We employed a serious game called Negotiation Battle (Figure 1), which is a board game developed by a nonprofit organization called Citizen’s Science Initiative Japan [26]. We chose Negotiation Battle because it is a board game in which 2 teams with different views on lifestyle-related diseases can discuss the issues while playing the game, and we thought it would have an educational effect through discussion.

Negotiation Battle is played by 6 people on 2 sides, with one side playing the seducer (3 people) and one side playing the human (3 people). The seducer team persuades the human team, but the seducer team also exchange opinions with each other and with the human team. The duration of the game is 20 to 30 minutes per set, and the game set includes a dilemma card (Figure 2) and health sheet (Figure 3). For the dilemma situation, unhealthy points and happy points are listed on the card about each specific behavior. On the health sheet (one for each seducer role and only the seducer team can record on and view the sheet), unhealthy values of the human team are recorded. When the health points reach a certain value, the seducer is hospitalized.

The goal of the seducer is to add up the unhealthy points of the person he is in charge of, while the goal of the human is to accumulate happy points without being hospitalized. (1) The seducer reads out the card’s dilemma situation and seduces the human to focus on work, hobbies, and relationships and continue with behaviors that are unhealthy. (2) The human decides whether or not to accept the temptation while negotiating and interacting with the seducer and other human players. (3) If the human accepts the temptation, the seducer adds the value of the unhealthy points on the dilemma card to the health sheet at hand. (4) The human earns happy points. (5) After each human's decision is made on one card, the player moves on to the next card and repeats steps 1 through 4. The health sheet shows the threshold of ill health (ie, hospitalization, which is game over for the human), but the human cannot know the current value of ill-health points. (6) At the end of the game, the seducer discloses the information on the health sheet to the human, and (7) together they reflect on and discuss the types of temptations to which they were vulnerable. Thus, participants are stimulated to engage in interactive communication by applying the imaginary dilemma to real-life situations.

We recruited participants from 3 age groups—adults (including postgraduate students), undergraduate students, and high school students—because behaviors and attitudes toward lifestyle-related diseases could differ between adults and young people. In addition, among young people, we expected that there would be differences between high school students, who mainly live with their parents, and university students, who are often independent from their parents (living alone).

Adults, who were invited to participate and applied through a social networking service (Facebook, mailing list), attended a total of 3 Negotiation Battle sessions held between November and December 2016. For university students, we invited second-year students of the Tokyo University to participate in January 2017, and for high school students, we invited third-year students of a high school in Tokyo to participate as part of their classes in January 2017.

The participants were asked to submit a questionnaire (Gaming Event Assessment Form for Lifestyle-related Diseases) before, immediately after, and 2 to 4 weeks after the intervention. Prior to the intervention, an instruction document that described the survey was given to each participant, with the questionnaire and a return envelope to be filled in 2 to 4 weeks after the intervention and returned by mail.

There is a rule of thumb regarding sample size in structural equation modeling analysis that the sample size should be at least 5 times the number of parameters [27]. In this study, the questionnaire had 16 items (on the theory of planned behavior), and the minimum required sample size was estimated to be 80. However, since this study was conducted in the context of actual classes for university students and high school students, feasibility was given priority, and it was considered inevitable that there would be some groups below that size in each age group.

The questionnaire was constructed to assess components of the theory of planned behavior (Figure 4) to allow structural equation modeling of before-and-after comparisons, and additional items were inserted to obtain background information on the participants (age, family composition, occupation, and history of lifestyle-related diseases as a background profile of the participants). The questionnaire was scaled according to previous studies that examined 4 aspects, namely, attitude toward behavior [28], subjective norm [29], perceived behavioral control [30], and behavioral intention [31]. The questionnaire included additional items for free description of what the participants learned or whether they newly started healthy behaviors afterward: “What was your learning or awareness about healthy lifestyle?” (for both time points) and “What kind of healthy behaviors have you recently started?” (2 to 4 weeks after).

Responses were analyzed using SPSS Statistics and Amos software (version 23; IBM Corp). Structural equation modeling was performed to determine the relationship (causality or correlation) between factors of the theory of planned behavior. We used the comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA) as fitness indices. Missing values were imputed (using means within each item). Statistical significance was set to P<.05. Content analysis was used for the free descriptions; characteristic concepts were extracted from entire free descriptions, and the frequency with which these concepts were observed was counted for each category.

Participants were given a written and verbal explanation of the study protocol, and only those who consented were included in the study. Anonymity was ensured; the contents of the questionnaire were viewed only by the researchers, and no identifiable information was disclosed. The researchers securely stored collected data. The undergraduate and high school students were assured that their responses would not place them at any academic disadvantage. Ethical procedures [32] of the Ministry of Education, Culture, Sports, Science, and Technology and Ministry of Health, Labor, and Welfare in Japan were followed; formal ethical approval is not mandated for this type of study under these guidelines.

The adult group ranged in age from 23 to 67 years, university students ranged in age from 19 to 34 years, and high school students ranged in age from 17 to 18 years (Table 1).

In adult participants, structural equation models demonstrated that there was a significant causal relationship between perceived behavioral control and behavioral intention both before (path coefficient 1.030, P<.001; CFI 0.562; TLI 0.464; RMSEA 0.259) and immediately after (path coefficient 2.045, P=.01; CFI 0.755, TLI 0.700, RMSEA 0.223) the intervention (Figure 5).

In undergraduate students, before the intervention, significant causal relationships between perceived behavioral control and behavioral intention (P=.004; CFI 0.781; TLI 0.731; RMSEA 0.138) and between attitude toward behavior and behavioral intention (P=.04) were evident. In contrast, however, the relationship between attitude toward behavior and behavioral intention was no longer significant immediately after the intervention (P=.22; CFI 0.785; TLI 0.701; RMSEA 0.140), which suggests that perceived behavioral control alone influences behavioral intention (Figure 6).

In high school students, prior to the intervention, no factors significantly influenced behavioral intention (CFI 0.785, P=.97; TLI 0.701, P=.97; RMSEA 0.154, P=.97); however, a significant causal relationship (P<.001; CFI 0.709; TLI 0.596; RMSEA 0.210) was observed between perceived behavioral control and behavioral intention immediately after the intervention (Figure 7).

The number of valid responses for free descriptions immediately after the intervention were 20, 66, and 19 for adults, undergraduate students, and high school students, respectively; the number of valid responses for free descriptions 2 to 4 weeks after the interventions were 12, 54, and 16 for adults, undergraduate students, and high school students, respectively. A total of 8 concepts were observed (Table 2): dilemma, intention, learning, and status quo explanation with dilemma and learning description as dominant descriptions.

Responses 2 to 4 weeks after the intervention contained descriptions of behavior (adults: 10/17 concepts, 59%; undergraduates: 42/74 concepts, 57%; high school students: 14/18 concepts, 78%) with less descriptions of dilemma (adults: 3/17 concepts, 18%; undergraduates: 3/74 concepts, 4%; high school students: 2/18 concepts, 11%) or other concepts (Table 3).

Our findings suggest that perceived behavioral control is a determinant of behavioral intention in adult participants who played Negotiation Battle, a game in which players face an imagined situation that induces dilemma and engage in dialog. Free descriptions revealed that adults frequently experienced dilemma (13/44, 30%) and learning (14/44, 32%), which were the expected characteristics of the game. Thus, it seems that exposure to different perspectives during dilemmas in the simulated scenarios and the subsequent dialog led to self-reflection and transformative learning, which reinforced their perceptions that they can control health-related behaviors.

Transformative learning pertains to “a deep, structural shift in basic premises of thought, feelings, and actions [33].” According to transformative learning theory, critical self-reflection on the assumptions of learners facing disorienting dilemma may occur, which leads them to explore new options regarding roles, relationships, and actions. After undergoing such phases, they build competence and self-confidence in new roles and relationships [34]. In the game, players faced imaginary dilemma situations, which may lead to critical self-reflection and cognitive change in their perceptions of their health behaviors.

For undergraduate students, 2 factors—namely, attitude toward behavior (path coefficient 0.241; P=.04) and perceived behavioral control (path coefficient 0.568; P=.004)—influenced behavioral intention prior to the intervention. Immediately after the intervention, the influence of perceived behavioral control on behavioral intention was maintained (path coefficient 0.737; P=.001), whereas that of attitude toward behavior was not (path coefficient 0.155; P=.22). This finding indicates that undergraduate students also reinforced perceived behavioral control toward healthy behavior by facing dilemmas and undergoing transformative learning.

For high school students, no significant factors for behavioral intention were observed prior to the intervention (P=.97). Afterward, perceived behavioral control contributed to behavioral intention (P<.001). Descriptions of dilemma (10/30, 33%) and learning (14/30, 47%) were mainly observed immediately after the intervention, whereas those of behavior (14/18, 78%) and cognitive change (2/18, 11.1%) appeared after 2 to 4 weeks, suggesting that playing Negotiation Battle triggered transformative learning in high school students as it did in adults and undergraduate students.

Notably, the serious game related to lifestyle-related diseases resulted in transformative learning even in high school students, with the majority of concepts (14/18, 78%) suggesting that participants started a new behavior after 2 to 4 weeks. All high school students who participated in the study lived with and were dependent on their parents; thus, we assumed that they would pay less attention to the context of their diet or health behaviors. However, our findings revealed that the game can increase health awareness even among high school students. This finding is in line with those of previous studies [35-37], which demonstrated that serious games designed for health behavior change can be effective for adolescents.

Moreover, we observed that Negotiation Battle reinforced perceived behavior control out of all factors of theory of planned behavior, which led to behavioral intention change. The findings of previous studies [38,39], that digital serious games for health promotion among older adults enhanced perceived behavioral control, support this.

The study has 3 major strengths. First, we used the theory of planned behavior framework to reveal which factors led to change of behavioral intention after playing Negotiation Battle, which revealed that perceived behavioral control was a major influencing factor. Second, we found that Negotiation Battle can induce critical reflection and transformative learning by placing learners in simulated dilemmas. If transformative learning can be triggered, then the health consciousness transformed by learning will be likely sustained. Third, we compared the effects of Negotiation Battle on adults and on younger people (high school students); thus, the findings are observable across ages.

The study’s limitations are the relatively small sample size, which limits the generalizability of this study, and the lack of a control group in the study’s design to see the true effect of the intervention.

Through the simulation of dilemma and dialog in Negotiation Battle, participants were encouraged to reflect on their health behaviors, and enhanced perceived behavioral control contributed to the change in health consciousness. Serious game interventions based on the framework of cognitive change processes appear to foster self-reflection and dialog, which encourages transformative learning and the improvement of specific lifestyle behaviors.