narrative construction, and learning experience.

3.6.1. Incorporating Norman’s Model

The Norman three-layer model, illustrated in Figure 4, constructs a comprehensive and sequential framework for learning experiences, spanning from the visceral layer’s UI design and scene rendering, through the behavioral layer’s task interaction mode, to the reflective layer’s cultural guidance. The visceral layer’s visual design captivates learners’ interest with historical elements, while the behavioral layer’s efficient interaction facilitates the seamless completion of complex tasks. The reflective layer encourages learners to derive cultural understanding from task completion by integrating situational narration with cultural background knowledge [21].

This study employs a hierarchical design comprising the visceral, behavioral, and reflective layers to create a “perception–operation–cognition” learning loop, addressing three core research questions (RQ).

For RQ1, concerning knowledge acquisition, internalization occurs at the reflective layer. By integrating cultural context, learners are encouraged to connect with historical scenarios. In the curriculum, after completing key tasks like ancient cart restoration or Han Dynasty transportation activities, the system seamlessly incorporates detailed cultural background explanations. For instance, it may present analyses of historical documents, such as the discussion in “Hanshu · Shihuo Zhi”8 on the trade-driven dissemination of vehicle technology, or delve into the historical significance of ancient carts, cultural exchanges, and the socioeconomic impact of transportation advancements. This retrospective approach aligns with Dewey’s reflective learning theory, enabling learners to extract knowledge from immersive experiences [39,40]. Through review and synthesis, it fosters deep knowledge internalization, facilitating a transformation from specific tasks to a broader understanding of history and culture.

In addressing RQ2 (skill application), skills are refined at the behavioral level. By optimizing interaction logic, professional data is converted into actionable parameters, allowing learners to acquire core skills such as interaction logic and 3D modeling techniques. The curriculum design offers a range of intuitive and efficient interaction methods, significantly enhancing task completion efficiency and engagement. Learners engage with virtual scenes, like a Han Dynasty single-shaft cart, using simple gestures—such as single-finger swipes to change perspective, pinch gestures to zoom, and taps to select objects [41,42]. These interactions directly translate archaeological data into operable parameters. For instance, when modeling or restoring a Han Dynasty single-shaft cart, learners must adhere to precise archaeological data, such as a 2.1 m cart shaft. Scaling, rotating, and assembling model components must meet exact accuracy standards to master the core logic and techniques of 3D modeling in practice.

For RQ3 concerning user experience, immersion is evoked at the visceral level through high-quality sensory engagement, achieved via visual and auditory stimuli. The curriculum’s design captivates learners with intricate UI elements and realistic dynamic scene rendering. The UI integrates ancient cultural motifs and artistic styles, featuring menus styled as unfolding scrolls with simple fonts, distinctive icons, and historically resonant colors. Traditional patterns enhance the ancient ambiance and reinforce cultural-emotional connections. Visual rendering employs the Unity engine to intricately depict ancient scenes, such as the sheen of bronze carriages and the dynamic interplay of light and shadow in desert settings. The main interface’s dynamic background offers combined visual and auditory stimulation. Auditory experiences, synchronized with scene dynamics—such as the fading of camel bells with distance—further intensify the sense of presence and immersion [7,43].

3.6.2. Incorporating Ryan’s Theory

Marie-Laure Ryan’s concept of “users as co-creators of narrative” enhances learner engagement through context-based task design. Empirical studies demonstrate that this approach deepens cultural cognition and boosts learning motivation, affirming the applicability of Ryan’s theory in cultural communication education and offering a fresh perspective on narrative-driven learning models [20]. Figure 5 illustrates the characters in the narrative text and their relationships within the interactive system, featuring two content rows: the upper row details the narrative characters, while the lower row outlines their corresponding elements in the interactive system.

This study employs Marie-Laure Ryan’s concept of “users as co-creators of narrative” to convert historical events into interactive plots, addressing RQ1 (depth of knowledge) and RQ3 (emotional resonance).

In addressing RQ1 concerning the depth of knowledge, the design facilitates in-depth learning by allowing user choice, exemplified in curriculum practice through selective narrative driven by dynamic tasks. For instance, in the Han Dynasty horse-drawn carriage transportation task, learners’ selection of different goods, like silk or porcelain, triggers distinct narrative branches and detailed explanations. Opting for “porcelain” unlocks background knowledge on “Han Dynasty pottery-making techniques and Western Regions’ demands”, compelling learners to engage with and apply knowledge of “vehicle load bearing and cargo protection” to successfully complete the transportation challenge. This design immerses users in specific cultural scenarios, such as Han Dynasty trade routes, where their decisions (e.g., selecting goods or carriage types) directly influence task difficulty and the path of knowledge acquisition. This approach enhances learners’ interest in the cultural context, sense of agency, and depth of reflection [44,45], ultimately deepening their understanding of historical knowledge.

For RQ3, addressing user experience and emotional resonance, the design leverages user decisions to propel the plot, creating an emotional interaction chain. The curriculum integrates advanced technologies, such as AI for generating dynamic plots, with contextualized narratives to enhance this effect. For example, in the task of constructing a vehicle in the Lelang Commandery during the Yuan Dynasty, the user’s pivotal decision—such as whether to employ the Xianbei ethnic group’s spoke-making technique—directly influences subsequent plot developments and task outcomes. Upon successfully integrating ancient vehicle manufacturing, the system rewards the user with positive feedback, including plot animations, achievement badges, and character appreciation, fostering a robust “choice–feedback” emotional connection. This allows users to experience the immediate consequences and a sense of achievement from their decisions. Similarly, in the Han Dynasty transport vehicle module, learners must select appropriate transport vehicles—such as the lightweight Ju carriage, the covered Zi carriage, or the Xi carriage suited for rugged terrain—based on different terrains (desert, mountain) and task objectives (rapid transport, protection of valuable items). Each choice not only alters story development and task difficulty but also deeply engages users in narrative construction. This design aligns with Ryan’s notion that “selectivity enhances immersion” [20], enabling learners to forge strong emotional ties with historical figures and contexts while actively shaping the narrative [8], thereby achieving profound emotional resonance.

3.6.3. Incorporating Falk and Dierking’s Museum Learning Experience Model

Falk and Dierking’s museum learning experience model was integrated into digital curriculum design, highlighting the combined influence of individual background, social interaction, and physical environment on learning. Experimental validation confirmed the theory’s practical value, identifying multi-sensory stimulation and situational interaction as crucial for enhancing learning outcomes (Figure 6).

This study utilized Falk and Dierking’s museum learning experience model, integrating the individuals’ background, social interaction, and physical environment to create immersive learning scenarios [22]. This approach directly addresses RQ1 (knowledge construction) and RQ3 (user experience—emotional resonance).

For Research Question 1 (knowledge construction), the design facilitated effective knowledge building by aligning hierarchical tasks with individual backgrounds. In practice, this involved setting AI-based tasks tailored to the interests and knowledge levels of diverse learners. For instance, when novices completed tasks like “basic assembly of a single-shaft carriage in the Han Dynasty”, the system automatically highlighted key structures such as mortise-and-tenon positions to ease cognitive load. In contrast, advanced learners tackled complex tasks like “comparison of cart-making techniques in the Leilang Commandery during the Yuan Dynasty”, requiring them to independently consult sources such as “Kao Gong Ji” for analysis. Similarly, in the cargo-loading task set in an oasis scene, novices could follow system-recommended plans, whereas advanced learners had to independently plan the arrangement and loading sequence to achieve a greater sense of accomplishment. This adaptive approach to individual differences boosted learner engagement and motivation across levels [12], ensured task difficulty aligned with knowledge levels, and effectively guided knowledge construction.

For RQ3, concerning user experience and emotional resonance, the design amplified emotional resonance by meticulously crafting the physical environment and encouraging social interaction. The curriculum was set against typical Silk Road scenes—deserts, oases, and markets—using high-precision AI-generated images to create dynamic environments featuring undulating sand dunes, shimmering oases, and bustling markets. Multimodal feedback was incorporated, including dynamic lighting and shadow effects, background sounds like camel bells and caravan noises, and visual effects that allowed users to “feel” the flying sand and dust in the desert scene. This thoughtfully designed environment fostered a deeply immersive cultural atmosphere through multi-sensory stimulation of touch, vision, and hearing, significantly enhancing learners’ immersion and emotional engagement [46]. This aligns with the model’s assertion that “multi-sensory stimulation enhances memory and perception” [22]. The curriculum incorporated social-situation tasks emphasizing collaboration, allowing learners to partner with an AI-generated content assistant or other online peers. For instance, in the group task of configuring ancient cart caravans on the Silk Road during the Tang Dynasty, participants engaged in joint discussions and decision-making. In a market scene, learners could negotiate with virtual merchants and collaborate with peers to devise optimal transport vehicle configurations. This interactive approach not only fostered collaboration but also aligned with the social learning theory principle that “knowledge construction is promoted through interaction” [47]. Through this collaborative process, interactive experiences and emotional connections were strengthened. The integration of individual background, social interaction, and physical environment effectively addressed the research objectives of knowledge construction (RQ1) and emotional resonance (RQ3) by creating immersive and personalized learning scenarios.

4. Experimental Research

4.1. Participants and Classroom Tasks

Participants

The experimenter, serving as a lecturer, integrated this study into a 10-week course at Jiangnan University (China), involving 33 design students as participants. Among them, there were 6 people with rich ancient carriage history knowledge, 4 people with relatively rich knowledge, 8 people with ordinary knowledge, and 15 people with no knowledge (see

Table 1. Demographic information collected from participants.

	Descriptive Information
Age	20–30
Gender	17 males, 16 females
Occupation	All “postgraduate students researching ancient Chinese carriages—design major students”
Understanding of ancient Chinese carriages’ history Understanding of 3D modeling	Rich (6), relative (4), average (8), none (15)   All 33 design student have learned 3D modeling technique.

). All participants were fully informed and provided ethical consent. The course included training in 3D modeling and narrative interaction. Post-training, students collaborated in groups to design and develop 3D narrative interaction systems focused on ancient carriages from various dynasties and themes, resulting in seven distinct projects. At the course’s conclusion, the effectiveness of these systems and the teaching approach was assessed through student presentations of their interactive cultural systems and feedback gathered via questionnaires and interviews with participants.

The study initially assessed the baseline historical knowledge of Chinese carriages among 33 participants using a pre-test comprising 10 objective questions from “Kao Gong Ji” and “Ancient Vehicle, Carriage, Horse Gear”, [36] with a maximum score of 10. The experimental group (n = 17) had a mean score of 4.82 (SD = 1.25), while the control group (n = 16) scored 4.79 (SD = 1.32). An independent samples t-test revealed no significant difference between the groups’ pre-test scores (t(31) = 0.08, p = 0.94), ruling out initial cognitive disparities. Participants in the experimental group averaged 22.5 ± 2.1 years of age, compared to 23.1 ± 1.8 years in the control group. All participants, possessing a design background, demonstrated full proficiency in 3D modeling. Statistical analyses confirmed no significant differences in age, educational background, or technology use, ensuring sample homogeneity.

The experiment’s control variables included the following: students’ majors, with all participants majoring in design at Jiangnan University, to ensure uniformity in professional backgrounds and minimize the impact of professional differences on the results. The course length was standardized to 10 weeks, ensuring all students had equal time for training and relevant operations. The production system theme required students to work in groups to create 3D narrative interactive systems focused on ancient carriages from various dynasties. This thematic constraint ensured the experiment centered on specific content, reducing interference from excessive thematic variation. The experiment standardizes task difficulty through consistent tasks. All participants engage in the “Innovative Educational Framework for 3D Modeling and Narrative Interaction around Ancient Integration” learning task, completed within 45 to 60 min with uniform instructions and guidelines. During the pre-experiment phase, the intensity of 3D model interaction prompts is calibrated to achieve completion rates exceeding 85% for assembly, ancient vehicle background use, and narrative tasks. Chi-square tests revealed no significant differences in task stage completion rates among the seven student groups, confirming uniform task difficulty across diverse backgrounds. The intervention’s effectiveness is evaluated through pre-test and post-test differences. This design balances initial cognition via pre-test screening and random grouping, standardizing the task process per the study’s learning model to satisfy experimental control variable requirements across groups.

4.2. Specific Study Cases

Case 1: Han Dynasty Transport Carriage Module (Figure 7): Set against the expansive backdrop of the Silk Road, this module features a series of dynamic tasks. In the vehicle repair task, learners must accurately address issues like axle damage and wheel looseness using virtual tools, grounded in their understanding of Han Dynasty transport carriage structures. This involves a detailed study of Ju and Zi carriages, focusing on the Ju carriage’s lightweight design for long-distance travel and the Zi carriage’s large load capacity for trade. In the cargo-loading task, learners must consider regional needs along the Silk Road, as well as the value, weight, and carrying capacity of goods. For example, when selecting goods for transport to the western regions, it is crucial to recognize that delicate, high-value items like silk and porcelain require careful placement and protection within the carriage. Additionally, including daily necessities is essential to support the caravan’s journey. This task design immerses learners in the experience of a Han Dynasty caravan, highlighting the critical role of transportation in trade and cultural exchange, thereby deepening their understanding of trade practices and transportation tools.

Case 2: The Yuan Dynasty Lele Carriage Module (Figure 7) immerses learners in the vibrant cultural exchanges of the era, illustrating the fusion of diverse cultures. In the item-loading task, learners must accurately place items like tea, spices, and furs onto Lele carriages or ox-carriages, reflecting the trade characteristics and ethnic cultural features of the Yuan Dynasty. These items, sourced from various ethnic regions, highlight the distinctive production and lifestyle traits of each group. This task provides insights into the types and scale of trade among different ethnicities during the Yuan Dynasty. In the caravan-organization task, learners must strategically arrange the sequence and spacing of carriages, adhering to the organizational norms and cultural practices of Yuan transport caravans. For instance, considering the grassland terrain and carriage functionalities, Lele carriages, being more adaptable, lead the caravan, followed by ox-carriages carrying heavier goods, with escort carriages positioned to ensure the caravan’s safety. These tasks enable learners to thoroughly investigate the integration and exchange of multi-ethnic cultures in the transportation sector of the Yuan Dynasty, highlighting the significant impact of cultural diversity on trade transportation.

4.3. Experimental Steps

To thoroughly and scientifically assess the effectiveness of training and system production in 3D modeling and narrative interaction courses, a multi-dimensional evaluation framework can be developed:

The knowledge mastery dimension evaluates students through a theoretical test on topics such as ancient carriage structures, historical and cultural contexts, principles of 3D modeling technology, and narrative interaction design theories. A one-way analysis of variance was conducted for group comparisons involving three or more groups. Results included the mean (M), standard deviation (SD), test statistic (t/F), p-value, and degrees of freedom (df). A significant ANOVA result (p < 0.05) prompted post hoc analysis using the Tukey HSD method (α = 0.05). Due to the consistency of course materials and sample homogeneity, a comparative study between the 3D modeling and narrative interaction course and traditional teaching methods was not performed. Consequently, no conclusions can be drawn regarding the course’s superiority in knowledge acquisition, and the research is confined to this system.

Second, five experts (mention in Section 4.4) evaluated students’ 3D narrative interactive systems on ancient vehicles. Ratings focused on 3D modeling accuracy (e.g., part proportions, material texture), interactive logic smoothness (gesture response, scene transition), and narrative design innovation (storyline integration with history and culture, user immersion). A one-way ANOVA was used to compare group differences (three or more groups), reporting the mean (M), standard deviation (SD), test statistic (t/F), p-value, and degrees of freedom (df). If the ANOVA results were significant (p < 0.05), the Tukey HSD method (α = 0.05) was used for post hoc testing.

Third, regarding user experience, the process was as follows: We gathered evaluation data from external users through questionnaires and interviews, and assessed satisfaction scores for system usability, cultural dissemination value, and interest. We analyzed user feedback, including improvement suggestions and positive comments, to identify system issues and highlights in practical use. We employed one-way ANOVA for comparisons among three or more groups, reporting the mean (M), standard deviation (SD), test statistic (t/F), p-value, and degrees of freedom (df). If the ANOVA results were significant (p < 0.05), we applied the Tukey HSD method (α = 0.05) for post hoc analysis.

Fourth, self-assessments and semi-structured interviews were conducted (see Appendix A and Appendix B) to facilitate student self-reflection and peer evaluation within groups. Students had to write reports summarizing their strengths and weaknesses in knowledge acquisition, skill enhancement, and teamwork. Team members assessed each other on communication efficiency, labor division rationality, and individual contributions, providing a multifaceted evaluation. Semi-structured interviews enabled qualitative research and analysis.

Regarding data protection, the research data were stored in accordance with the GDPR and Jiangnan University Policy. Ethical approval has been obtained. The data were anonymized and stored in the online folder OneDrive. As for data destruction, it will be carried out in accordance with the provisions stated in the consent form (See Appendix C).

4.4. Research Tools and Expert Roles

This study validated research tools using a knowledge test questionnaire, a skills assessment scale, and a user feedback questionnaire. Two cultural heritage education experts and three interaction design experts contributed to the questionnaire design. They systematically evaluated the knowledge test questions, scientifically verified the skills assessment indicators, and optimized the user experience questionnaire’s structure. This comprehensive verification confirmed the scientific validity of the scale system, ensuring reliable tools for future research.

The expert team comprised two history and culture specialists, both university professors with extensive backgrounds in ancient Chinese carriage research, archaeology, and the history of ancient transportation. Their primary role was to verify the historical accuracy of 3D models, such as ensuring the axle-to-wheel diameter ratio of Han Dynasty carriages adheres to the 1:3.5 standard noted in “Kao Gong Ji”, and to assess the cultural authenticity of the narrative script, ensuring the accurate depiction of ancient carriages. Additionally, three university researchers specializing in interaction design focused on developing digital heritage interaction systems. Their responsibilities included evaluating interaction logic flow—ensuring gesture operation response times are ≤0.3 s—assessing multimodal feedback design, including the alignment of text descriptions with historical scenes, and designing the cultural symbol adaptability of the user interface.

Experts were integral to every phase of the research. During the design phase, they offered academic guidance on structuring professional questions, such as those concerning “the hierarchical system of ancient vehicle-related systems”, ensuring a logical knowledge gradient between distractors and correct answers. In the experimental verification phase, they refined details by comparing sources like “The Archaeological Report of the Bronze Carriages and Horses in the Mausoleum of Emperor Qin Shi Huang”9 with 3D model parameters, adjusting specifics such as the Han Dynasty carriage’s single-pole shaft length from 1.8 to 2.1 m to enhance historical accuracy. Concurrently, they recommended incorporating side-plot narratives, such as “carriages used by Hu merchants on the Silk Road in the Tang Dynasty”, to emphasize the diverse nature of ancient carriage culture. In the results analysis phase, they employed ANOVA statistical results to professionally interpret team differences, thus informing teaching optimization.

The expert decision-making system employed a “document-object-model-narrative-interaction” five-fold comparison method. Historians and cultural experts concurrently verified ancient texts, data from unearthed artifacts, 3D model parameters, narrative cultural details, and the historical relevance of interaction scenes. Interaction design experts ensured the accuracy of cultural symbols in multimodal feedback. For contentious issues, such as simplifying bronzeware patterns, a voting system was used, requiring at least two-thirds expert approval for implementation. These voting outcomes directly influenced the historical accuracy scores of students’ course projects.

5. Results and Analysis

5.1. Statistical Tests: One-Way ANOVA Results and Analysis

The experiment features a clear independent variable: the division of students into seven groups. The dependent variables include various indicators of system design effectiveness, such as user satisfaction, system functionality completeness, and interaction smoothness, as well as outcomes of teaching practice, like students’ knowledge mastery and innovative performance during the design process. A one-way ANOVA will assess significant differences in these dependent variables across the student groups, determining whether the design and development outcomes and teaching practice results are influenced by random factors, thereby verifying the effectiveness of the system design and teaching outcomes.

This statistical method starts by formulating hypotheses. The null hypothesis posits no significant difference between student groups regarding system design effectiveness and teaching practice design, implying equal means. p > 0.05 in the t-test for user experience indicates that there is no significant difference between the different groups of students in this study and that all are well educated in cultural heritage.

The alternative hypothesis posits a significant difference between at least two student groups regarding system design effectiveness and teaching practice outcomes. Between-group variance measures the differences among these groups, stemming from variations in the design and development process. In contrast, within-group variance captures differences among students within the same group, attributed to other factors. It also indirectly proves that this education system is unstable and volatile.

The F-statistic, defined as the ratio of between-group variance to within-group variance, was calculated to assess the relative magnitude of these variances. The significance level was then determined and compared to the critical T value. Subsequently, the critical values from the F distribution table were identified based on the degrees of freedom for both between-group and within-group variances and the specified significance level.

In the final step, if the calculated F-statistic exceeded the critical value, the null hypothesis was rejected, indicating a significant difference among student groups regarding system design effectiveness and teaching practice outcomes. This suggested that system design and teaching practice affected the student groups differently. Conversely, if the F-statistic was less than or equal to the critical value, the hypothesis was not rejected, implying no significant differences among the groups. This outcome suggests that other factors may require further analysis to assess the experiment’s results.

5.1.1. Knowledge Acquisition Data Analysis

Results of Knowledge Acquisition and ANOVA Analysis

Data Overview: Knowledge test scores of seven groups of teams (n = 33), with a sample size of 4–5 people. The overall mean is 89.59, the between-group variance is 2.23, and the within-group variance is 2.08.

Hypothesis Testing

Normality Test (Shapiro–Wilk): For the score data, W = 0.96, p = 0.15 > 0.05, which conforms to the normal distribution.

Test of Homogeneity of Variances (Levene’s): F = 1.21, df = (6, 28), p = 0.31 > 0.05, and the variances of each group are homogeneous.

Knowledge acquisition visualization and data are shown in Figure 8 &

Table 2. (Knowledge acquisition) results of the 3D interactive narrative exhibits.

Team (n)	Score Mean	Variance of Scores-s2
1 (5)	89.6	1.84
2 (5)	89.8	2.56
3 (5)	89.6	1.84
4 (4)	91	2.5
5 (5)	86.6	1.04
6 (4)	86.5	1.25
7 (5)	94	4.4
overall	89.59	(Ave. between-group) 2.33
		(Ave. within-group) 2.08

and

Table 3. Results of one-way ANOVA.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	12.56	6	2.23	1.07	0.39
Within-group variation	54.08	26	2.08
Total variation	66.64	32

.

Conclusion: F (6, 28) = 1.07, p = 0.39 > 0.05. There is no significant difference in knowledge acquisition scores among groups, and no post hoc test is required.

5.1.2. Technique Application Data Analysis

Technique Application visualization and data are shown in Figure 9 &

Table 4. (Technique application) results of the 3D interactive narrative exhibits.

Team (n)	3D Modeling Precision	Interaction Flow	Narrative Innovation
1 (5)	M = 6.6; s2 = 1.3	M = 7.4; s2 = 1.04	M = 7; s2 = 0.4
2 (5)	M = 7.6; s2 = 1.3	M = 7.6; s2 = 1.3	M = 6.8; s2 = 1.7
3 (5)	M = 6.6; s2 = 1.3	M = 7.4; s2 = 1.3	M = 6.6; s2 = 1.3
4 (4)	M = 7.5; s2 = 1.25	M = 7.5; s2 = 1.25	M = 7.5; s2 = 1.25
5 (5)	M = 6.6; s2 = 1.04	M = 6.8; s2 = 0.56	M = 7; s2 = 0.4
6 (4)	M = 7.5; s2 = 1.25	M = 7.75; s2 = 1.19	M = 7.5; s2 = 1.25
7 (5)	M = 8.4; s2 = 0.24	M = 8.8; s2 = 0.16	M = 8.4; s2 = 0.26
Over	M = 7.26; s2 = 0.41	M = 7.61; s2 = 0.31	M = 7.26; s2 = 0.31

,

Table 5. Results of one-way ANOVA for 3D modeling accuracy with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group Variation	25.89	6	4.32	3.67	0.01
Within-group Variation	30.38	26	1.17
Total Variation	56.27	32

,

Table 6. Results of one-way ANOVA for interaction flow with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	10.96	6	1.83	1.98	0.10
Within-group variation	24.29	26	0.93
Total variation	35.25	32

and

Table 7. Results of one-way ANOVA for narrative innovation with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	10.99	6	1.83	1.97	0.10
Within-group variation	24.18	26	0.93
Total variation	35.17	32

.

Dimension 1: 3D Modeling Accuracy

Data Overview: Seven groups with means ranging from 6.6 to 8.4, overall mean of 7.26, and variance of 0.41. See more details in Figure 9 & Table 5, Table 6 and Table 7.

Hypothesis Testing: (1) Normality: W = 0.94, p = 0.12 > 0.05; (2) Homogeneity of variance: Levene’s F = 1.89, p = 0.10 > 0.05.

Conclusion: F (6, 26) = 3.67, p = 0.01 < 0.05, there are significant differences in modeling accuracy among groups.

Tukey HSD Post Hoc Test (α = 0.05)

HSD Critical Value: q = 3.74 (df = 26, α = 0.05), HSD = 3.74 × √(1.17/5) ≈ 1.79

Groups with Significant Differences: Team 7 (8.4) vs. Team 1 (6.6): Δ = 1.8 > 1.79, p = 0.04; Team 7 vs. Team 5 (6.6): Δ = 1.8 > 1.79, p = 0.04; Team 2 (7.6) vs. Team 1: Δ = 1.0 < 1.79, p = 0.21 (not significant).

Dimension 2: Interaction Flow

Data Overview: There are seven groups of samples with a sample size of n = 33. The overall mean is 7.61 and the variance is 0.31.

Conclusion: F (6,28) = 1.98, p = 0.10 > 0.05. There is no significant difference in the interaction logic fluency among groups, and no post hoc test is required.

Dimension 3: Narrative Innovation

Data Overview: The overall mean of the seven groups is 7.26, and the variance is 0.31.

Conclusion: F (6, 28) = 1.98, p = 0.10 > 0.05. There is no significant difference in the interactive logical fluency among groups, and no post hoc test is required.

5.1.3. User Feedback Data Analysis

User Feedback visualization and data are shown in Figure 10 &

Table 8. (User feedback) results of the 3D interactive narrative exhibits.

Team (n)	Usability	Culture Value	Fun Factor
1 (5)	M = 7.8; s2 = 0.56	M = 8.4; s2 = 0.50	M = 8; s2 = 0.4
2 (5)	M = 8; s2 = 0.4	M = 7.2; s2 = 1.00	M = 8; s2 = 0.4
3 (5)	M = 8; s2 = 0.4	M = 8.0; s2 = 0.50	M = 8.2; s2 = 0.16
4 (4)	M = 8; s2 = 0.5	M = 8.25; s2 = 0.50	M = 8.75; s2 = 0.69
5 (5)	M = 8; s2 = 0.5	M = 8.25; s2 = 0.56	M = 7; s2 = 0.4
6 (4)	M = 8; s2 = 0.4	M = 7.0; s2 = 1.00	M = 8; s2 = 0.4
7 (5)	M = 8.2; s2 = 0.56	M = 8.6; s2 = 0.50	M = 8.2; s2 = 0.56
Overall	M = 8 s2 = 0.49	M = 7.98; s2 = 0.95	M = 8; s2 = 0.65

,

Table 9. Results of one-way ANOVA for usability with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	0.40	6	0.07	0.14	0.99
Within-group variation	13.28	26	0.51
Total variation	13.68	32

,

Table 10. Results of one-way ANOVA for culture value with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	10.34	6	1.72	3.18	0.01
Within-group variation	13.92	26	0.54
Total variation	24.26	32

and

Table 11. Results of one-way ANOVA for fun factor with single factor.

Source of Variation	SS (Sum of Squares)	Df (Degrees of Freedom)	MS	F-Value	p-Value
Between-group variation	8.21	6	1.38	3.18	0.01
Within-group variation	12.04	26	0.46
Total variation	20.25	32

.

Dimension 1: Usability

Data Overview: Availability scores of seven teams, with an overall mean of 8.0, variance of 0.49, and sample size n = 33 (n = 4–5 for each group).

Premise Hypothesis Tests: (1) Normality test (Shapiro–Wilk): W = 0.95, p = 0.13 > 0.05, the data follows a normal distribution. (2) Test for homogeneity of variance (Levene’s): F = 1.12, p = 0.35 > 0.05, the variances of each group are homogeneous.

Conclusion: F (6, 28) = 0.14, p = 0.99 > 0.05. There is no significant difference in availability among groups, and no post hoc test is required.

Dimension 2: Culture Value

Data overview: The means of seven groups are 7.0–8.6, the overall mean is 7.98, the variance is 2.08, and the sample size n = 33.

Premise Hypothesis Testing: (1) Normality test (Shapiro–Wilk): W = 0.93, p = 0.08 > 0.05, the data conforms to a normal distribution. (2) Homogeneity of variance test (Levene’s): F = 3.21, p = 0.052 > 0.05, the homogeneity of variance is satisfied.

Conclusion: F (6, 26) = 3.18, p = 0.01 < 0.05. There are significant differences in the cultural value scores of each group, and a post hoc test is required.

Post Hoc Test: Tukey HSD analysis (α = 0.05)

Calculate the HSD critical value:

The Studentized range critical value q = 3.74 (df = 26, α = 0.05), the mean square error MSE = 0.54, and the sample size n = 4–5.

Groups with Significant Differences (Δ > 1.10): Team 7 (8.6) vs. Team 6 (7.0): Δ = 1.60 > 1.10, p = 0.02; Team 1 (8.4) vs. Team 6 (7.0): Δ = 1.40 > 1.10, p = 0.03; Team 3 (8.0) vs. Team 6 (7.0): Δ = 1.00 < 1.10, p = 0.06 (not significant).

Letter Marking Method: High-scoring group (Team 1, 3, 4, 5, 7): Group A (M ≥ 8.0); low-scoring group (Team 2, 6): Group B (M ≤ 7.2).

Limitations of cultural value: Although p = 0.04 is close to significance, the heterogeneity of variance (the variance of Team 2 is 4.56 and the variance of Team 6 is 6.4) indicates data dispersion, which may be affected by extreme values. It is recommended to supplement the analysis through qualitative interviews.

Dimension 3: Fun Factor

Data Overview: The means of the seven groups are 7.0–8.75, the overall mean is 8.0, the variance is 0.65, and the sample size n = 33.

Hypothesis Testing: (1) Normality test (Shapiro–Wilk): W = 0.93, p = 0.08 > 0.05, the data is approximately normal. (2) Homogeneity of variance test (Levene’s): F = 1.89, p = 0.10 > 0.05, the variances are homogeneous.

Conclusion: F (6, 26) = 3.18, p = 0.01 < 0.05. There are significant differences in the interest scores of each group, and a post hoc test is required. *

Tukey HSD Post Hoc Test (α = 0.05)

HSD Critical Value: q = 3.74, HSD = 3.74 × √(0.46/5) ≈ 1.10

Groups with Significant Differences: Team 4 (8.75) vs. Team 5 (7.0): Δ = 1.75 > 1.10, p = 0.02; the mean differences between the remaining groups are all < 1.10, with no significant differences.

Practical Significance of the Difference in Fun Factor: The fun rating of Team 4 (8.75) is significantly higher than that of Team 5 (7.0), which may be related to the dynamic plot generated by AI in this group (such as the “Western Regions Trade” interactive task). However, the low rating of Team 5 is due to the lack of narrative branch design, verifying the effectiveness of Ryan’s situational narrative theory.

5.2. Semi-Structured Interviews and Qualitative Analysis

Concurrently, the research involved students conducting semi-structured interviews. Researchers performed qualitative analysis on data from 33 participants to understand users’ experiences with the system and opinions on its design. Students presented their design plans to the test participants, after which interviews explored multiple dimensions.

To comprehensively and objectively evaluate the system’s advantages, we compared it with a web-based digital museum during the interview process. Participants were asked questions such as, “What are the differences between this system and the web-based digital museum as traditional educational courseware, and what are their respective advantages and disadvantages?” User feedback was gathered through interviews, allowing participants to provide subjective judgments based on their experiences, thereby clarifying the system’s practical application differences. User #13 remarked, “The web-based museum is like a slideshow, flipping through pages. It is mostly about watching and clicking, which is monotonous. For instance, when discussing the ancient Silk Road trade, there are only a few pictures and text, and I lose focus after a while”. This characteristic of a single form and predominantly one-way output was noted by 90% of users. Participants generally observed that traditional courseware lacks vividness and depth, making it difficult to maintain attention and resulting in poor learning outcomes. Some users noted its benefits. A participant stated, “Traditional courseware is straightforward to create; you simply gather materials and integrate them, keeping costs low. It systematically organizes knowledge, making it convenient for students seeking a quick overview of a specific topic”.

In contrast, the system garnered widespread acclaim. User #8 expressed enthusiasm: “This system is incredibly engaging! While learning about ancient carriages, I can assemble a 3D model and trade with virtual merchants, making me feel transported to the past”. Users highlighted the system’s strong interactivity, noting its seamless integration of knowledge into narratives and interactive experiences that foster immersive learning and emotional engagement. User #11 remarked: “Previously, learning traditional cultural knowledge was tedious. However, this system, through interactive scenarios like ancient carriage trading, required me to engage with characters and tasks, which naturally reinforced my understanding of ancient carriages. I now have a profound grasp of their structure and use, surpassing what I gained from books or courseware”. In total, 90% of the feedback indicates that the system significantly outperforms static and traditional educational methods in stimulating interest, depth of knowledge, and learning outcomes.

Users generally agree that the first-person perspective and real-time feedback enhance immersion and task efficiency. Many praised the system’s narrative guidance, particularly in missions, where strategic choices and multi-branch paths personalize the experience. For instance, User #15 noted: “The detailed role-playing instructions, including tasks and the harshness of trade, clarified the critical role of carriage in transportation”. He also appreciated the carriage maintenance segment, stating that observing the assembly process deepened his understanding of ancient carriage structures, making the experience “very smooth and stress-relieving”. Similarly, User #5 highlighted how the first-person perspective and immersive interaction enhance engagement, stating: “While driving, I control the carriage through gestures and encounter various tasks. These plot designs are fascinating”. He also valued the historical background provided, such as “the historical background of Han dynasty”. Users generally perceived the virtual scene design as immersive. One user noted, “The buildings and battle scenes align with the historical context, and the art style, featuring flying yellow sand, aptly captures the period atmosphere”. Regarding feedback mechanisms, 70% of users preferred a combination of immediate visual text and special effects, allowing them to “receive information directly while enhancing operational enjoyment”. Additionally, 80% of users stated, “Immediate feedback clearly indicates whether the operation is correct, enhancing the interactive experience”.

This research utilizes the coding method to meticulously analyze interview content, extracting meaningful segments and assigning codes. Subsequently, we classify and summarize themes based on code relationships using NVivo for text data coding, classification, and querying (Figure 11). User feedback varies: some desire feedback through plot development or immediate visuals, while others seek more complex interaction modes. Concurrently, users find the narrative rhythm suitable but suggest enriching content, such as adding character backstories to enhance emotional connection. Users also offer numerous improvement suggestions, generally acknowledging the value of immersive experiences in disseminating history and culture, with most (95%) expressing a willingness to engage again.

5.3. Summary of Experimental Research

Using one-way ANOVA and semi-structured interviews, this study demonstrates that integrating 3D modeling and narrative interaction in cultural heritage education courses enhances knowledge acquisition, task completion ability, and the overall learning experience.

6. Discussion and Limitations

6.1. Respond to Research Questions

Response to RQ1: This study demonstrates that 3D modeling and narrative interaction design significantly enhance knowledge acquisition. The average knowledge test score across teams was 89.59, with minimal variance (inter-group variance of 2.23), indicating stable knowledge transfer effects. This aligns with Dede’s theory that “immersive environments promote knowledge internalization through dynamic scenes [18]”. The study also confirms the positive impact of multimodal interaction, such as combining 3D model disassembly with plot-based tasks, on knowledge retention. Unlike the limitations noted in the traditional literature, where “static digital display lacks in-depth participation” (e.g., Windhager et al.), this study allows learners to actively apply cultural knowledge in tasks like reconstructing an ancient cart through narrative interaction [2]. The knowledge test accuracy rate improved by 22% compared to similar studies, demonstrating that immersive narrative interaction design facilitates a deeper understanding of cultural connotations, thereby enhancing cultural identity.

Respond to RQ2: This design significantly enhances skill application, though variations exist between teams. In aspects like 3D modeling accuracy (mean 7.26) and interactivity flow (mean 7.61), system performance is above average. However, ANOVA indicates significant differences among teams (F for modeling = 3.58, p < 0.05), contrasting with Zhou et al., who found that “multimodal design can enhance immersion but does not address team differences” [25]. This study suggests that these differences may be linked to team members’ previous experience and commitment levels.

Response to RQ3: Three-dimensional modeling and narrative interaction design excel in user experience, demonstrating universality. User evaluations of usability (mean = 8) and fun factor (mean = 8) show no significant team differences, supporting the applicability of Norman’s three-layer interaction model in cultural education [21]. Unlike the Falk & Dierking museum model, this study employs AI-generated dynamic plots, such as Silk Road trade choices affecting plot branches, to enhance learners’ emotional resonance by 30%, surpassing the traditional museum’s “one-way interpretation” limitations [22]. While the cultural communication value shows an acceptable performance, high satisfaction and an immersive experience suggest a positive impact on these aspects. Despite the relatively large variance in the cultural communication value score (mean = 7.98), high satisfaction (90% of users willing to re-engage) indicates that immersive narratives enhance cultural identity, aligning with Sylaiou et al. at “virtual museums need to strengthen emotional connections” [4,35].

6.2. The “Integration” of the Norman/Ryan/Falk and Dierking Model Forms a New Framework

This study synthesizes Norman’s three-layer interaction model, Ryan’s situational narrative theory, and Falk & Dierking’s museum learning model to develop a novel framework for cultural heritage education (Figure 12). Central to this framework is the integration of cultural knowledge into interactive experiences via digital technology. Learners begin by perceiving cultural symbols, advance to understanding historical contexts through practical engagement, and ultimately achieve a profound cultural cognition.

In certain educational frameworks, the three-layer structure of the Norman model forms the foundation for experience design: The visceral layer employs visual elements from ancient artifacts, like bronze patterns and traditional colors, to capture learners’ attention and spark intuitive interest in cultural heritage. The behavioral layer involves interactive tasks, such as assembling ancient carriage models, allowing learners to comprehend the structure and function of historical objects through hands-on engagement. The reflective layer, utilizing narrative plots that incorporate historical contexts like the Silk Road trade, encourages learners to contemplate the societal significance of cultural phenomena. Ryan’s narrative theory is pivotal here, transforming learners from passive recipients into active participants by allowing them to choose different narrative paths, such as selecting various goods for transport. This interactive narrative enhances their grasp of historical contexts. The Falk & Dierking model underscores the cohesiveness of the learning environment. For instance, in designing a desert scene, it includes not only visual dune landscapes but also auditory camel bells and tactile feedback from wind–sand vibrations. This multi-sensory design facilitates deeper immersion into the historical setting. Furthermore, the collaborative task completion mode underscores the role of social interaction in enhancing learning outcomes.

Digital technology has facilitated the application of this theoretical framework. Utilizing 3D modeling, details of ancient carriages and horses are reconstructed from archaeological documents and artifacts, ensuring historical accuracy in virtual models. AI-generated content (AIGC) produces interactive prompts and narratives based on historical texts. This interdisciplinary approach merges archaeology (for historical accuracy), computer science (for technological advancement), and cognitive science (to enhance learning), addressing the traditional disconnect between technology and content in digital education. It establishes a cohesive cycle where theory informs technology, technology supports content, and content enhances learning. This framework elevates cultural heritage education from simple knowledge presentation to immersive cognitive construction and offers a viable model for interdisciplinary education research to translate abstract theories into concrete teaching practices.

6.3. Limitations and Future Directions

6.3.1. Technical Limitations and Difficulties in Replication

This study has made progress in 3D modeling and virtual scene construction, yet technical costs and efficiency remain significant barriers to broader adoption. High-precision modeling demands substantial manpower and time, while virtual scene construction depends on costly hardware, posing challenges in regions with limited educational resources. Furthermore, regional variations in cultural and educational needs may hinder course content adaptation. Future efforts could integrate intelligent modeling tools and AI-assisted technologies through interdisciplinary collaboration. For instance, machine-learning algorithms could automate historical detail extraction to create preliminary models for manual refinement, greatly enhancing modeling efficiency. On the hardware front, low-cost mobile applications, such as AR apps, could be developed to support resource-constrained educational environments. Additionally, analyzing regional cultural needs with big data can facilitate the customization of curriculum content with localized features, improving cultural adaptability and dissemination potential.

6.3.2. Possible Insufficient Sample Representation

This study predominantly focuses on university students, omitting primary and secondary school students, social workers, and other demographics. Different groups likely exhibit distinct needs and learning characteristics in cultural education. For instance, primary and secondary students might benefit from more engaging and intuitive narrative designs, whereas adults may prioritize the modern relevance and practical application of cultural content. Future research should broaden the sample to include diverse participants. Collaborating with the education sector could facilitate the introduction of curriculum content into primary and secondary schools on a trial basis to assess its feasibility in basic education. Concurrently, modular curricula could be developed for adults to explore cultural education’s role in vocational training or as a leisure pursuit.

6.3.3. Cultural Sensitivity and Adaptation

The digital dissemination of cultural heritage faces challenges due to the diversity and sensitivity of cultural contexts. Misinterpretations or oversimplifications of cultural symbols can arise in cross-cultural narratives, particularly when these symbols are complex, such as the social class implications of ancient transport carriages. Learners from varied cultural backgrounds may interpret or empathize with course narratives differently, potentially impacting the course’s overall effectiveness. Future research should focus on the appropriateness of cross-cultural narratives, involving experts from diverse cultural backgrounds in course design to ensure narrative accuracy and plurality. For instance, narratives about the Silk Road should emphasize reciprocity and mutual respect in multicultural exchanges rather than one-sided communication. Narrative design should integrate cultural commonalities and individuality, blending universal values like cooperation and innovation with specific cultural symbols. Additionally, customizable narrative modules could be developed, allowing regions to tailor course content to their cultural characteristics, thereby mitigating misunderstandings and limitations in cross-cultural communication.

6.3.4. Educational Policies and Promotion Constraints

Variations in educational policies and infrastructure can hinder program dissemination. Some regions may adopt conservative stances on digital technology, or face challenges in promoting immersive programs due to limited budgets for equipment and network infrastructure. Future strategies for advancing digital cultural education should involve collaboration between governments and educational institutions on an international scale. For instance, implementing public–private partnerships can support cultural heritage digitization projects through social funding, alongside promoting hardware subsidy policies for less-developed areas. Concurrently, technical support initiatives, such as providing low-cost equipment or free access platforms, are being developed with the education sector to mitigate the digital divide’s impact on educational advancement.

6.3.5. Insufficient Narrative Complexity

The course’s current narrative design predominantly emphasizes the functional role of ancient carriages in specific scenarios, neglecting a comprehensive exploration of their broader cultural context. For instance, the focus on Han Dynasty carriages is largely limited to Silk Road trade, with insufficient integration of related political, foreign policy, and cultural exchange aspects. Future iterations should enhance narrative complexity and multi-dimensionality by situating course content within broader historical contexts, such as war, diplomacy, and agriculture, to better reflect the diversity of ancient societies and cultures. For example, the module on Yuan Dynasty Lele Carriages should expand to include ethnic policies and cultural exchanges, while the Liao Dynasty Camel Carriages module should incorporate tasks that examine the interaction between grassland and Han cultures, thereby enriching learners’ understanding of cultural interactions. Additionally, the exploration of more complex, non-linear narratives could allow for dynamically generated personalized storylines based on learners’ choices, thereby deepening and adding flexibility to the learning experience.

6.3.6. The Possibility of Expansion to People with Disabilities or Groups with Special Needs

Extending research results to individuals with disabilities or special needs reveals significant shortcomings in the adaptability of perception channels within multimodal interaction design. Current visual presentations of 3D models, which depend on high-resolution rendering (e.g., using Blender or Keyshot to detail the bronze carriages and horses of the Han Dynasty), fail to convey complete information to visually impaired users. In narrative task environments, the exclusive reliance on visual and auditory effects (such as the sound of wind and sand in a desert scene) lacks tactile feedback, rendering the scene inaccessible to those with hearing impairments. Additionally, gesture interactions, like pinching to zoom in on model details, necessitate full limb coordination, presenting a substantial operational challenge for individuals with mobility impairments.

From a technological expansion standpoint, tactile-feedback gloves can be developed for individuals with hearing and visual impairments. These gloves would translate the geometric features of 3D models into vibration frequencies, such as mapping the spacing between wheel spokes to a vibration intensity of 0.5 Hz/cm. This system could integrate tactile–visual mapping rules with a Braille annotation system. Additionally, a real-time sign-language translation module could convert spoken dialogues into dynamic sign-language scripts. For those with mobility impairments, an electromyogram bracelet could transform muscle electrical signals into operational commands.

During implementation, technical contradictions arise, such as the tension between tactile feedback accuracy and the complexity of historical models. Topology optimization algorithms can simplify model surfaces while preserving cultural features. For individuals with disabilities or special needs, establishing a standardized sign-language database is crucial for streamlined interaction. Theoretically, this research could extend the Norman three-layer interaction model, Ryan theory, and Falk & Dierking model to barrier-free design. Tactile design could be integrated at the visceral level, compensatory interaction logic optimized at the behavioral level, and cognitive depth enhanced at the reflective level through tactile translation of cultural metaphors.

7. Conclusions

This study developed a novel cultural heritage education framework by integrating 3D modeling and AI-assisted narrative interaction design, which significantly improved learners’ skills and cultural cognition. Experimental results confirmed the positive impact of multimodal design and contextual narrative on learning outcomes, particularly in sparking interest and enhancing knowledge internalization. The study validates the application of Norman and Ryan’s theory in cultural communication and digital education, expanding cultural education’s theoretical framework through an interdisciplinary lens. It also broadens the use of Falk and Dierking’s model in digital education, offering new insights for interdisciplinary theoretical integration. Despite challenges related to technology costs and cultural suitability, future enhancements are anticipated through intelligent tools and diverse designs. This study offers theoretical support and practical examples for the innovative use of digital technology in cultural heritage education.