Exploring the Key Factors Influencing the Plays’ Continuous Intention of Ancient Architectural Cultural Heritage Serious Games: An SEM–ANN–NCA Approach

Abstract

Serious games (SGs) have been widely employed in the digital preservation and transmission of architectural heritage. However, the key determinants and underlying mechanisms driving users’ continuance intentions toward ancient-architecture cultural heritage serious games (CH-SGs) have not been thoroughly investigated. Accordingly, a conceptual model grounded in the stimulus–organism–response (S–O–R) framework was developed to elucidate the affective and behavioral effects experienced by CH-SG users. Partial least squares structural equation modeling (PLS-SEM) and artificial neural networks (ANNs) were employed to capture both the linear and nonlinear relationships among model constructs. By integrating sufficiency logic (PLS-SEM) and necessity logic (necessary condition analysis, NCA), “must-have” and “should-have” factors were identified. Empirical results indicate that cultural authenticity, knowledge acquisition, perceived enjoyment, and design aesthetics each exert a positive influence—of varying magnitude—on perceived value, cultural identification, and perceived pleasure, thereby shaping users’ continuance intentions. Moreover, cultural authenticity and perceived enjoyment were found to be necessary and sufficient conditions, respectively, for enhancing perceived pleasure and perceived value, which in turn indirectly bolster CH-SG users’ sustained use intentions. By creating an immersive, narratively rich, and engaging cognitive experience, CH-SGs set against ancient architectural backdrops not only stimulate users’ willingness to visit and protect heritage sites but also provide designers and developers with critical insights for optimizing future CH-SG design, development, and dissemination.

1. Introduction

Cultural heritage is broadly defined as the tangible and intangible assets that have been transmitted through history [1]. In addition to tangible and intangible cultural heritage [2], digital heritage has been formally recognized by UNESCO as a third category, and natural heritage has been acknowledged as an important complement to both forms [3]. Architectural heritage—including monuments, vernacular buildings, and archaeological landscapes—has been regarded as an irreplaceable repository of cultural memory, identity, and aesthetic achievement [4]. As precious legacies passed down through generations, the degradation or loss of these built heritage assets would result in the irreversible depletion of global cultural resources. However, such sites are increasingly threatened by climate change, overtourism, armed conflict, and rapid urban redevelopment [5]. Consequently, serious games that integrate playfulness, interactivity, immersion, and user-friendliness have emerged as a novel solution for the digital intelligent dissemination of architectural heritage [6].

Serious games for cultural heritage (CH-SGs) have been recognized as both a practical necessity and a source of considerable applied value [7]. CH-SGs have been shown to effectively facilitate the preservation and dissemination of ancient architectural heritage [8]. By focusing on particular historic structures and employing multimodal interaction technologies to simulate heritage exploration experiences, users’ cultural identity can be strengthened [9]. In comparison to traditional educational methods, the informal learning mechanisms embedded within CH-SGs have been demonstrated to produce stronger communicative outcomes and to more effectively elevate public awareness of heritage conservation [10]. Through gamification elements—such as virtual reconstructions, skill-based challenges, and narrative-driven exploration—serious games create immersive learning environments that establish an organic connection between architectural heritage and contemporary values.

Cultural heritage serious games (CH-SGs), such as Forbidden City Panorama (https://pano.dpm.org.cn/, accessed on 27 March 2025) and Digital Library Cave (https://dlc.e-dunhuang.com/, accessed on 29 March 2025), leverage advanced digital technologies—including high-precision digital twins, artificial intelligence, and immersive game mechanics—to create highly authentic virtual reconstructions of historic architecture. These CH-SGs effectively deliver immersive learning experiences, enhancing users’ understanding and fostering commitment toward heritage preservation [7,11]. By combining entertaining content with rigorously researched historical narratives, it deepens users’ understanding of and commitment to preserving the Dunhuang architectural heritage [7]. Moreover, Black Myth: Wukong merges contemporary culture with ancient mythological themes to reignite interest in Shanxi’s architectural heritage [12]. Cross-media narrative research has shown that when heritage elements are foregrounded in game tools, they can stimulate curiosity and strengthen players’ intentions to continue engaging [13]. From this perspective, Black Myth: Wukong can serve as a powerful stimulus to draw users into the gaming experience—especially when Shanxi’s architectural heritage is complemented by related digital content or thematic events. By linking serious games with physical heritage sites, audiences are enabled to gain a deeper appreciation of the historical and cultural contexts that inspire their narratives [14]. This synergy illustrates how ancient architectural heritage can be harnessed through serious games to enhance the educational value of heritage destinations and to establish sustainable dissemination mechanisms.

However, despite the demonstrated potential of serious games in the field of ancient architectural heritage, significant limitations persist and empirical evidence remains fragmented [15]. To date, only around twenty related titles have been released in China by leading institutions such as Tencent Games and the Palace Museum [16]. Existing research has unfolded along two primary dimensions. On one hand, scholars have focused on optimizing game design elements: for example, Xu [17] demonstrated—using the Shenyang Palace Museum’s architectural heritage—that both functional and aesthetic design enhancements yield positive effects on user experience. On the other hand, most studies have concentrated on the educational functions of serious games [18,19]. Yet current research does not explain how specific design stimuli are translated into user behaviors and continuance intentions. These gaps have posed serious constraints on the development of CH-SGs. Therefore, the present study will undertake a more systematic and in-depth empirical investigation to address these deficiencies and to enrich both theoretical understanding and practical guidance.

Three potential innovations are proposed in this study. First, under the guidance of the S–O–R framework, novel stimulus and organism constructs are introduced to investigate users’ continuance intentions toward CH-SGs. The focus is thus shifted from merely identifying direct psychological antecedents to exploring how external CH-SG feature factors interact with those antecedents to reshape continuance intention. Second, a multi-method analytical strategy integrating PLS-SEM, ANN, and NCA is adopted to capture both linear and nonlinear relationships, thereby offering a more comprehensive understanding of the key determinants of continuance intention. Third, the core factors driving users’ engagement decisions are examined in order to unlock the latent needs of CH-SG users. Finally, insightful recommendations will be offered to designers and developers to inform strategic planning for CH-SG initiatives.

2. Literature

2.1. Stimuli–Organism–Response (S–O–R) Model

The S–O–R model, rooted in the behaviorist “stimulus–response” paradigm, comprises three interrelated components—stimulus, organism, and response—to explicate the mechanisms underlying individual behavior formation [20]. Within this framework, stimuli (S) denote external drivers of users’ behavioral intentions; the organism (O) encapsulates users’ internal emotional and cognitive states; and the response (R) represents the resultant behavior, typically manifested as approach or avoidance, arising from the interaction between external stimuli and internal states [21].

The S–O–R model has been widely applied across diverse research domains, demonstrating robust explanatory power for individual behavior, particularly within consumer psychology [22]. As the framework has evolved, its use has been extended to gamification contexts—including hybrid approaches that combine gamified elements with traditional surveys [23]. Gamification has been shown to effectively stimulate user engagement motivations, thereby subtly influencing cognition, affect, and decision-making processes, and serving as a key mechanism for driving consumer behavior change [24]. In their study, Gatautis et al. [25] constructed a gamified marketing S–O–R framework to examine how game elements function as stimuli that trigger shifts in consumers’ psychological states, which in turn significantly affect decision outcomes and behaviors. Moreover, prior research has further extended the S–O–R model by incorporating constructs such as perceived value to enhance the prediction of user behavior [26]. Building on these precedents, the present study investigates the applicability of the S–O–R model within the context of serious games for ancient architectural heritage. In this configuration, functional and hedonic features of CH-SGs are conceptualized as external stimuli; perceived value, cultural identity, and perceived enjoyment serve as organismic mediators; and continuance intention is treated as the behavioral response. Thus, a gamification-adapted S–O–R path framework is proposed to provide a theoretical foundation for future research on cultural heritage serious games.

2.1.1. Characteristics of CH Serious Games (Stimulus)

Serious games are essentially regarded as a form of digital games [27]. Digital games have been delineated into four core components—narrative, visual representation, game mechanics, and technical support [28]. In the present study, the stimulus factors of serious games for ancient architectural heritage have been classified into two primary dimensions: functional features and hedonic features. Functional features are defined by the authenticity of cultural elements and opportunities for knowledge acquisition, both of which are posited to enhance users’ perceived learning value and engagement intentions. Conversely, hedonic features are constituted by perceived enjoyment and aesthetic design, which are posited to increase the game’s appeal and visual expressiveness, thereby eliciting psychological resonance and influencing subsequent behavioral responses.

2.1.2. User Experience (Organism)

The S–O–R model posits that environmental stimuli influence behavioral responses indirectly through their effects on user experience [29]. Within this framework, organismic factors related to user experience include perceived value, cultural identity, and perceived enjoyment. Perceived value is widely recognized as a critical predictor of user behavior [30]. For instance, Chiu et al. [31] demonstrated that a thoughtfully crafted user experience generates value, encompassing users’ perceptions of that value. In empirical research, Deng et al. [32] confirmed that both entertainment and educational experiences contribute synergistically to the formation of cultural identity, underscoring their joint role in reinforcing users’ sense of belonging. Moreover, prior studies have identified perceived enjoyment not only as a central concept in media entertainment [33] but also as a vital component of user experience; perceived enjoyment reflects intrinsic motivation and emotional needs, substantially shaping attitudes toward a product or content and continuance intentions [34]. Accordingly, the present study provides a theoretical foundation for understanding players’ cognitive and affective responses in serious games for ancient architectural heritage and offers guidance for designing more engaging tools for cultural heritage dissemination.

2.1.3. Game Continuance Intention (Response)

Functional features in serious games for ancient architectural heritage, such as the authenticity of cultural elements and opportunities for knowledge acquisition, are posited to exert an indirect effect on continuance intention via the mediating role of user experience. However, because actual usage behavior in heritage-focused games is often difficult to observe and quantify directly, behavioral intention has been established as a valid proxy for predicting actual behavior. As Venkatesh and Davis [35] demonstrated, behavioral intention can effectively forecast individuals’ real-world actions. Accordingly, this study employs game continuance intention as the key metric for assessing users’ acceptance of and sustained engagement with serious games for ancient architectural heritage, thereby evaluating the dissemination efficacy and user retention of CH-SGs.

2.2. Hypotheses Development and Conceptual Model Construction

2.2.1. Authenticity of Cultural Elements

Authenticity is defined as users’ perception of a product or service’s uniqueness, originality, and credibility, reflecting the degree of novelty and veracity in its content, and is regarded as a critical factor for enhancing user identification and engagement [36]. In the evaluation of cultural products, authenticity has often been prioritized as a primary criterion [37,38]. Its core definition, “authenticity based on traditions created by local communities” [39], emphasizes originality and trustworthiness. Originally conceptualized within cultural heritage research [40], the notion of authenticity has since been extended to cultural heritage tourism [41] and to cultural product domains [42].

In the context of serious games, indexical cultural elements (e.g., ancient architectural ruins, traditional music) have been found to significantly enhance players’ perceptions of authenticity and cultural identification by directly linking game content to historical realities [43,44,45]. Such elements are used to construct an authentic cultural backdrop, which in turn deepens immersion [46] and heightens users’ perceived cultural value [47]. Accordingly, it is hypothesized that the authenticity of cultural elements in CH-SGs exerts a positive influence on perceived value, cultural identity, and perceived enjoyment, mediated by strengthened trust and cultural resonance, and thereby fosters continuance intention. The specific research hypotheses are thus presented as follows:

H1a: 

Authenticity of cultural elements in cultural heritage serious games positively affects perceived value.

H1b: 

Authenticity of cultural elements in cultural heritage serious games positively affects cultural identity.

H1c: 

Authenticity of cultural elements in cultural heritage serious games positively affects perceived enjoyment.

2.2.2. Knowledge Acquisition

Knowledge acquisition is typically defined as the process by which individuals integrate and expand upon existing information to learn and master new knowledge [48]. Prior research has demonstrated that knowledge acquisition plays a critical role in the adoption and application of technological innovations, effectively enhancing users’ understanding of and adaptability to emerging technologies, and thereby promoting their widespread uptake and sustained use [49,50]. In the context of serious games for cultural heritage, Jiang et al. [43] proposed that knowledge acquisition enhances players’ comprehension of cultural value and strengthens their cultural identity through the gaming experience. Furthermore, studies have shown that when learners acquire more analytical knowledge, their understanding of experiential content deepens, resulting in a significant increase in perceived enjoyment [51]. In serious games for cultural heritage, knowledge acquisition not only deepens users’ understanding of cultural content but also markedly elevates their perception of the game’s conveyed cultural value, thereby reinforcing the overall sense of meaning and identification. Accordingly, the following research hypotheses are proposed:

H2a: 

Knowledge acquisition in cultural heritage serious games positively affects perceived value.

H2b: 

Knowledge acquisition in cultural heritage serious games positively affects cultural identity.

H2c: 

Knowledge acquisition in cultural heritage serious games positively affects perceived enjoyment.

2.2.3. Perceived Playfulness

Interest has been defined as the pleasurable experience derived during an activity and the intrinsic motivational inclination stimulated by interaction with that activity [52]. It can also be conceptualized as a preference for the enjoyment of participation [53], and is regarded as an intrinsic motivator that enables individuals to derive satisfaction from novel experiences, leading them to engage in tasks for the enjoyment of the process rather than solely for outcomes. Consumers who exhibit high levels of interest are typically more creative, open-minded, and willing to embrace new experiences, integrating these experiences into their self-concept [54]. In the context of serious games for cultural heritage, perceived interest has been shown to enhance players’ emotional investment and cultural experience, thereby significantly fostering cultural identity [43]. Moreover, by elevating both enjoyment of cultural content and affective engagement, interest is posited to strengthen cultural identification and to promote the dissemination and preservation of cultural heritage. Accordingly, the following hypotheses are proposed:

H3a: 

Perceived playfulness in cultural heritage serious games positively affects perceived value.

H3b: 

Perceived playfulness in cultural heritage serious games positively affects cultural identity.

H3c: 

Perceived playfulness in cultural heritage serious games positively affects perceived enjoyment.

2.2.4. Design Aesthetics

Design aesthetics are recognized as a salient factor in serious games for ancient architectural heritage, whereby a harmonious integration of visual and affective elements is employed to enhance users’ emotional engagement and to foster cultural identity [43]. Specifically, aesthetic design is manifested through interface graphics, color schemes, and musical accompaniment, all of which serve to elevate visual appeal and emotional connection [55]. Empirical evidence indicates that aesthetic components in game design—particularly visual presentation, sound design, and interactive elements—exert a significant influence on user experience, constituting key determinants of perceived enjoyment [55]. Moreover, design aesthetics have been shown to enhance perceived value by deepening immersion and reinforcing cultural identification within the context of heritage serious games [7]. In sum, the application of high-quality aesthetic design in serious games for cultural heritage not only promotes users’ emotional investment and immersive experience but also intensifies their identification with and enjoyment of the heritage content. Accordingly, the following hypotheses are proposed:

H4a: 

Design aesthetics in cultural heritage serious games positively affects perceived value.

H4b: 

Design aesthetics in cultural heritage serious games positively affects cultural identity.

H4c: 

Design aesthetics in cultural heritage serious games positively affects perceived enjoyment.

2.2.5. Perceived Value

Perceived value is the subjective, overall assessment of the utility derived from a product or service in a given context [56] and exerts a significant influence on user behavior. Specifically, an increase in perceived value has been shown to markedly improve participants’ affective experiences, thereby strengthening their sense of belonging [57]. Liu et al. [58] further confirmed that perceived value—such as identification with ancient architectural heritage and satisfaction with its use—effectively promotes users’ continuance intentions, thereby reducing their acquisition costs. Moreover, Petrick [59] emphasized that perceived value, together with satisfaction and service quality, constitutes a key predictor of behavioral intentions. These findings collectively indicate that the sustainable development of serious games for cultural heritage must be built upon a multidimensional framework of perceived value, providing empirical support for digital cultural heritage preservation. Accordingly, the following hypothesis is proposed:

H5: 

Perceived value has a direct and positive impact on game continuance intention.

2.2.6. Cultural Identity

From a societal perspective, cultural identity is founded on a sense of belonging and is closely tied to traditions, lifestyles, and values, reflecting a strong feeling of affiliation [60]. It differentiates between cultural and value systems by encompassing unique bodies of knowledge [61]. Anderson [62] emphasized that, in cultural-education video games, alignment between game content and players’ cultural backgrounds can narrow cultural cognition gaps and enhance understanding and appreciation of local culture. Swoboda et al. [63] noted that cultural identity plays a pivotal role in digital cultural-heritage products, especially among users with deep local cultural affiliations, who are more likely to form positive emotional connections with culturally congruent offerings. Moreover, research has shown that cultural identity significantly influences continuance intentions in virtual tourism experiences for heritage sites [11]. Stronger cultural identity enhances users’ perceived value and satisfaction, thereby promoting sustained engagement. Accordingly, the following hypothesis is proposed:

H6: 

Cultural identity has a direct and positive impact on game continuance intention.

2.2.7. Perceived Enjoyment

“Enjoyment” refers to the sense of accomplishment, mental stimulation, and pleasure experienced by users during engagement with serious games [64]. In contrast, perceived playfulness denotes the degree to which game mechanics afford spontaneous exploration, interactive engagement, and feelings of autonomy, rather than the emotional satisfaction captured by perceived enjoyment [65]. Enjoyment constitutes a positive psychological state, manifested in the pleasure, fun, and aesthetic appreciation derived from gameplay [66]. It has been identified as a core element of the mobile gaming experience, being critical to a game’s appeal [67]. Bai et al. [64] further pointed out that in cultural heritage serious games, perceived enjoyment significantly affects users’ perception and has a positive effect on their behavioral intentions, especially their game continuation intentions. Moreover, across various domains such as online shopping, digital gaming, and mobile travel applications, perceived enjoyment has been demonstrated to strongly predict continuance intentions [68,69]. Lee & Tsai. [70] proposed an online game user acceptance model based on the theory of planned behavior and found that perceived enjoyment had a positive impact on players’ willingness to play the game. Thus, perceived enjoyment not only occupies a central role in the gaming experience but also plays a crucial part in fostering sustained engagement, especially within cultural heritage serious games, where heightened enjoyment substantially increases users’ continuance intentions. Accordingly, the following hypothesis is proposed:

H7: 

Perceived enjoyment has a direct and positive impact on game continuance intention.

2.2.8. Game Continuance Intention

The successful adoption of an innovation depends not only on users’ initial willingness to adopt it but, more importantly, on their intention to continue using it in the future [71]. Within the context of cultural-heritage serious games (CH-SGs), however, continuance intention signifies more than a simple retention metric [72]. Sustained digital engagement is understood to deepen players’ cognitive and affective bonds with heritage, facilitate ongoing knowledge sharing, stimulate virtual and on-site revisit intentions, and ultimately cultivate preservation-oriented behaviors [73]. Jiang et al. [43] further demonstrated that, in intangible cultural heritage serious games (ICH-SGs), continuance intention shapes user behavior and consequently shifts attitudes toward heritage conservation. Accordingly, in this study, continuance intention is conceptualized as a culturally embedded behavioral response. We therefore posit that perceived value, cultural identity, and perceived enjoyment each exert a positive influence on users’ continuance intention toward CH-SGs.

2.2.9. Conceptual Model

Based on the foregoing discussion, the research model depicted in Figure 1 was constructed. It is hypothesized that the four core features of serious games for ancient architectural heritage—authenticity of cultural elements, knowledge acquisition, perceived interest, and design aesthetics (H1–H4)—will exert significant effects on users’ perceived value, cultural identity, and perceived enjoyment. Furthermore, as proposed in hypotheses H5–H7, these perceptions and experiences are expected to influence users’ psychological responses, specifically manifested in continuance intention toward the game.

3. Research Methodology

3.1. Research Process

The study was conducted in five sequential phases (see Figure 2). In Phase 1, a literature review was performed to identify the key determinants of user engagement with serious games for ancient architectural heritage. A conceptual model was then developed, a survey instrument was designed, and data were collected; the valid responses were subjected to descriptive analysis. In Phase 2, partial least squares structural equation modeling (PLS-SEM) was employed to test the linear relationships posited in the conceptual model. Additionally, PLS-SEM’s sufficiency logic was used to identify “sufficient” conditions—i.e., factors that, when present, guarantee the outcome. In Phase 3, the latent variable scores generated by PLS-SEM were used as inputs to an artificial neural network (ANN), which tested for nonlinear relationships among factors, ranked their relative importance, and enabled a comparison between PLS-SEM and ANN findings. In Phase 4, necessary condition analysis (NCA) was applied—based on necessity logic—to identify “necessary” factors without which the outcome cannot occur. By integrating the results of PLS-SEM and NCA, the conceptual model was refined to distinguish between “necessary” and “sufficient” determinants. Finally, in Phase 5, theoretical and practical recommendations were formulated, drawing on the combined insights from SEM, ANN, and NCA, to guide the optimization of continuance intention in serious games for ancient architectural heritage.

3.2. Data Collection

3.2.1. Pre-Research Analysis

A quantitative survey methodology was adopted to empirically evaluate the proposed research model. Several clear advantages are afforded by the survey approach. First, because the required data primarily reflect respondents’ subjective perceptions—specifically their perceived value, cultural identity, and perceived enjoyment derived from engagement with CH-SGs—questionnaire administration was deemed highly appropriate [74]. Second, the use of standardized surveys enhances the clarity of interpretation and the statistical robustness of the findings [6]. Given the unique nature of the data—continuance intention driven by user experiences in serious games is not available in public repositories—it was necessary to collect primary data through original fieldwork.

In March 2025, a pilot survey was administered to a randomly selected cohort of 50 enthusiasts of games related to Chinese ancient architectural heritage to assess the questionnaire’s validity. The pretest results indicated that both the overall reliability and Cronbach’s α coefficients exceeded 0.80, confirming strong internal consistency and reliability. The questionnaire comprised three sections. The first section presented an informed-consent procedure and screened respondents for eligibility based on prior experience with serious games set in ancient architectural contexts; those who declined consent or did not meet the experience criterion were excluded from further participation. The second section gathered demographic data, including gender, age, and educational level. The third section contained measurement items for the eight constructs specified in the research model. The constructs’ conceptual definitions, number of items, specific measurement statements, and source references are presented in

Table 1. Construct conceptualization and items.

Variable	Option	Items	Key References
Authenticity of Cultural Elements (ACE)	ACE1	The digital representations of ancient architecture presented in CH-SGs were perceived as highly realistic.	[75]
	ACE2	The visual scenes in CH-SGs were characterized by a diverse array of traditional cultural elements, including architectural ruins, figures, color schemes, and motifs.
	ACE3	The historical context and traditional aesthetics of the ancient architectural sites were found to align with my cultural understanding.
Knowledge Acquisition (KA)	KA1	New knowledge pertaining to architectural cultural heritage was acquired through CH-SGs.	[43,76]
	KA2	Specific cultural symbols and markers of ancient architectural sites were identified within the game.
	KA3	The level design of the serious game was optimized to facilitate the learning process of heritage knowledge.
	KA4	CH-SGs provide multidimensional pathways for the acquisition of cultural knowledge.
Perceived Playfulness (PP)	PP1	Engagement interest was stimulated by the cultural content presented in CH-SGs.	[43,53]
	PP2	The game experience enhanced curiosity regarding the aesthetics of architectural heritage sites.
	PP3	Participation in CH-SGs elicited a sense of enjoyment.
	PP4	Completion of game tasks in CH-SGs was perceived as genuinely enjoyable.
Design Aesthetics (DA)	DA1	Profound significance was conveyed by the architectural cultural symbols presented in the game.	[53]
	DA2	Aesthetic pleasure was provided by the visual design of the ancient architectural elements.
	DA3	The crafted scenes of ancient architectural sites, along with the game’s artwork and sound effects, enabled deeper immersion in the game world.
	DA4	Engagement with the game’s narrative was enhanced by the integration of ancient architectural cultural elements.
Perceived Value (PV)	PV1	Significant cultural and educational value is afforded by CH-SGs.	[77,78,79]
	PV2	Positive psychological responses are elicited by the game experience.
	PV3	Participation in CH-SGs is perceived as a worthwhile investment of time.
	PV4	Deeper value is provided by CH-SGs in comparison with entertainment games.
Cultural Identity (CI)	CI1	A sense of pride was experienced regarding the architectural cultural heritage presented in CH-SGs.	[80]
	CI2	A willingness was expressed to further explore the cultural connotations of ancient architectural sites within CH-SGs.
	CI3	It is believed that the CH-SG gaming experience can effectively enhance national cultural confidence.
	CI4	Continued time will be devoted to exploring the cultural depictions of ancient architectural sites in the game.
Perceived Enjoyment (PE)	PE1	Participation in CH-SGs was perceived to confer mental relaxation.	[81,82,83]
	PE2	The gameplay process was perceived as highly enjoyable.
	PE3	The gaming experience was perceived to generate pleasurable memories.
	PE4	The overall engagement with CH-SGs was greatly enjoyed.
Game Continuance Intention (GCI)	GCI1	The continued engagement with CH-SGs was perceived as highly valuable.	[84,85]
	GCI2	It was intended that serious games for ancient architectural heritage would be played in the future.
	GCI3	It was planned to experience as many ancient architectural site scenes in CH-SGs as possible.
	GCI4	A strong recommendation of CH-SGs to others was intended.

. Notably, all items were adapted and refined from established scales to ensure their appropriateness for the CH-SG context. Responses were recorded on a seven-point Likert scale (1 = strongly disagree; 7 = strongly agree).

3.2.2. Formal Data Collection

Following the prescribed methodology, the study’s hypotheses were tested using survey data. Respondents were recruited via gaming forums and social media communities dedicated to serious games with historic architectural settings. Eligibility criteria required prior experience with such serious games. Of the 613 complete questionnaires initially collected, 125 were discarded due to data irregularities. An additional 25 responses were removed for incompleteness or duplication, yielding a final sample of 532 valid questionnaires for structural validation and hypothesis testing.

The data collection process of this study is shown in Figure 3.

3.3. Respondent Profile

As shown in

Table 2. Participants’ demographic information (n = 532).

Item	Number	Proportion
Gender
Male	286	53.76%
Female	246	46.24%
Age
18–25	227	42.67%
26–35	165	31.03%
35–45	97	18.23%
46–55	35	6.57%
55–60	8	1.51%
Education background
High school and below	53	9.96%
University college	148	27.82
University undergraduate	213	40.03%
Master and above	118	22.19%
Degree of familiarity with CH-SGs
Completely unfamiliar: never played	2	0.37%
Slightly unfamiliar: rarely played	16	3.01%
Medium: some experience	257	48.3%
Familiarity: often played	162	30.5%
Very familiar: play regularly	95	17.82%
The frequency of using CH-SG
Never experienced	5	0.93%
1–2 times	49	9.21%
3–5 times	82	15.41%
6–10 times	169	31.77%
More than 10 times	227	42.68%
Total	532	100%

, a total of 532 valid questionnaires were obtained. The sample was predominantly male (53.76%), with a strong concentration in younger age cohorts: 42.67% were aged 18–25, and 31.03% were aged 26–35, together comprising 73.70% of respondents. Educational attainment was high, with undergraduates (40.03%) and associate-degree holders (27.82%) accounting for 67.85% of the sample, and master’s degrees or higher representing 22.19%. Notably, the sample demonstrated deep engagement with serious games for ancient architecture (CH-SGs): nearly half of respondents (48.30%) reported moderate experience levels, while 48.32% indicated high familiarity (“familiar” 30.50%; “very familiar” 17.82%). In terms of usage frequency, 74.45% of participants had played CH-SGs more than six times (31.77% for 6–10 sessions; 42.68% for more than 10 sessions), and only 0.93% had no prior experience. This profile confirms that the study’s participants were primarily young, mid- to high-frequency CH-SG users, providing a robust empirical basis for examining continuance intentions. In addition, a late-respondent versus early-respondent extrapolation test [86] revealed no significant differences in demographic or substantive variables, indicating that nonresponse bias was not a concern.

4. Results

4.1. PLS-SEM Results

Following the procedures outlined by [87,88], a two-step modeling approach was conducted using SmartPLS 4.1 software [89]. In the first step, the measurement model was assessed using the PLS algorithm, and in the second step, the research hypotheses were evaluated via a bootstrap resampling procedure.

4.1.1. Assessment of Measurement Model

Composite reliability (CR) and Cronbach’s α were employed to assess the internal consistency reliability of the measurement model. Following the recommendation of [90], CR and Cronbach’s α values exceeding 0.70 are indicative of satisfactory reliability. As shown in

Table 3. Cronbach’s α, corporate reliability, and average variance extracted.

Variables	Cronbach’s Alpha	Composite Reliability	AVE	VIF	Items	Factor Loading
Authenticity of Cultural Elements	0.908	0.908	0.844	3.177	ACE1	0.916
				3.790	ACE2	0.937
				2.574	ACE3	0.902
Knowledge Acquisition	0.916	0.919	0.799	2.683	KA1	0.881
				3.537	KA2	0.913
				3.965	KA3	0.922
				2.661	KA4	0.858
Perceived Playfulness	0.908	0.910	0.784	2.627	PP1	0.859
				3.346	PP2	0.899
				3.136	PP3	0.895
				2.835	PP4	0.887
Design Aesthetics	0.907	0.914	0.782	2.302	DA1	0.846
				3.159	DA2	0.898
				3.264	DA3	0.914
				2.576	DA4	0.878
Perceived Value	0.850	0.863	0.688	1.789	PV1	0.828
				2.166	PV2	0.857
				2.446	PV3	0.864
				1.885	PV4	0.767
Cultural Identity	0.875	0.880	0.728	2.356	CI1	0.852
				2.956	CI 2	0.896
				2.370	CI 3	0.861
				1.909	CI 4	0.801
Perceived Enjoyment	0.861	0.871	0.706	1.928	PE1	0.837
				2.334	PE2	0.868
				2.395	PE3	0.857
				1.994	PE4	0.796
Game Continuance Intention	0.893	0.903	0.757	2.367	GCI1	0.872
				2.891	GCI 2	0.899
				2.959	GCI 3	0.894
				2.092	GCI 4	0.813

, all latent constructs in the present study exhibited CR and Cronbach’s α coefficients above 0.80, thereby demonstrating excellent internal consistency reliability and satisfying rigorous academic standards.

Additionally, high inter-variable correlations can bias parameter estimates and significance levels in the structural model. Therefore, variance inflation factors (VIFs) were computed to assess multicollinearity. A VIF threshold of less than 5 was applied to indicate the absence of problematic collinearity [91]. As reported in Table 3, all VIF values for the key constructs were below 5, confirming that no multicollinearity issues were present in the structural model.

Additionally, convergent and discriminant validity were also assessed to evaluate the measurement scales. Convergent validity was examined via factor loadings, with loadings greater than 0.70 indicating adequate representation of the latent constructs and demonstrating strong convergent validity. Discriminant validity was assessed using the Fornell–Larcker criterion and cross-loading analysis. According to the Fornell–Larcker standard, a latent construct exhibits discriminant validity if its average variance extracted (AVE) exceeds 0.50 and if the square root of its AVE is greater than its correlations with other constructs. As shown in

Table 4. Fornell–Larcker criterion.

	ACE	CI	DA	GCI	KA	PE	PP	PV
ACE	0.919
CI	0.289	0.853
DA	0.044	0.229	0.884
GCI	0.177	0.407	0.228	0.870
KA	0.000	0.283	−0.054	0.185	0.894
PE	0.229	0.242	0.185	0.369	0.318	0.840
PP	−0.006	0.221	−0.045	0.204	0.059	0.284	0.885
PV	0.138	0.147	0.125	0.414	0.139	0.188	0.156	0.830

Note: ACE = authenticity of cultural elements, CI = cultural identity, DA = design aesthetics, GCI = game continuance intention, KA = knowledge acquisition, PE = perceived enjoyment, PP = perceived playfulness, PV = perceived value.

, all factor loadings and AVE values satisfied these requirements, thereby confirming the measurement model’s convergent and discriminant validity [92].

4.1.2. Assessment of Structural Model

The structural model was evaluated by examining path coefficients, the coefficient of determination (R²), and predictive relevance (Q²). A bootstrap procedure with 5000 resamples was performed using SmartPLS 4.0. As shown in

Table 5. A summary of the results of hypotheses testing.

Hypothesis	Path Coefficient	p-Values	t-Values	Confidence Interval	Hypothesis	Result
				2.5%	97.5%
ACE→PV	0.133 ***	0.001	3.310	0.052	0.209	Supported
ACE→CI	0.280 ***	0.000	7.699	0.208	0.349	Supported
ACE→PE	0.222 ***	0.000	5.957	0145	0.293	Supported
KA→PV	0.137 **	0.003	2.927	0.042	0.224	Supported
KA→CI	0.283 ***	0.000	8.515	0.218	0.348	Supported
KA→PE	0.313 ***	0.000	8.321	0.237	0.385	Supported
PP→PV	0.155 ***	0.000	4.042	0.080	0.230	Supported
PP→CI	0.217 ***	0.000	6.028	0.145	0.286	Supported
PP→ PE	0.276 ***	0.000	7.349	0.197	0.345	Supported
DA→PV	0.133 **	0.001	3.328	0.052	0.210	Supported
DA→CI	0.242 ***	0.000	6.481	0.169	0.312	Supported
DA→PE	0.205 ***	0.000	5.726	0.134	0.275	Supported
PV→GCI	0.325 ***	0.000	7.441	0.239	0.412	Supported
CI→GCI	0.302 ***	0.000	7.833	0.227	0.376	Supported
PE→GCI	0.235 ***	0.000	6.424	0.164	0.306	Supported

Note: *** p < 0.001, ** p < 0.01.

, all 15 hypotheses received support to varying degrees.

At the level of functional features (S), the authenticity of cultural elements was found to exert significant positive effects on perceived value (H1a: β = 0.133, p < 0.01), cultural identity (H1b: β = 0.280, p < 0.001), and perceived enjoyment (H1c: β = 0.222, p < 0.001). Moreover, knowledge acquisition was shown to positively influence perceived value (H2a: β = 0.137, p < 0.01), cultural identity (H2b: β = 0.283, p < 0.001), and perceived enjoyment (H2c: β = 0.313, p < 0.001). At the level of hedonic features, perceived interest was positively associated with perceived value (H3a: β = 0.155, p < 0.001), cultural identity (H3b: β = 0.217, p < 0.001), and perceived enjoyment (H3c: β = 0.276, p < 0.001). Additionally, design aesthetics demonstrated significant positive relationships with perceived value (H4a: β = 0.155, p < 0.01), cultural identity (H4b: β = 0.217, p < 0.001), and perceived enjoyment (H4c: β = 0.276, p < 0.001).

Within the organism (O) dimension, perceived value was found to be significantly and positively related to continuance intention (H5: β = 0.325, p < 0.001), as were cultural identity (H6: β = 0.302, p < 0.001) and perceived enjoyment (H7: β = 0.235, p < 0.001).

As illustrated in Figure 4, the model accounted for 34.4% of the variance in continuance intention toward serious games for cultural heritage (R² = 0.344), 7.8% of the variance in perceived value (R² = 0.078), 26.4% of the variance in cultural identity (R² = 0.264), and 26.7% of the variance in perceived enjoyment (R² = 0.267). Predictive relevance (Q²) was assessed using a blindfolding procedure for the four endogenous constructs. All Q² values exceeded zero-continuance intention (Q² = 0.149), perceived value (Q² = 0.063), cultural identity (Q² = 0.252), and perceived enjoyment (Q² = 0.253), indicating that the model possesses adequate predictive accuracy [88].

4.2. ANN Results

4.2.1. Model Building

Because structural equation modeling (SEM) assumes linear and compensatory relationships among constructs [93], its accuracy has been increasingly challenged in recent years. This assumption overly simplifies the complex, multifactorial decision-making processes involved. To address this limitation and enhance predictive accuracy, artificial neural network (ANN) analysis was employed to capture noncompensatory and nonlinear relationships among variables. Drawing on [94] guidance, and using the SEM results (

Table 6. RMSE values for models A, B, C, and D.

Neural Network	Model A		Model B		Model C		Model D
	Input: ACE, KA, PP, DA		Input: ACE, KA, PP, DA		Input: ACE, KA, PP, DA		Input: PV, CI, PE
	Output: PV		Output: CI		Output: PE		Output: GCI
	Training	Testing	Training	Testing	Training	Testing	Training	Testing
ANN1	0.468	0.430	0.374	0.341	0.368	0.369	0.325	0.355
ANN2	0.453	0.536	0.366	0.349	0.371	0.396	0.317	0.375
ANN3	0.446	0.688	0.369	0.345	0.378	0.274	0.334	0.296
ANN4	0.467	0.629	0.380	0.544	0.369	0.388	0.332	0.296
ANN5	0.438	0.468	0.379	0.377	0.379	0.349	0.344	0.211
ANN6	0.501	0.346	0.375	0.291	0.443	0.414	0.348	0.326
ANN7	0.483	0.401	0.371	0.408	0.372	0.320	0.337	0.284
ANN8	0.465	0.956	0.364	0.509	0.375	0.434	0.315	0.557
ANN9	0.463	0.389	0.372	0.370	0.376	0.312	0.332	0.335
ANN10	0.494	0.380	0.366	0.356	0.401	0.407	0.334	0.416
Mean	0.468	0.522	0.372	0.393	0.381	0.362	0.332	0.345
SD	0.020	0.189	0.005	0.083	0.023	0.052	0.011	0.093

and Figure 4) as a foundation, multilayer perceptron (MLP) networks were trained in SPSS 27.0. Four ANN models were developed (see Appendix A Figure A1) to predict perceived value (PV), cultural identity (perceived usefulness), perceived enjoyment (PE), and game continuance intention (GCI), respectively.

4.2.2. Validation of ANN

To avoid overfitting the model, we used a ten-fold cross-validation method for testing. In this process, 90% of the data is used for training and 10% for testing [95]. The results indicated that the RMSE values for both training and test sets across ANN models A, B, C, and D ranged from 0.332 to 0.522 (see Table 6). Although these results demonstrate acceptable predictive performance, model A exhibited a comparatively higher RMSE (mean testing RMSE = 0.522), which may be attributed to the inherent subjectivity of users’ perceived value and heterogeneity in individual evaluations, thus posing challenges to prediction precision. Nevertheless, the observed RMSE values remain within acceptable limits, suggesting that the ANN models employed in this study maintain satisfactory reliability.

Coefficients of determination (R²) were calculated using Equations (1) and (2) to assess model fit. The ANN models developed in this study explained 7.22%, 54.23%, 56.57%, and 68.32% of the variance in cultural identity, perceived usefulness, perceived enjoyment, and game user engagement, respectively. All four ANN models exhibited high goodness-of-fit and demonstrated strong generalization capability.

$$RMSE=\sqrt{MSE}={\displaystyle \frac{SSE}{n}}$$

(1)

$$R^2=1-{\displaystyle \frac{SSE}{SST}}$$

(2)

In Equation (1), MSE represents the mean squared error, SSE represents the sum of squared errors, and n refers to the sample size of both the training and testing datasets.

4.2.3. Sensitivity Analysis

A sensitivity analysis was conducted in this study using the permutation method to determine the importance of each factor within the models [96]. As shown in

Table 7. Sensitivity analysis of models A and B.

Neural Network	Model A (Output: PV)				Model B (Output: CI)
	ACE	KA	PP	DA	ACE	KA	PP	DA
ANN1	0.234	0.240	0.239	0.287	0.224	0.291	0.253	0.231
ANN2	0.227	0.183	0.302	0.289	0.270	0.252	0.208	0.271
ANN3	0.230	0.196	0.337	0.237	0.257	0.270	0.210	0.264
ANN4	0.362	0.200	0.164	0.274	0.285	0.303	0.210	0.202
ANN5	0.300	0.149	0.266	0.285	0.267	0.265	0.245	0.224
ANN6	0.242	0.097	0.546	0.115	0.254	0.309	0.212	0.225
ANN7	0.218	0.226	0.101	0.455	0.260	0.294	0.221	0.225
ANN8	0.313	0.159	0.311	0.217	0.314	0.277	0.166	0.243
ANN9	0.325	0.194	0.232	0.249	0.248	0.253	0.224	0.275
ANN10	0.182	0.349	0.232	0.236	0.287	0.284	0.224	0.205
RI	0.263	0.199	0.273	0.264	0.267	0.280	0.217	0.237
NI (%)	96.336	72.893	100.000	96.703	95.281	100.000	77.663	84.532

and

Table 8. Sensitivity analysis of models C and D.

Neural Network	Model C (Output: PE)				Model D (Output: GCI)
	ACE	KA	PP	DA	PV	CI	PE
ANN1	0.231	0.257	0.274	0.238	0.483	0.263	0.253
ANN2	0.227	0.274	0.274	0.224	0.436	0.300	0.265
ANN3	0.221	0.282	0.271	0.236	0.261	0.340	0.399
ANN4	0.218	0.281	0.282	0.219	0.369	0.307	0.324
ANN5	0.201	0.302	0.273	0.224	0.422	0.378	0.200
ANN6	0.424	0.087	0.391	0.097	0.366	0.392	0.242
ANN7	0.242	0.265	0.272	0.221	0.370	0.339	0.291
ANN8	0.187	0.285	0.273	0.256	0.377	0.298	0.325
ANN9	0.209	0.296	0.284	0.211	0.385	0.353	0.262
ANN10	0.337	0350	0.292	0.020	0.451	0.316	0.233
RI	0.250	0.268	0.289	0.195	0.392	0.329	0.279
NI (%)	86.505	92.733	100.000	67.474	100.000	83.928	71.177

, the importance rankings of the predictive variables for ANN model A were as follows: PP (100%), DA (96.703%), ACE (96.336%), and KA (72.893%). For ANN model B, the importance rankings were KA (100%), ACE (95.281%), DA (84.532%), and PP (77.663%). In ANN model C, the predictive variable importance was ranked as follows: PP (100%), KA (92.733%), ACE (86.505%), and DA (67.474%). Finally, for ANN model D, the importance order of the predictive variables was PV (100%), CI (83.928%), and PE (71.117%).

To compare the results of the two methods, the study further conducted ordinal comparisons between the path coefficients in the structural equation modeling (SEM) and the normalized relative importance in the artificial neural networks (ANNs) (see

Table 9. Comparison between PLS-SEM and ANN results.

PLS-SEM Path	PLS-SEM Path Coefficient	ANN Normalized Relative Importance (%)	Ranking (PLS-SEM)	Ranking (ANN)	Remark
Model A (Output: PV)
ACE→PV	0.133	96.336	3	3	Match
KA→PV	0.137	72.893	2	4
PP→PV	0.155	100.000	1	1	Match
DA→PV	0.133	96.703	4	2
Model B (Output: CI)
ACE→CI	0.280	95.281	2	2	Match
KA→CI	0.283	100.000	1	1	Match
PP→CI	0.217	77.663	4	4	Match
DA→CI	0.242	84.532	3	3	Match
Model C (Output: PE)
ACE→PE	0.222	86.505	3	3	Match
KA→PE	0.313	92.733	1	2
PP→PE	0.276	100.000	2	1
DA→PE	0.205	67.474	4	4	Match
Model D (Output: GCI)
PV→GCI	0.325	100.000	1	1	Match
CI→GCI	0.302	83.928	2	2	Match
PE→GCI	0.235	71.177	3	3	Match

). The results revealed that for models B and D, the findings from SEM and ANN were fully consistent. However, slight discrepancies were observed between the SEM and ANN results for models A and C. Overall, the ANN demonstrated strong predictive capabilities, while also validating the explanatory power of the PLS-SEM results for the endogenous variables.

4.3. NCA Results

The sufficiency logic assumes that “X can enhance Y”, while the necessity logic assumes that “X is essential for Y”. The boundary conditions in SEM–ANN models cannot determine the necessary conditions that trigger the outcome variable. In contrast, the boundary conditions in NCA focus on measuring the impact of X on the changes in Y. Therefore, the combination of SEM–ANN–NCA methods serves to complement one another. Based on the Cartesian coordinate system (Figure 5), NCA maps the relationship between the predictor variables (X-axis) and the outcomes (Y-axis). The ceiling line divides the coordinates into two regions, based on the presence or absence of observations, and is drawn using the ceiling envelope without disposability hull (CE-FDH), forming a stepwise function [97]. The top left corner identifies the degree required for X to reach a certain level of Y.

4.3.1. Effect Size and Significance Testing

When the effect size (d) of a single condition is greater than 0.1 and the p-value test reveals that the effect size is statistically significant (p < 0.05), the predictor is considered necessary for the outcome [98]. As shown in

Table 10. NCA effect sizes (method: CE-FDH).

Parameters	Ceiling Zone	Scope	Effect Size	p-Value
ACE→PV	2.080	25.0	0.083	0.481
KA→PV	1.390	18.5	0.075	0.566
PP→PV	2.360	19.0	0.124	0.020
DA→PV	1.710	20.0	0.086	0.107
ACE→CI	1.690	22.5	0.075	0.031
KA→CI	1.190	16.7	0.071	0.037
PP→CI	1.450	17.1	0.085	0.009
DA→CI	1.040	18.0	0.058	0.044
ACE→PE	2.410	22.5	0.107	0.001
KA→PE	1.250	16.7	0.075	0.148
PP→PE	1.270	17.1	0.074	0.193
DA→PE	1.100	18.0	0.061	0.160
PV→GCI	4.560	25.0	0.182	0.250
CI→GCI	4.670	22.5	0.208	0.114
PE→GCI	6.450	22.5	0.287	0.026

, three variables were thus identified as necessary conditions. First, perceived playfulness (PP) was found to be necessary for perceived value (PV) (d = 0.124, p < 0.05), indicating that users must experience a sufficient level of playfulness in order to attain satisfactory perceived value. Second, authenticity of cultural elements (ACE) was determined to be necessary for perceived enjoyment (PE) (d = 0.107, p < 0.01), underscoring that without the integration of adequately authentic cultural elements into CH-SGs, high levels of enjoyment cannot be achieved regardless of other design quality. Most notably, perceived enjoyment (PE) demonstrated the strongest necessary-condition effect for game continuance intention (GCI) (d = 0.287, p < 0.05), suggesting that enjoyment constitutes the core threshold for sustaining user engagement; only when enjoyment surpasses this critical minimum are users likely to maintain ongoing play intentions.

4.3.2. Bottleneck Analysis

Bottleneck analysis aims to identify the minimum threshold of predictor variables required to achieve a specific target level [99]. As shown in

Table 11. Bottleneck (in percentage).

	ACE	KA	PP	DA	CI	PE	PU
Perceived Value
0	NN	NN	NN	NN
10	NN	NN	NN	NN
20	NN	NN	NN	NN
30	NN	NN	NN	NN
40	NN	NN	NN	NN
50	NN	NN	NN	NN
60	NN	NN	NN	NN
70	NN	NN	NN	NN
80	14.0	6.8	24.4	8.1
90	30.4	34.7	51.2	36.0
100	46.8	62.7	78.8	63.9
Cultural Identity
0	NN	NN	NN	NN
10	NN	NN	NN	NN
20	NN	NN	NN	NN
30	NN	NN	NN	NN
40	NN	NN	NN	NN
50	NN	NN	NN	NN
60	NN	NN	NN	NN
70	NN	NN	NN	NN
80	13.8	8.3	3.3	7.6
90	28.3	26.7	30.3	23.9
100	42.7	45.1	57.3	40.2
Perceived Enjoyment
0	NN	NN	NN	NN
10	NN	NN	NN	NN
20	NN	NN	NN	NN
30	NN	NN	NN	NN
40	NN	NN	NN	NN
50	NN	NN	NN	NN
60	NN	NN	NN	NN
70	NN	NN	NN	NN
80	15.3	NN	2.3	NN
90	45.3	25.5	31.9	24.1
100	75.3	51.8	61.5	51.7
Game Continuance Intention
0					NN	NN	NN
10					NN	NN	NN
20					NN	NN	NN
30					NN	NN	NN
40					NN	NN	NN
50					NN	NN	7.4
60					NN	NN	22.4
70					14.4	21.3	37.3
80					37.8	43.4	52.3
90					61.2	65.5	67.2
100					84.6	87.6	82.1

Note: NN represents not necessary.

, achieving a high level of continuance intention (71–100%) requires that perceived value, cultural identity, and perceived enjoyment reach at least 14.4%, 21.3%, and 37.3%, respectively. Additionally, to attain a high level of perceived enjoyment, cultural authenticity must be ≥15.3%, knowledge acquisition ≥ 25.5%, perceived interest ≥ 2.3%, and design aesthetics ≥ 24.1%. Notably, the threshold requirement for perceived enjoyment in relation to continuance intention is the highest (37.3%), and a significant bottleneck effect is observed for perceived interest, which affects both perceived value (24.4%) and perceived enjoyment (24.1%).

Additionally, Figure A2 presents NCA scatter plots generated in R Studio 4.5.0, which further confirm the necessity of the aforementioned thresholds.

Integration of the PLS-SEM and NCA findings permits the classification of factors into four categories (Figure 6) [100]. The conceptual model developed in this study encompasses two of these categories. The first category consists of factors that are both “should-have” and “must-have” (located in the upper-right quadrant), namely, perceived interest → perceived value; cultural authenticity → perceived enjoyment; and perceived enjoyment → continuance intention toward CH-SGs. For this category, it is posited that increases in the levels of these predictors will facilitate the occurrence of the corresponding outcomes, while the manifestation of those outcomes also requires that predictor levels exceed a minimum threshold. The second category comprises factors that are “should-have” but not strictly “must-have” (located in the lower-left quadrant), which include all hypothesized paths H1–H15, each exhibiting significance to varying degrees. In this category, it is asserted that enhancements in predictor levels will produce corresponding increases in the expected outcomes.

5. Discussion

This study systematically examined the mechanisms underlying players’ continuance intentions in cultural heritage serious games (CH-SGs) focused on ancient architecture, thereby validating the hypothesized relationships. Functional features (cultural authenticity, knowledge acquisition) and hedonic features (perceived playfulness, design aesthetics) were integrated into the research model to incorporate cultural–cognitive experiences and to reveal potential pathways for the digital transmission of architectural heritage and informal learning. Empirical analyses were then performed using a hybrid SEM–ANN–NCA approach, with the results cross-validated and critically discussed.

Specifically, perceived value, cultural identity, and perceived enjoyment each exerted significant positive effects on players’ continuance intentions toward cultural heritage serious games (CH-SGs), thereby supplementing earlier findings [57,63,70]. Among these factors, perceived value had the strongest impact, a result consistent with [59,101]. Prior studies have underscored the pivotal role of perceived value in serious-game contexts, indicating that the greater the value users derive in a gamified learning environment, the more likely they are to believe that CH-SGs yield tangible benefits—and, consequently, to maintain their intention to continue playing [11]. Thus, CH-SGs must confer a clear sense of value; otherwise, sustained participation is unlikely to be stimulated. As anticipated, cultural identity also emerged as a key determinant of continuance intention in ancient-architecture CH-SGs. This conclusion supports [102], who reported that cultural identity significantly strengthens users’ emotional ties to, and identification with, cultural heritage, making it a critical driver of sustained behavioral intentions in CH-SG experiences. Notably, although perceived enjoyment positively influenced continuance intention, its effect was not the most pronounced. This finding complements [43], revealing that within serious-game learning settings, enjoyment can indeed spur active participation, but its influence is weaker than that of value- and culture-oriented drivers [68,69]. Accordingly, the design and development of CH-SGs should balance cultural, educational, and entertainment requirements to enhance player stickiness. In practice, designers should build robust heritage databases and conduct user-behavior research to ensure informational accuracy and alignment with learners’ needs, while also employing diverse CH-game mechanics and rich narrative structures to render otherwise tedious knowledge engaging.

Secondly, this study’s findings are discussed with respect to both the functional and hedonic features of cultural-heritage serious games. On the functional dimension, the results confirmed that cultural authenticity and knowledge acquisition exert direct, positive effects on perceived value, cultural identity, and perceived enjoyment in serious games for ancient architectural heritage; this finding is consistent with prior research [43,103]. This effect can be attributed to the inherently strong cultural attributes of such games. The authenticity and accuracy of heritage elements determine the educational quality and directly shape users’ value perceptions. Accordingly, developers should honor the integrity of cultural heritage by avoiding excessive gamification, balancing creative reinterpretation with factual fidelity, and accurately conveying cultural values to facilitate players’ understanding and acquisition of heritage knowledge—thereby crafting a truly immersive experience. On the hedonic dimension, perceived interest and design aesthetics were shown to function as environmental stimuli that positively influence perceived value, cultural identity, and perceived enjoyment in CH-SGs; this aligns with previous findings [104]. Interest enhances intrinsic motivation by stimulating curiosity and autonomy, while aesthetics reinforce positive evaluations of cultural symbols through coherent visual and auditory design. It is noteworthy that, although serious games leverage their “fun” aspects to promote efficient learning of “serious” content—rendering pedagogy and entertainment mutually reinforcing [105]—developers of serious games centered on architectural heritage must place particular emphasis on aesthetic design. Thus, integrating culturally rich, visually appealing, and engaging content is essential in CH-SG development.

Moreover, the ANN analysis elucidated the relative importance of functional features, hedonic features, and user-experience factors in CH-SGs. By addressing the limitations of SEM, ANN provided a more comprehensive understanding of players’ continuance intentions and behaviors [106]. Perceived value emerged as the most influential determinant of CH-SG engagement, while heritage knowledge acquisition was shown to be critical for both perceived value and cultural identity. Subsequent NCA revealed that perceived interest constitutes a core and necessary condition for sustained CH-SG use. It was also found that cultural authenticity is a necessary condition for perceived enjoyment, indicating that the verisimilitude of cultural elements not only deepens immersion and knowledge acquisition but also engenders enjoyment. In turn, perceived enjoyment was identified as a necessary precursor to continuance intention, thereby establishing a chained mediating effect. Across SEM, ANN, and NCA analyses, cultural authenticity was consistently shown—via the mediating role of perceived enjoyment—to influence continuance intention toward CH-SGs. This insight holds profound implications for the future design and development of cultural-heritage serious games.

However, several limitations should be acknowledged. Although the hybrid SEM–ANN–NCA approach integrates multiple analytical perspectives, the ANN component functions as a “black box”, and its nonlinear mapping process cannot be directly interpreted; at present, only feature-importance rankings have been employed to infer interpretations indirectly. Furthermore, participants were recruited via Chinese gaming forums and social media communities, resulting in a culturally homogeneous sample that may limit the cross-cultural generalizability of the findings. Future research is encouraged to employ longitudinal and experimental designs, to recruit more culturally diverse samples, and to incorporate explainable AI techniques—such as SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME)—to enhance the transparency and interpretability of ANN results.

6. Implications, Limitations and Conclusions

6.1. Implications

First, at the theoretical level, this study extends the S–O–R framework by uncovering novel mechanisms underlying user behavior in serious games for ancient architectural heritage (CH-SGs). Cultural authenticity and knowledge acquisition are innovatively introduced as core stimulus constructs, and—together with individual factors such as perceived value and cultural identity—a dynamic mediational pathway of “stimulus → organismic response → continuance intention” is specified [11]. Unlike prior research that has primarily examined the direct effects of the experience economy [7], the present model empirically demonstrates that cultural elements influence continuance intention via the mediating role of perceived enjoyment. This finding deepens understanding of CH-SG user behavior mechanisms and reaffirms the explanatory power of the S–O–R framework in this domain.

Second, from a methodological standpoint, an integrated SEM–ANN–NCA model was proposed to overcome the limitations inherent in single-method approaches. SEM was employed to test linear hypotheses, ANN to capture nonlinear relationships (e.g., threshold effects in perceived value), and NCA to identify necessary conditions (e.g., the necessity of interest for continuance intention). This hybrid approach successfully addressed SEM’s inability to detect nonlinear effects [107] and NCA’s incapacity for sufficiency analysis [108], thereby constituting both an innovative extension of the SEM–NCA framework and an enhancement through multi-model comparative analysis. The result is a novel paradigm for the study of complex behavioral phenomena [109].

Finally, at the practical level, and based on the results of the three-stage quantitative analyses (

Table 12. Summary of results.

Relationship	PLS-SEM	ANN	NCA
ACE→PV	Significant	Third important	Not Necessary
KA→PV	Significant	Fourth important	Not Necessary
PP→PV	Significant	First important	Necessary
DA→PV	Significant	Second important	Not Necessary
ACE→CI	Significant	Second important	Not Necessary
KA→CI	Significant	First important	Not Necessary
PP→CI	Significant	Fourth important	Not Necessary
DA→CI	Significant	Third important	Not Necessary
ACE→PE	Significant	Third important	Necessary
KA→PE	Significant	Second important	Not Necessary
PP→PE	Significant	First important	Not Necessary
DA→PE	Significant	Fourth important	Not Necessary
PV→GCI	Significant	First important	Not Necessary
CI→GCI	Significant	Second important	Not Necessary
PE→GCI	Significant	Third important	Necessary

Note: ANN = artificial neural network; PLS-SEM = partial least squares-structural equation modeling; NCA = necessary condition analysis.

), a tiered development strategy is proposed.

For developers, the core “must-have” elements identified—namely, perceived playfulness, cultural authenticity, and perceived enjoyment—should be incorporated into design priorities and continuously refined through user testing to optimize the game experience. For policymakers, a comprehensive support system for the gamification of digital cultural heritage should be established, encompassing targeted funding mechanisms, the promotion of multilingual and localized development, the formulation of technical and content standards, and the creation of cross-sector collaboration platforms among government, industry, academia, and research institutions, thereby facilitating the broad application and sustainable development of CH-SGs in education, cultural dissemination, and heritage preservation.

6.2. Limitation and Future Research

Despite the theoretical, practical, and methodological contributions of this study, there are several limitations to consider. Firstly, the cultural representativeness of the sample was constrained, as data were collected exclusively from Chinese players, thereby limiting the generalizability of the findings to contexts with markedly different cultural backgrounds. Secondly, self-report questionnaires were employed to assess players’ psychological processes; such instruments capture only subjective perceptions and cannot directly reveal underlying cognitive mechanisms. Finally, a cross-sectional design was adopted to examine the drivers of continuance intention in cultural heritage serious games (CH-SGs), and temporal effects were not considered.

In light of the limitations of this study, future research can explore the following avenues to guide research directions. Firstly, participants from diverse cultural contexts should be recruited to examine cultural differences and commonalities in players’ continuance intentions toward CH-SGs. Secondly, neuroscientific techniques—such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI)—should be integrated into experimental designs to obtain more precise measurements of users’ psychological processes during CH-SG gameplay. By recording brain activity and event-related potentials (ERPs), a more comprehensive understanding of players’ cognitive processes—including attention, emotion, and decision making—can be achieved, and their relationships with continuance intentions can be explored. Finally, to investigate temporal effects on continuance intentions, a longitudinal study design should be employed to track changes in players’ psychological trajectories over multiple time points. Combining survey methods with focus group discussions would enable the capture of dynamic user experiences and provide deeper insights into the evolution of decision-making intentions.

6.3. Conclusions

This study extends the S–O–R framework to investigate the underlying mechanisms driving players’ continuance intentions in serious games for ancient architectural heritage (CH-SGs). A hybrid methodology integrating PLS-SEM, ANN, and NCA was employed to offset the limitations of any single technique: PLS-SEM verified linear hypotheses, ANN captured nonlinear effects, and NCA identified necessary conditions (e.g., the necessity of cultural authenticity for perceived enjoyment). Empirical findings indicate that cultural authenticity (ACE), knowledge acquisition (KA), perceived interest (PP), and design aesthetics (DA) each exert significant positive effects on perceived value (PV), cultural identity (CI), and perceived enjoyment (PE), with PV and CI exhibiting the strongest influences. ANN analysis further confirmed that PV is the most powerful predictor of continuance intention, while NCA revealed that ACE and PP indirectly drive behavior via PE, thereby delineating the “must-have” elements for CH-SG design. Based on these results, a tiered development strategy grounded in necessity thresholds is proposed to theoretically support the enhancement of player retention in serious games for ancient architectural heritage.

Notably, serious games for cultural heritage (CH-SGs) enable users to be virtually immersed in historical environments, thereby reaching a global audience and fostering widespread appreciation for ancient architectural sites. Such tools can complement physical visits by raising players’ awareness of preservation and cultural respect, ultimately transforming them into advocates for heritage conservation. In sum, this research highlights the promise of integrating digital innovation with cultural-heritage interpretation—a strategy that can stimulate tourism interest while deepening the impact of conservation efforts. These insights not only address gaps in the literature but also inspire further exploration of gamified narratives, affective design, and their capacity to achieve a balanced synergy between player engagement and heritage preservation.