Understanding, expressing, and communicating emotions constitute fundamental human abilities. Integrating positive emotional elements into design significantly enhances users’ emotional experiences (Yoon et al., 2020). In his bestseller, “Emotional Design: Why We Love (or Hate) Everyday Things,” Norman (2004) introduced an innovative theory of emotional design (EDT), challenging the preceding practicality and usability paradigm. Norman hypothesized the heightened importance of the emotional aspect in design, proposing three interdependent design levels: visceral design, behavioral design, and reflective design (Norman, 2004). Norman (2004) elucidated these concepts as follows: Visceral design, preceding consciousness and cognition, lays the foundation for esthetics, appearances, and initial impressions encompassing visual and auditory senses. Behavioral design pertains to user experience, including functionality, performance, and usability, representing external expressions of user pleasure and utilitarian emotions. Reflective design involves consciousness, higher emotions, and perceptions, which vary across cultures, experiences, education, and individual differences. However, EDT elucidated the impact of emotions on responses to design objects (Zachry, 2005), spanning product, interior, and game design. Therefore, when considering emotional studies in design, the consequential themes are emotional effects on user behavior and reflection (Yoon et al., 2020), especially in the realm of human-computer interaction (Imbesi, 2010). The intricate relationship between emotional responses to design and users’ visceral, behavioral, and reflective reactions directly influences the gaming experience of MOBA players. Abbasi et al. (2023) found that visually appealing design and graphics may lead to a positive attitude toward games and the games’ visual and auditory aspects may create a sensory experience that is positively related with game engagement. Patzer et al. (2020) examined the potential for game involvement in terms of behavioral manifestations such as gamer motivation and intention. In addition, gamers’ imagination, emotions, and sensory experiences could trigger cognitive, emotional, and behavioral engagement with the game (Bojan et al., 2021). The above studies show that games are designed with cognitive and affective factors, but cognition is related to behaviors such as motivation.

Hence, this paper based on the correlation between visceral perception (e.g., visual, auditory, tactile, and other sensory variables), behavior (e.g., player skill level and actual combat experience), reflection (e.g., player acquisition of game culture and information), and game addiction, the following hypotheses are posited:

Leisure is a distinct and subjective mental state, intricately linked to motivation and time (Voss, 1967). When an individual attains a specific psychological state of freedom outside of work, it is considered leisure time. In the early days of the Industrial Revolution, leisure was confined to the “idle class,” exempt from labor-related obligations (Carnevali and Strange, 2014). However, with technological advancements and increased productivity, employers reduced working hours, providing workers with more free time for leisure activities. The evolving work-leisure relationship prompted the study of Leisure Theory (LT) to adopt a multidisciplinary approach (Spracklen et al., 2017). While recent studies on leisure have embraced interdisciplinary fields such as sociology, geography, and culturology, scholars from each discipline have contributed varying definitions of leisure to conceptualize the concept and construct theoretical models. Despite increased scholarship to distinguish the relationship between leisure and other disciplines, this study introduces a theoretical model within leisure activities, specifically focusing on MOBA.

Digital technology has transformed the social dynamics of leisure activities, creating a leisure space intertwined with social attributes (Rojas de Francisco et al., 2016). Social media, recognized as a leisure activity by an increasing number of young people, plays a significant role in users’ lives (Andreassen, 2015; Kuss and Griffiths, 2017; Sharma et al., 2020). Authors have explored the impact of social media on individual users’ self-reflection and emotions within the framework of LT (Sharma et al., 2020). Variables such as age, education level, income, gender, occupation, and social class are commonly assessed for their influence on social media use (Whitfield et al., 2020). However, some studies emphasize social media’s role in enabling individuals to connect without constraints of time and space, contributing to social addiction (Kuss and Griffiths, 2017).

Gaming, identified as a new form of social media, serves as an entertainment platform abundant with leisure resources (Javier Cabeza-Ramirez et al., 2021), boasting an estimated two billion players globally (Matuszewski et al., 2020). In the virtual gaming world, individuals can momentarily escape reality, alleviating pressures from real life and making it a popular leisure activity for various age groups (Tobon et al., 2020). Consequently, studies on gaming culture have gained attention, highlighting the close relationship between leisure and social activities (Andreassen et al., 2017). Engaging in screen-based leisure is significantly linked to an individual’s physical activity, body mass index, social situation, and emotions (Li et al., 2020). Thus, this study investigates whether leisure influences gamers’ emotional experiences, potentially leading to addiction through leisure variables such as decompression, entertainment, and virtual engagement. The following hypotheses are posited:

User and Gratification Theory (UGT) is a mass communication theory employed to investigate societal behaviors and motives in the pursuit of specific media to fulfill their needs (Whiting and Williams, 2013). A multidimensional analysis often examines basic needs, individual differences, and cultural backgrounds to understand an individual’s interactions and intentions to use specific media (Rosengren, 1974). With the evolution of new and digital media, UGT is now applied to elucidate why and how individuals choose social media platforms like Instagram, Snapchat, Facebook, and Twitter (Sheldon and Bryant, 2016; Alhabash and Ma, 2017; Phua et al., 2017), games such as League of Legends (LOL) and Pokémon Go (Macedo and da Cunha, 2017; Hamari et al., 2019; Jang and Liu, 2019; Bueno et al., 2020), or video streaming sites like YouTube and Vimeo (Ruggiero, 2000; Khan, 2017; Golob et al., 2021). Not surprisingly, diverse outcomes are expected due to media use being based on individual differences (Greenberg et al., 2010).

UGT has recently been introduced to studies on game behavior and motivation, exploring factors that attract players to video games (Hamari et al., 2019; Jang and Liu, 2019; Bueno et al., 2020; Patzer et al., 2020). Previous research has established a connection between game gratification and motivation, as well as the intention of players’ sustained involvement in these games (Jang and Liu, 2019) and the level of satisfaction felt while playing (Phan et al., 2016). Using UGT, researchers have analyzed the relationship between player gratification and in-game consumption behavior (Ghazali et al., 2019; Hamari et al., 2019), in-game consumption behavior (Wang and Yu, 2017), as well as the connection between player viscosity and experience (Bueno et al., 2020).

If a player has a positive experience during the game, they will be focused, curious, interested, and willing to continue the gaming experience (Choi and Kim, 2004). This may lead to the possibility of further involvement and continuation in the game. Some researchers maintain there is a significant correlation between gratification and players’ motivation to continuously play these games (Abbasi et al., 2017). The motivation for game continuity and players’ long-term gratification experience with the game are the main reasons most players become addicted (Hamari and Keronen, 2017). A recent report by the World Health Organization highlighted the high risks of negative effects of game addiction on various aspects of players’ personal lives, including family, social, professional, and other related areas (Gros et al., 2020). Players’ curiosity about the game, their achievements, and gratification can influence their gaming experience, potentially leading to game addiction. Therefore, this study advances the following assumptions:

Building upon the review of these theoretical frameworks, it is proposed that leisure and gratification may exert a positive influence on the visceral perception, behavior, and reflection of MOBA game players. Furthermore, a consistently positive gaming experience is identified as the crucial factor in fostering prolonged engagement among MOBA game players. In light of the aforementioned assumptions, an exploratory research model (refer to Figure 1) for a emotional analysis of MOBA game addiction has been formulated.

A total of 298 samples were systematically gathered using a non-probabilistic sampling approach from 24th Nov. 2021 to 24th Mar. 2022 in China. Inclusion criteria encompassed individuals expressing a willingness to volunteer and those engaged in playing MOBA games, specifically League of Legends (LOL) and Honor of Kings (HOK), recognized as the prevailing MOBA games in China (Wirtz, 2022). Additionally, respondents fell within the age bracket of 12 to 40. According to 2017 data from the Nielsen Company, two-thirds of the U.S. population aged 13 and above identified as game players. Notably, 62 participants did not partake in MOBA games and were consequently excluded from the study. The dissemination of online questionnaires transpired randomly through Wenjuanxing, a platform akin to Amazon Mechanical Turk. The conclusive analysis was based on 236 validated questionnaires, equating to a commendable 79.2% response rate among the entire sample pool. It is crucial to note that the act of agreeing to participate in the experiment inherently signified informed consent.

The questionnaire was structured into two sections. The initial segment gathered socio-demographic data, encompassing information such as respondents’ gender, age, education level, employment status, salary, gaming experience, and proficiency. The second section comprised 43 five-point Likert scales, ranging from 1 (totally disagree) to 5 (totally agree). For the emotional analysis of gamers’ addictive behavior, a validated scale for design, adjustment, modification, and optimization was employed. The scales developed by this study are attached in Appendix.

In measuring the average weekly/daily MOBA gaming hours, survey items are adapted from “Usage (UG)” construct from Wu and Holsapple’s (2014) scale.

Visceral perception experience include visual experience and auditory experience. Therefore, survey items are adapted from the following studies: (1) survey items in “Sensory Experience” construct from Abbasi et al.’s (2019, 2021, 2023) measurement models are adapted; (2) survey items in “Sensory & imaginative Immersion” category from Phan et al.’s (2016) GUESS scale are adapted; (3) survey items from Tomlinson et al.’s (2018) BUZZ scale are adapted.

MOBA gamers’ behavioral performance involves both game skills and challenges. Therefore, survey items are adapted from the following studies: (1) survey items in “Behavioral intention to play (BI)” construct from Wu and Holsapple’s (2014) scale are adapted; (2) survey items concerning with skills and challenges from Phan et al.’s (2016) GUESS scale are adapted; (3) survey items in “Perceived behavioral control” construct from Hollebeek et al.’s (2022) measurement model are adapted; (4) survey items concerning with challenges from Bueno et al.’s (2020) scale are adapted.

Gamers’ reflection of the game include both perceptions of the game information and perceptions of the game culture. Therefore, regarding the measurement of gamers’ perception of information, the survey items are adapted from the following studies: (1) survey items in “Cognitive Engagement” construct from Abbasi et al.’s (2023) scale are adapted; (2) the item “I am likely to recommend this game to others.” from Phan et al.’s (2016) GUESS scale is adapted. For measuring cultural cognition, survey items are adapted from following studies: (1) survey items in “Cognitive Engagement” construct from Abbasi et al.’s (2023) scale are adapted; (2) survey items from the “Perceived behavioral control” construct from Hollebeek et al.’s (2022) measurement model are adapted; (3) survey items in “Game Knowledge” construct from Jang and Liu’s (2019) scale are adapted.

Playing games is one of the most popular forms of leisure activities Abbasi et al.’s (2021). Playing games can create escapism and relieve stress for gamers, and immerse themselves in the virtual worlds. A large number of gamers take games as a leisure activity to entertain themselves and spend their free time. Therefore, survey items are adapted from the following studies: (1) survey items concerning escapism and avatars from Abbasi et al.’s (2023) scale are adapted; (2) survey items in “Escapism” construct from Bueno et al.’s (2020) scale are adapted; (3) survey items concerning characters or avatars from Phan et al.’s (2016) GUESS and Segaran et al. (2021) scale are adapted.

Achievement and satisfaction are the two core factors that enhance MOBA gamers’ gratification. Therefore, regarding the measurement of satisfaction, survey items are adapted from the following studies: (1) survey items in “Enjoyment” construct from both Abbasi et al.’s (2023) and Hamari et al.’s (2019) scale are adapted; (2) survey items in “Entertainment” construct from Jang and Liu’s (2019) scale are adapted; (3) survey items in Patzer et al.’s (2020) GUESS Subscales are adapted.

For measurement of MOBA game addiction, survey items are adapted from the following studies: (1) survey items in “Continuance Intention (CI)” construct from Bueno et al.’s (2020) scale are adapted; (2) survey items in “Emotional involvement (EI)” construct from Wu and Holsapple’s (2014) scale are adapted; (3) survey items in “Continuance use intention (CUI)” construct from Jang and Liu’s (2019) scale are adapted.

In summary, addiction factors among MOBA gamers were analyzed in this study across six constructs: Leisure, Gratification, Visceral, Behavioral, Reflective, and Games Addiction. The construct of visceral include visual perception (VP) and auditory perception (AP). The construct of behavioral performance include skill (Sk) and challenge (Ch). The construct of reflective include sharing information (SI) and cultural (Cl) cognition. The construct of gratification include satisfaction (Sat) and achievement (Ach).

The socio-demographic analysis in this study was conducted by Statistical Package for the Social Sciences (SPSS, Version 26). For model evaluation, the Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to assess the relationships between indicators and the measurement models, as well as the structural models. PLS-SEM, known for its capacity to manage complex models and maximize the variance in dependent variables, was selected for its suitability in conducting formative measures on latent variables and fitting the exploratory nature of this theoretical model analysis (Hair et al., 2017; 116). The analysis utilized SmartPLS 4.1.0, involving two key stages:

In Stage 1, four First-order models, Gratification, Reflective, Behavioral and Visceral, were evaluated for measuring structure modeling. As reviewed above, Gratification contains two First-order measurement models, Ac and Sat; Reflective contains two First-order measurement models, Cl and SI; Behavioral contains two First-order measurement models, Ch and Sk; and Visceral contains two First-order measurement models, VP and AP. Thus, in the First-order measurement structural model, we evaluated Factor Loading (FL), Cronbach’s Alpha (Cα), Composite Reliability (CR) value, Average Variance Extracted (AVE) value, and Variance Inflation Factor (VIF) value, intending to assess the reliability and convergent validity of First-order measurement structural models.

In Stage 2, this study constructed a Second-order structural model consisting of six constructs: Leisure, Gratification, Visceral, Behavioral, Reflective, and Games Addiction. First, the reliability and convergent validity of its measurement constructs were assessed by the same assessment method as in Stage 1. Second, the indirect and overall effects between each construct were evaluated, assessing reflective constructs to determine the significance of the relationships posited in the hypotheses.

Demographic analysis revealed seven key variables among the respondents: gender, age, education, financial status, employment status, MOBA gaming skills, and years of playing MOBA games. The distribution between male (46.2%) and female (53.8%) participants indicated a balanced gender representation in MOBA gaming. The age range of 18 to 30 years dominated the sample, comprising 80.9%, with no minors participating. This age distribution aligns with findings from the Pew Research Center (2018) and the Nielsen Company (2017), which highlighted a predominant young adult demographic in gaming, especially among the Z Generation and Millennials (Dimock, 2019; Lim and Parker, 2020). A significant majority of the respondents were unmarried (91.1%) and well-educated, with over 78.0% holding a bachelor’s degree. This educational status correlates with the high percentage of college students (64.0%) in the sample, who generally have more leisure time compared to high school students and working adults. Employees formed the second-largest group at 25.9%. In terms of gaming proficiency, 75% of respondents were high-level players, and a substantial 83.0% had over 3 years of gaming experience. These demographic details are further elaborated in Table 1.

The statistics of gamers’ weekly or daily engagement in MOBA games are shown in Figure 2. This study measured the length of time related to gaming activities in gamers’ daily lives from four aspects: (1) the length of time that gamers spend on games per day on weekends; (2) the length of time that gamers spend on games per day on weekdays; (3) the length of time that gamers spend on watching videos related to MOBA games per day; and (4) the days that gamers spend on games every week.

Table 2 shows that among 236 samples, the proportion of gamers playing MOBA games for more than 2 h (2 ~ 3 h, 3 ~ 4 h, >5 h) on weekends (27.54, 16.95, 15.25%) is higher than that on weekdays (13.14, 2.54, 4.66%), with a distinct increase on weekends. The proportion of gamers playing MOBA games for less than 2 h (<1 h, 1-2 h) on weekends (13.56, 26.69%) is smaller than that on weekdays (44.92, 34.75%). It indicates that gamers are more likely to get engaged in MOBA games with longer leisure time on weekends.

Second, the average length of time gamers spend on watching MOBA online game videos in their daily life is around 1 ~ 2 h (Mean = 1.44, see Table 2), of which 89.83% of gamers watch MOBA online game videos for 0 ~ 2 h per day. It indicates that gamers also spend their leisure time watching MOBA online game videos when they are not engaged in MOBA games.

Third, as shown in Table 2, 58.90% of gamers spend 3 days in average playing MOBA games per week (Mean = 3.12). 80.30% of the gamers spend up to 3 h per day (Mean = 2.41) in playing MOBA games, indicating that gamers have developed the habit of playing MOBA games to some extent. This finding is consistent with the study by De Grove et al. (2016), who found that gamers’ behavior of repeated engagement in games contributed to the development of personal habits.

The convergence effectiveness of the conceptual model was evaluated using CR, Cα, AVE, and VIF. Abbasi et al. (2021) and Hair et al. (2019) established that for the structural model to pass reliability and validity tests, Cα, Rho A, and CR values should exceed 0.70, AVE should be greater than 0.50, while VIF should be lower than 5. As shown in Table 3, survey items (15 items) with FL value lower than 0.70 are deleted, leaving a total of 28 question items in the four First-order reflective models of Gratification, Reflective, Behavioral and Visceral with FL value (0.707 ~ 0.954) greater than 0.70.

With 15 survey items deleted, the Cα value (0.787 ~ 0.928) and CR value (0.794 ~ 0.928) of the eight constructs (AC, Sat, Cl, SI, AP, VP, Ch and Sk) were greater than 0.70, and the VIF value (1.504 ~ 4.964) was less than 5, which indicated that survey items in each First-order model had good reliability. Second, the AVE value (0.612 ~ 0.874) of each construct, which is greater than 0.50, indicates that these survey items and constructs have good convergent validity. Third, the Cα (0.885 ~ 0.891), AVE (0.557 ~ 0.653) and CR (0.869 ~ 0.891) values of the four constructs Gratification, Reflective, Behavioral and Visceral passed the test and the p values was less than 0.001, indicating that these four First-order models passed the reliability test.

In terms of differential validity assessment, Henseler et al. (2015) proposed a new assessment method, heterotrait-monotrait (HTMT), to measure the correlation between different constructs and thus assess the differential validity between each construct. In this case, an HTMT ratio below 0.85 indicates good discriminant validity between the constructs (Hair et al., 2019). In this study, the four First-order models are tested for differential validity separately, and the results are shown in Table 4, where the HTMT ratios between the constructs are significantly lower than 0.85, indicating that the four reflective First-order models of Gratification, Reflective, Behavioral and Visceral all have good differential validity.

In the Second-order reflective models analysis, the Partial Least Squares (PLS) algorithm was employed to assess the structural model’s reliability and validity. The maximum sub-sample size was set at 5000, and confidence intervals (CI) were calculated using the Percentile Bootstrapping method. The testing was conducted using a two-tailed approach, with a significance level set at p < 0.05. Table 5 displays FL and VIF value for each item, and constructs FL, mean, standard STDEV, t-values, p-values, CR, Cα, and AVE values for Leisure, Gratification, Visceral, Behavioral, Reflective and Games addiction.

The findings indicate that I-FL values (0.853–0.949) for each item are greater than 0.70, and VIF values (1.324–3.501) are less than 5, indicating good reliability of the measurement constructs. C-FL (0.744–0.895), CR (0.853–0.945), and Cα (0.662–0.895) values are greater than 0.70 for each of the six constructs. AVE values (0.744–0.895) greater than the minimum threshold of 0.50 and the p-values are less than 0.001. These findings confirm that the Second-order structural model successfully meet the criteria for reliability and validity testing.

The model’s uniqueness was affirmed through a discriminant validity test, an essential criterion for verifying reflective structural models as noted by Hair et al. (2017). Latent variable correlation (Fornell-Larker, F-L) is another way to test the discriminant validity, where the value of the square root of AVE should be greater than the correlation coefficient between each constructs (Hair et al., 2017, 2019, 2022). In other words, the explanation of observed variables by latent variables should be better than the explanation of other latent variables in order to pass the discriminant validity test.

The findings, presented in Table 6, show that HTMT value of Gratification <− > Behavioral, Reflective <− > Behavioral, and Reflective <− > Gratification are 0.937, 0.955, and 0.904, respectively, which are all greater than 0.85, indicating that internal correlation exist among constructs of Behavioral, Gratification and Reflective. The HTMT values of the other constructs are less than 0.85, indicating that certain internal correlation exist among some constructs in the model. Therefore, this study is further tested by F-L. It is also found that the value of the square root of AVE between each construct is greater than the correlation coefficient between each construct, which passed differential validity. It indicates differential validity is supported, displaying good internal consistency between each constructs in the model of this study.

In the structural model and hypothesis testing, this study utilized the Partial Least Squares (PLS) automatic estimation method for measuring structural models and verifying hypotheses. The final verification model is depicted in Figure 3, and the total effects are detailed in Table 7. The t-values for H1, H2, and H3 were 3.391, 2.415, and 3.541, respectively, with corresponding p-values of <0.01, < 0.05, and < 0.001, respectively, and positive path coefficients. These results support H1, H2, and H3, indicating that visceral, behavioral, and reflective aspects positively influence over-engagement in MOBA games by gamers.

The t-values for H4a is 1.876, with p-value of 0.061, greater than a 0.05 threshold. This leads to the rejection of H4a, suggesting that leisure does not significantly influence gamers’ visceral perception in MOBA games. However, H4b and H4c, with t-values of 8.471 and 6.569 respectively, and p-values were less than 0.05 and 0.001 respectively, are supported, indicating a significant influence of leisure on gamers’ behavioral and reflective states. Additionally, the indirect effect of leisure on MOBA game engagement, with a t-value of 7.735 and a p-value less than 0.001, suggests a significant overall impact of leisure on game engagement.

For H5a-c, the t-values were 6.569, 12.747, and 7.928, respectively, with all p-values <0.001 and positive path coefficients, validating these hypotheses. These findings indicate that gratification significantly influences gamers’ visceral perception, behavior, and reflection. Moreover, the t-value for the indirect effect of gratification on MOBA game engagement was 4.364, with a p-value <0.001, demonstrating a substantial overall impact of gratification on engagement in MOBA games.

PLS-SEM model quality assessment includes the coefficient of determination (R²), which is used to evaluate the explanatory and predictive power of the model. The value of R² ranges from 0 to 1, with higher values indicating greater explanatory power. The criteria for assessing the R² value are inconsistent across different disciplines and research areas (Hair et al., 2022).

The R² of the theoretical model of MOBA addiction was assessed by SmartPLS 4.1.0 based on Bootstrapping algorithm. The results are shown in Table 8. The effects of Leisure and Gratification on Visceral (T = 4.258, p < 0.001), Behavioral (T = 9.133, p < 0.001), and Reflective (T = 11.550, p < 0.001) constructs were 26.8, 47.5, and 55.9%, respectively. The explanatory power of Visceral, Behavioral, and Reflective for Games Addiction was 32.0%. It shows that the theoretical model of this study has good quality.