Games With Safety Training Purposes in the Industry: Game Design Method and Its Demonstration

Work accidents are a concern for the industry as they can generate human, material, and economic losses. One way to mitigate them is to conduct efficient safety training, but they cannot always arouse the necessary interest and engagement of those trained. Games with a purpose are tools widely used for training in different contexts. They are identified as innovative, immersive, and attractive approaches, but they are complex to develop. This article presents SafetyPlay Game Design (SpGD), a structured method for designing safety training games with well-defined steps that support translating training elements into an immersive and playful game environment. Since we seek to build an artifact as an intervention to a problem in a real context, we based the research methodology on Action Research to conduct the investigations. Therefore, the intervention in the context of security training continued with creating and demonstrating the SpGD method based on the construction of a digital game for training. In this way, the evaluation undertaken in this research considered two moments: i) the ability of SpGD to develop the game, considering when we evaluated it by volunteers regarding game experience and perception of safety training, and ii) the validated method through interviews with game designers, experts in designing training games. The results point to evidence that the SpGD method supported the design of a game that allowed a positive gaming experience considered by game designers to be viable and useful for this purpose, in addition to allowing people to learn training concepts and risks. Therefore, this research contributes to the game design field and the industry, providing opportunities for creating interactive training that can complement traditional security training.


Introduction
Year after year, work accidents claim the lives of thousands of people (TRT04, 2022).According to the United Nations (ONU, 2022), between 2012 and 2021, more than 20 thousand workers lost their lives as a result of an accident involving their work activity (TRT12, 2023).
An efficient way to provide workers with the skills and knowledge necessary to develop work activities safely is by implementing safety training (Volpe andLorusso, 2009; Venturi et al., 2021).In this context, safety training incapable of transmitting insight into the dangers present in the work environment to workers may incorporate risks to a given organization, such as serious injuries, permanent disabilities, and even loss of life, as well as possible damage to property and/or the environment (Correa and Cardoso Junior, 2007).
However, safety training professionals (such as: managers and trainers) report low participant interest and engagement.They attribute this problem to factors such as the passivity of the training (e.g.slides, videos, handouts), which fail to arouse a sense of belonging in the people being trained (Rufino Júnior et al., 2023).As a result, there is a demand for more interesting and interactive safety training sessions (Gallerati et al., 2017).
In this context, purposeful games emerge as an interesting approach to meet this demand (Rufino Júnior et al., 2022).
They can present a dynamic training environment using a simulated risk scenario without the risk being present (Mayer et al., 2013), providing learning and engagement of those trained (Bruzzone et al., 2013; Chittaro andRanon, 2009).
Nevertheless, even though purposeful games can benefit safety training sessions, designing them is not a simple task (Forbes, 2022; Wolf et al., 2022).Even though industry managers are interested in such approaches, there is concern about the predictability of investing time and money to design them (Rufino Júnior et al., 2022).Therefore, creating a methodology capable of making the game development process for security training more systematic and simple, with less risk of mistakes in understanding and transmitting training information, is necessary.
With this in mind, this paper aims to present the proposal for the SafetyPlay Game Design (SpGD) method and its demonstration, explicitly conceived for designing digital games for safety training in the industry.SpGD focuses on adding agility and traceability to developing games for safety training.We based it on consolidated concepts and techniques relating to training evaluation and the development of digital games to align the essential characteristics of safety training with game elements.
Since the artifacts of this research (SpGD method and training games) seek to solve a problem in an organizational context, changing the work environment, the research methodology chosen was Action Research (AR) (Davison et al., 2004).Thus, given the diagnosis of the need for safety training games, the SpGD method is designed to intervene in the organizational context, with both the training games designed using the technique and the SpGD itself being evaluated and its results leading to reflection on changes in day-to-day organizational life.
A digital game was developed for safety training using the steps provided in SpGD to demonstrate the feasibility of using the method.
We evaluated the game in a quasi-experimental study to verify whether its design, with support from SpGD, provided players with a gaming experience and perception of safety training.Finally, we evaluated the SpGD through interviews with game designers to elucidate its strongness, possible improvements, and feasibility.The interviews followed the Underlying Discourse Explanation Method (Nicolaci-da Costa, 2007), and we analyzed the responses based on qualitative discourse analysis techniques grounded on the work of Strauss and Corbin (1990).
We organized the article as follows: Section 2 presents the backgrounds and related works.Section 3 introduces Action Research as the research methodology used in this article.Section 4 shows the SpGD method, and next, Section 5 demonstrates it from the design of a digital game for training purposes.
Section 6 presents the game evaluation and the validation of SpGD with game design specialists.We briefly reflected on our results in Section 7. Finally, in Section 8, we presented our final remarks.

Backgrounds
This section introduces the conceptual basis of our research, aiming to better understand the context, concepts, and approaches we used to design the research.We also present some related works that are directly associated with our study.

Safety Training in the Industry
Industrial activities refer to the various operations and processes performed in the industrial sector to produce, transform, and manufacture goods.
These activities encompass a wide range of sectors, such as manufacturing, energy production, construction, mining, chemistry, food, and beverages, among others, which present a series of risks and challenges to the professionals involved, such as exposure to dangerous substances, use of complex equipment and operations with high physical demand (Rodrigues and Santana, 2010).
In this context, it is essential to implement effective training programs that enable employees to deal with risky situations and adopt appropriate preventive measures (Martins, 2021).One of the main objectives of safety training in the industry is to provide workers with the necessary knowledge to identify and assess the risks in the work environment and the skills to implement best safety practices (Venturi et al., 2021).
With the advancement of technology, new training approaches have emerged in the industry.One of these innovative approaches is using digital games as a training tool (Vigoroso et al., 2021).Digital games offer an immersive and interactive experience, allowing realistic simulations of risky situations and the practice of specific skills.This innovative approach has effectively engaged participants, facilitating learning and retention of security-related knowledge (Bruzzone et al., 2013; Lu et al., 2022).

Training Assessment
Training assessment is an essential step in measuring the effectiveness and impact of training activities on the development of skills and performance of individuals (Lacerda and Abbad, 2003).
Several methods were proposed to evaluate training, each with specific approaches and focus.One of the widely recognized methods is that proposed by Baldwin and Ford (1988), which emphasizes results-based evaluation, seeking to identify and measure observable changes in the behavior and performance of individuals after training.Another relevant method is the one developed by Kraiger et al. (1993), which focuses on evaluating reactions, learning, transfer, and training results.
However, the Kirkpatrick assessment model is one of the most well-known and widely used methods (Kirkpatrick and Kirkpatrick, 2016).This model establishes four levels of training evaluation (Figure 1): reaction, learning, behavior, and results.Although each method has its advantages and specific focus, the Kirkpatrick model is one of the most robust for training evaluation.It covers different assessment levels and allows a complete analysis of the effects of training in multiple dimensions (Agarwal et al., 2019).

MDA Framework
There are many different frameworks and ways to design digital games (Winn, 2009; Rogers, 2010; Schell, 2019; Classe et al., 2019).Still, for this research, we choose to adopt the MDA framework (Mechanics, Dynamics, and Aesthetics) because it provides a simple way to analyze games, divided into clear elements.
The MDA framework (Mechanics, Dynamics, and Aesthetics) is a conceptual model used in game design to describe and analyze the fundamental elements that make up a game (Hunicke et al., 2004).MDA divides the game elements into three layers: Mechanics, Dynamics, and Aesthetics.
The Mechanics layer refers to the rules, systems, and components that structure the game.It covers the most tangible elements, such as movement, environmental interaction, combat, puzzles, and other game mechanics.This layer describes the game's basic features and how players interact with them.
The Dynamic layer deals with interactions between players and the game system.It describes how mechanics and systems interact and combine to create specific game experiences.Dynamics encompasses how players' choices and actions affect the state of the game, generating challenges, rewards, and consequences.In this layer, the players' complex behaviors and strategies emerge.
Finally, the Aesthetic layer refers to the player's emotional and subjective experience when playing the game.It involves elements such as narrative, atmosphere, visual and sound aesthetics, themes, and sensations transmitted by the game.This layer creates an emotional connection between the player and the game, providing pleasure, immersion, and engagement.
The MDA framework enables a structured approach to game design, helping developers understand and balance the elements that compose the gaming experience.By using MDA, designers can describe and analyze a game's mechanical, dynamic, and aesthetic aspects in a separate and integrated way, which can lead to a deeper understanding of the desired gaming experience and assist in making more informed design decisions (Hunicke et al., 2004).

Related Works
In previous studies, we performed some literature reviews to understand how the industry has used games for training.Our investigations led us to perceived use contexts, technologies, and concepts that allowed us to propose our method.
For instance, Rufino Júnior et al.In another research, Rufino Júnior et al. (2023) looked into the game and information systems literature to explore information about the game design process, elements, and coasts, considering the context of safety training.In that work, the authors identified types of "games for training" based on augmented reality, virtual reality, and 3D environments.
At last, Serra and Classe (2023) performed a systematic mapping of the literature to explore how the industry has been using gameplay data originating from games for training.They identified some uses of those data, such as game analytics, data visualization, and game balance.All those applications were responsible for giving training information to the organizational manager.
The search for methods that can support the creation of games for organizational training found research byClasse et al. (2019), which proposes the Play Your Process (PYP) method for designing games based on business processes and the research by Lopes et al. (2022), which presents PYP4Training, intended to guide the design of games for business process training.
Based on related work, it is possible to observe that although some studies investigate the use of games as training tools and there are some methods to design them, it is not possible to identify specific techniques aimed at safety training or risk mitigation in the industry.Therefore, evidence points to an opportunity for research into developing specific methods that meet the demands of safety training.

Action Research
Action Research is a flexible method suitable for scenarios where you want to obtain evidence about using an artifact applied in practice under a real problem (Santos and Travassos, 2011).
Such an artifact is then planned, developed, and used as an intervention to the problem.By assessing data, the results lead to reflection on the issue, potentially giving rise to a new investigative cycle (Figure 3) (Davison et al., 2004).As an investigation plan, we conducted a literature study to identify how organizations have used purposeful games in training contexts and what elements are necessary for this, correlating them with training evaluation proposals (Rufino Júnior et al., 2023).Still, during planning we proposed the SpGD method to systematize these games' designs.
As intervention/action, we used the SpGD method to construct a digital safety training game to solve the research problem.In this way, we carried out the evaluation in two moments: i) the game, which was used to demonstrate the feasibility of the SpGD method, in addition to the analysis of experience, learning, and identification of security risks by the players/trained, and ii) the method itself, submitted to evaluation by game designers through the exemplification of its application in the creation of the game, collecting perceptions of the feasibility of use, strongness and weaknesses on the part of game designers.
Finally, the assessment results led to reflection on the research artifact about solving the problem.Our research presented two reflection questions in line with the evaluation moments: 1) Did the players/trained positively perceive the game experience and safety training (learning and risks) through the gameplay?2) Was the SpGD viable to support the game's design for safety training purposes?
According to Carr (2007), AR is a process that involves action-reflection cycles in which researchers and research participants collaborate to identify a problem, plan and implement actions to resolve it, evaluate the results, and reflect on learned lessons.These repeated cycles allow for the adaptation and continuous improvement of approaches and interventions over time, promoting the progressive enhancement of practices and the production of knowledge relevant to solving real problems.

Plan: SafetyPlay Game Design Method
This research proposes the Safetyplay Game Design method to improve safety training in the industry by making it more playful, engaging, interactive, and motivating.We designed the SpGD method to guide game development teams for security training games.To do so, its stages encompass assimilating the fundamental concepts of training, translating them into game elements, going through validation and assessment cycles, and finalizing the game's production and implementation as a training alternative in industrial environments.
The SpGD method has well-defined and structured stages (Figure 4), which aim to guarantee the integration of safety training characteristics with game design elements, namely: The second stage (Mapping the training to game elements), also conducted by game designers, process managers, and trained personnel, corresponds to the translation of training information into game elements.We did this through a semantic alignment between Kirkpatrick's four training assessment levels and the MDA framework elements (Figure 6).This step aims to mitigate problems in understanding the aspects of safety training, which could lead to players' misunderstanding during gameplay.
It is important to highlight that it is possible to associate any elements of Kirkpatrick's model with those of MDA (red dashed lines).However, when semantically analyzing the meaning of the elements (considering Bloom's taxonomy (Ferraz and Belhot, 2010), procedural rhetorics (Bogost, 2008) and other theories), some relationships are more strongly indicated than others, for example, Mechanics are suggested based on information coming from Kirkpatrick's second level (learning), which contains the main actions, rules and elements that those trained must learn.
Furthermore, learning is strongly related to mechanics when thinking about procedural rhetorics.Dynamics benefit from information collected at Kirkpatrick's third level (behavior).At this level, we analyzed situations where the trained can use acquired knowledge.Finally, the game's Aesthetics are fed by references from Kirkpatrick's levels 1 (reaction), 3 (behavior), and 4 (results), benefiting from the feelings intended for those trained, the aspects of the workplace in which they can apply the knowledge and business indices that, transported into a game, could be affected by the actions of the trained.
In the third stage, Performing a brainstorming, game designers and process managers are trained to share ideas and thoughts about how the game should be designed based on the mapping from the previous stage.The objective of this stage is to create a documentary view of the game project (Game Design Document -GDD), which will assist developers and artists in the production of the game, with a detailed description of the game's levels, history, mechanics, rules, sounds and interfaces.
In the fourth stage, (Developing the game), the game is developed based on the GDD established in the previous stage.After the game's development, the fifth stage (Validating the game) occurs, where the game is validated by process managers and trained using information from the Kirkpatrick assessment model defined in stage 1.Such validation is important because it verifies whether the game is aligned with training objectives and conveys correct information to players.If it is not validated, we can adopt  three decisions: 1) the training context was poorly understood, making it necessary to return to the understanding training stage; ii) the mapping between training and game elements has gaps, so it is required to go back to the second stage of SpGD; or iii) there is an error in the game software so, you must return to the developing game stage.If approved, the game moves on to the sixth stage (Assessing the game) to evaluate the game experience, immersion, and engagement in the training process, besides using the training evaluation with Kirkpatrick's four levels to measure whether training was able to add the desired knowledge and experiences and ensure that those trained were aware of the risks of the presented situation.
Finally, the seventh stage is Delivering the game for use in safety training in the industry.At this stage, the game is made available to those who are trained and will benefit from the immersive and interactive learning experiences offered by the game.

Act: Game for Fire Training
We demonstrated The SpGD method to exemplify its application in creating a safety training game in the industry.In this demonstration, the "Bob Ruff in Deck is on Fire" 1 game was developed for fictional firefighting training, teaching about the different types of extinguishers used in these cases (Figure 5).
Firstly, those involved in designing the game (game designers and managers) realized the stage of understanding the proposed training, considering the levels in Kirkpatrick's assessment model.At each level of the model, we expected to answer some pre-determined questions.As this is a relatively simple training, Table 1 2 presents each level of Kirkpatrick, highlighting the questions and the responses of the managers of this training so that it is considered adequate in the organization.
At SpGD, that spreadsheet is developed jointly between managers and game designers so that the team can understand the context of what they will transform into a game from the beginning of the process.This approach aims to ensure a certain coherence between what organizations expect from training and which elements should be present during gameplay 3 .
In the second stage, Mapping the training to game elements, the responses obtained and organized based on Kirkpatrick's levels were aligned with the MDA framework elements (Figure 7).The level of reaction, where we expect the trained to be motivated and react promptly to a possible fire situation, is connected to the game's aesthetics through fire outbreaks (Figures 5A and 7A) which, if not extinguished in time, evolve into explosions (Figures 5E 1 Game: https://joccom.uniriotec.br/games/deckfire| INPI nº: BR512023001633-0 2 Letters in Figure 5, Table 1, and Figure 7 were prepositionally allocated to reference the game design traceability.It is possible to trace the answers we got in SpGD step 1, which we used to map game elements in step 2 and develop the game in step 3. 3 Meetings document: https://bit.ly/BobRuff_DesignMeet22-23 and 7E).Learning the types of fire and correct extinguishers and possible consequences are linked to the mechanics of obtaining information about them (Figures 5A and 7A) and choosing the appropriate extinguisher (Figures 5B and 7B) for success in the game.Situations where the worker can demonstrate a behavior change, that is, fire situations in the workplace encourage dynamics (identifying information on the type of fire or choosing the best extinguisher for that situation) (Figures 5 C and 7C) and aesthetics (fire spots in places inspired by the work environment) (Figures 7A and E).Finally, the result, a company metric (fire completion rate), is transported to the game's scorecard (Figures 5C-D and 7C-D).
In the third stage, Performing brainstorming, the game designers and managers fed a game design document with ideas inspired by the alignment information between the training elements and the MDA, organizing the ideas around the level design (the game occurs on the deck of an oil tanker), storytelling (the player takes on the role of a member of the company's fire brigade) and the already established mechanics, dynamics and aesthetics.
With this document, a technical team (e.g.artists, programmers) developed the game.It is possible to see that the planned mechanics (obtaining information about fire and suitable extinguisher) in Figures 5A e 5B. Figure 5C represents the dynamics intended for correct or incorrect firefighting.The aesthetics (scorecard, fire source, and explosion) can be seen in Figures 5A, 3D e 3E.Therefore, we affirm the existence of an alignment between the main elements of safety training and the game's elements.
When identifying a fire source, the player can move close to it, and when hovering the mouse over the fire animation, information about which extinguisher is suitable is displayed (Figure 5B).Having identified the type of fire, the player must select the correct extinguisher, position it on the "Equip" field, and click on the flame.If the extinguisher is suitable, it will extinguish the fire.If not, the flame increases and could lead to an explosion.The player also has information such as the monetary value of the damage caused by the fires, time, and health bar (Figure 5C)."We expected the trained person to be calm even in a stressful situation."(C) "Safety training participants must reflect on this complex situation and its risks."(C) "They must choose the better option (extinguisher) to extinguish the fire outbreak."(B)

-Learning
What is the main safety information that they must learn?"Type of fire extinguishers and what kind of fire they must be used.Thus, this aimed to demonstrate the viability of the SpGD method, considering the coherence of the steps that lead to the creation of the training game conducted by game designers, managers, and trained.We conducted the validation steps with the managers who supported this game's initial stages.If they considered the result valid, we evaluated the game with the target audience: industry professionals who require fire training.Therefore, we present the evaluation with training participants in the evaluation section of this research article.

Evaluate
We defined two assessment steps to guide this evaluation: a quasi-experimental evaluation using the game for fire training and; semi-structured interviews with the support of the game designer.
At least, we planned a reflection/communication step providing a synthesized discussion of the results of two evaluations.A summary of this process can be seen in Figure 8.

Game for Training Purpose Evaluation
At this stage of the research, we based the study design on an empirical evaluation along the lines of a quasi-experimental study (Campbell and Stanley, 2015), following the steps: 1) definition; 2) planning; 3) execution; 4) analysis and; 5) results and conclusions.According to these authors, quasi-experiments are less controlled empirical studies than traditional experiments, with no need for random selection of participants.Hence, we selected this research method because the survey's audience comprises volunteer workers from an industry.

Definition
The study definition presents the research objectives within a specific context.Then, we decided to follow the GQM (Goal-Question-Metric) (Basili, 1992)  The assessment organized by GQM seeks to answer questions which, in this research, can be stated as: (Q1) Did the participants have a positive perception of the game experience?(Q2) Did the participants positively perceive the safety training in learning and identifying risks?
We based the metrics used to answer the questions on the dimensions of game experience and MEEGA+ security training (Petri et al., 2019), the latter dimension covering learning variables and identification of training risks.This questionnaire consists of statements in which participants indicate their degree of agreement with the statement using a Likert scale.This ordinal/interval scale allows descriptive, correlational, and inferential statistical analyses (Coelho et al., 2020).

Planning
When planning scientific evaluations, it is necessary to detail the context of the study so that it can be understood, its limitations mitigated, and academic peers can replicate it.It is essential to explain who the participants were, the study steps, what will be the instrumentation, and how the data were collected, processed, and analyzed.
The study participants were volunteer workers from a Brazilian industry who had already undergone fire training within their respective work environments at some point in their professional lives.
We conducted the study by collecting the opinions of these professionals and accordingly in the form of an opinion survey.No sensitive information identifying the worker or company was requested, collected, analyzed, or revealed.We presented all information about the conduct and objective of the study to participants in a free and informed consent form (FICF), and their participation was recorded in the participation questionnaire.
It is worth noting that, according to CNS (Brazilian National Health Council) resolution 510/2016 4 , in its Art.1, studies that are characterized as opinion research, where there is no identification of participants, do not need to be submitted and evaluated by a research ethics committee.Furthermore, all data collected follows the guidelines of Brazilian data protection law 5 .
Participants performed the research tasks at three different times.Initially, participants had access to the electronic form with the FICF and questions related to their profile (we defined a time limit of 10 minutes for this).They were then given access to the game and asked to play it for at least 15 minutes, considering it a safety training section.Finally, participants answered their perceptions based on statements from the adapted MEEGA+ questionnaire (with a maximum time of 15 minutes).
We based the instrumentation on an adaptation of MEEGA+.Thus, it consisted of 52 items: 8 questions related to the participant's profile and 44 statements related to the game and its relationship to safety training (Table 2): 29 about game experience and 15 about safety training.The 44 statements were composed using a Likert scale, ranging from -2 (totally disagree) to +2 (totally agree).Data collection 6 was undertaken using an electronic Google form.
The participants' responses were collected independently, although the classes allowed interaction and collaboration.
Rigorous empirical studies require identifying and addressing validity threats to the study.Data analysis used quantitative approaches, descriptive statistics (percentage, mean, median, mode, standard deviation), and inferential statistics (normality analysis, correlation, and hypothesis).The analysis software used was R Statistics 4.1.3.We defined the significance level for inferential statistics as 95% (alpha=0.05).We used tables and figures to summarize the quantitative analyses and the results.

Execution and Analysis
The execution of the evaluation occurred online from September 29, 2023, to October 13, 2023.It took place individually, in which 21 people voluntarily participated in the study, as stated in the questionnaire's free and informed consent form.None of them attended the study more than once.

Participant's Profile
Regarding the profile of the 21 participants, the predominantly observed characteristic was the majority were men (95%), aged between 40 and 49 years old (45%), with an educational degree (41%), all workers in the oil and gas sector and who have already participated in a fire safety training session, and who rarely use games (45%).
Although the majority of participants in this study are men, they are not unique participants in the training process.Many women have been working in training contexts and hazardous situations.They will be the focus of future investigations related to this research context.

Reliability and Validity Questionnaire
To validate the reliability of the items in the MEEGA+ questionnaire, we used a statistical method, which, although the literature indicates limitations (Yang and Green, 2011), is still widely used for this purpose: the analysis of Cronbach's alpha coefficient.When interpreting the alpha value, one must observe the values (Cronbach, 1951) and interpret them as >0.9 excellent; >0.8 good; >0.7 acceptable; >0.6 questionable; >0.5 poor; and <= 0.5 unacceptable.According to Freitas and Rodrigues (2005), in exploratory studies, Cronbach's alpha values between 0.6 and 0.7 can be considered acceptable, and this argument is used in the reliability analysis presented in this paper.
Table 4 shows the Cronbach's alpha coefficient results applied to the MEEGA+ items.As can be seen, considering the items and their scales as a single questionnaire, the general alpha (Benreal Reliability) obtained a value of 0.91, which indicates excellent reliability in the participants' responses.
Individually, the categories of gaming experience and training perception (learning and risks) were obtained, respectively, with alpha values of 0.88 (good) and 0.73 (acceptable), which corroborates the reliability and general validity of the responses.
When observing some of the variables individually, it is clear that some values demonstrate unacceptable reliability, such as aesthetics (alpha = 0.40) and trust (alpha = 0.32).Regarding aesthetics, we understand that the prototype game may not have a very attractive design or the best fonts and color palettes.Therefore, we consider that this result may have influenced this value.Concerning the trust variable, we noticed that, at first, there is a lot of information for the player, which may have influenced an initial perception of a lack of trust and organization of the content perceived by the players.Consequently, both dimensions need attention in the future design process of these types of games.
However, we considered the participants' responses reliable and valid data for analyses, reflections, and considerations on the research questions defined for the study.

Participant's Perceptions and Category Correlations
When observing the percentage of responses (percentage column) and, mainly, the mode statistics in Table 4, it is observed that many of the items have values above 0 (some level of agreement with the statement -with the only exception of ACE01).Such a previous conclusion is easy to see when analyzing Figure 9.
Still, mode statistics observations cannot affirm positive answers to questions Q1 and Q2.
Confirming the participants' perception, we made statistical inferences considering questions Q1 and Q2 as alternative hypotheses and, as neutral hypotheses, whether participants had negative or neutral perceptions related to them.

EST01
The game's design is attractive.

EST02
The texts, colors and fonts match and are consistent.

Learnability (APR)
APR01 I needed only a few things to start playing.

APR02
Learning to play this game was easy for me.APR03 I think most people would learn to play this game pretty quickly.

Operability (OPE)
OPE01 I consider the game to be easy to play.OPE02 The rules are clear and understandable.

ACE01
The fonts (size and style) used in the game are legible.

ACE02
The colors used in the game are understandable.

PTE01
The game protects me from making mistakes.

PTE02
When I make a mistake, it is easy to recover quickly.

Trust (CONF)
CONF01 When I first looked at the game, I had the impression that it would be easy.CONF02 The content's organization helped me be confident that I would learn from this game

DES01
This game is suitably challenging for me.

DES02
The game offers new challenges (offers new obstacles, situations or variations) at a good pace.

DES03
The game does not become monotonous in its tasks (repetitive or tedious tasks).

SAT01
Completing the game's tasks gave me a sense of accomplishment.

SAT02
It is due to my effort that I can advance in the game.SAT03 I feel satisfied with the things I learned in the game.SAT04 I would recommend this game to my colleagues Fun (DIV)

DIV01
I had fun with the game.

DIV02
Something that happened during the game (game elements, competition, etc.) made me smile.Focused Attention (AF)

AF01
Something interesting at the start of the game captured my attention.

AF02
I was so involved in the game that I lost track of time.AF03 I forgot about my surroundings while playing this game.

REL01
The game content is relevant.

REL02
It is clear how the game's content relates to real situations.

REL03
Play is a suitable teaching method for learning about real-world situations.REL04 I would rather learn with this game than another way (another method).

APC01
The game contributed to my learning about prevarication crime.

ACP02
The game was efficient for my learning compared to other information and news (newspapers, social media, etc.).

Risks (RISC)
RISC01 The game presented the number of fires that occurred and the total number of fires extinguished, which serves as a firefighting metric in the company.RISC02 The game showed that if I choose the wrong extinguisher I could cause malfunctions, explosions or other risks to the work environment.RISC03 The game showed a financial decrease, suggesting that poorly fought fires bring losses to the company.RISC04 The game showed that the longer fires took to be fought, the more outbreaks could emerge.RISC05 The game showed that if there is a delay in fighting fires, they could explode, even leading to the total compromise of the workstation.RISC06 The game shows that your health depends on you extinguishing the fire quickly but safely.

Table 3. Threats to validity
Type Threat Description Treatment

Conclusion
The statistical power of the analysis method It is related to using statistical methods that could reach wrong conclusions.
Scales and statistical methods more consistent with the metrics were selected and applied Violation of assumptions of statistical methods Wrong use of the statistical tests to data that could not be evaluated with them Use of statistical methods consistent with the scale and characteristics of the data sample.

Bias in data selection
Favoring of data by researcher The data used have been published so that others can repeat the analysis.

Lack of training It happens when participants do not know how to operate the study object
To lessen the threat, participants could answer at their homes.

Participant wear
This influences the participant behavior in studies that demand hard effort Assessments are designed to be completed in 50 minutes.

Research expectations
Influence of researchers above participants in a conscious way or not.
We planned the study so that the researcher did not communicate with participants, except in issues related to analysis objectives.

Instrumentation
It happens when the study instruments are not adequate.
We used game evaluation questionnaires that were already used in previous research.Additionally, we made a reliability verification as the first analysis to guarantee the reliability and validity of participants' answers.

Planning
Use of reproducible methods.
To lessen this threat, the assessment was planned considering the design definitions of quasi-experimental studies (Campbell and Stanley, 2015)

Generalization
It consists of generalizing the study results to a population bigger than the study participants.
We treated it using distinct participants' profiles.Each one was a worker in a Brazilian industry that already had safety training sessions in their jobs.Hence, the first test was normality to determine which hypothesis test would be most suitable for sampling.In this study, as there were fewer than 30 participants, the most recommended test was the Shapiro-Wilk test (Asadoorian and Kantarelis, 2005).In addition to the mean and standard deviation values for gaming experience and perception of safety training, Table 5 also shows the results of the normality and inference tests.About the normality test, neither category follows normal behavior.In other words, their p-values are less than 0.05.By analyzing the Wilcoxon hypothesis test values, it is possible to state that all the values are below 0.05.As a result, it is possible to say with at least 95% certainty that the alternative hypotheses can be accepted.Consequently, it is possible to answer affirmatively that in Q1, the participants had a positive gaming experience.In Q2, the participants had a positive perception regarding learning and identifying training risks in the game.
The correlation analysis between the categories is vital because it identifies how each relates to the other.For this, we executed correlation tests among them.Since we perceived that the category data does not follow normal behavior (Table 5), the most indicated correlation test is the Pearson test (Benesty et al., 2009).According to Gasparin et al. (2010), we can interpret the value resulting from Pearson's correlation analysis as >0.5 large correlation; >0.3 medium correlation; >0.1 small correlation and <0.1 no correlation.
This range of values can vary both negatively and positively.
Figure 10 presents the correlation measures between the game experience and safety training perceptions; we observe a strong correlation (0.63) between these dimensions.When detailing the security training perception category, it is still possible to notice strong correlations between gaming experience and risk identification (0.63), gaming and learning experience (0.53), and risk identification and learning (0.58).
Therefore, it is possible to conclude that there is evidence that participants perceived the BobRuff in Deck is on Fire game as positive in the study, with the game experience strongly correlated with learning from the training section and identification of risks of the context presented in the game.Consequently, once we developed the game with the support of SpGD, it is possible to say that there is evidence that the method can support the creation of digital games for training purposes.

Definition and Planning
We based the protocol that guided this part of the study on the Underlying Discourse Explanation Method (UDEM) (Nicolaci-da Costa, 2007) and included five stages: 1) selection of the sample of participants, 2) construction of the interview script, 3) conducting the interviews, 4) transcribing the statements and 5) analyzing and communicating the results.Initially, the sample was selected (step 1), defining the desired profile of the participants, that is, game designers with experience in creating purposeful games.Next, we prepared a structured interview guide (step 2) containing specific questions that served as a guide during the individual interviews.The objective of this stage was to identify the perceptions, opinions, and suggestions of game designers regarding the SpGD method's strength, use feasibility, and improvements.
Having selected the participants according to the profile defined in the sample selection stage and having the structured script in hand, we conducted the interviews individually (stage 3), allowing direct interaction between the interviewer and the participant.The UDEM method provides that semi-structured interviews must have a pre-defined duration.For this reason, each interview lasted a maximum of 40 minutes, where the SpGD method was presented, with a detailed explanation of each of its steps and the demonstration of the technique, as well as a demonstration of the game developed using the method.Subsequently, we transcribed the interviews (step 4) to be analyzed and described in the present study (step 5).
We conducted a qualitative analysis, using Grounded Theory (GT) procedures as a basis (Strauss and Corbin, 1990) to analyze the interviews, which consists of data analysis in three phases: open coding, axial coding, and selective coding.In open coding, researchers separate data, conceptualize, and categorize.In axial coding, researchers identify possible relationships between categories.Finally, in selective coding, researchers create the central idea of the study.
Yet, since the study included interviews and qualitative data analysis, we must consider threats to validity.According to Pinto and Santos (2012), researchers in this type of research can inadvertently transmit personal bias in their interpretations since a speech coding stage was carried out.A group of researchers performed the qualitative analysis and, when discussing their interpretations, reached a consensus on the results to mitigate this threat.
The sample size may also threaten the generalization of this study.However, the UDEM method indicates the quality of the sample by considering the saturation of the statements given in the interviews.Such saturation was observed in the interviews performed in this study.Even so, specific contextual limitations must be considered in future studies to ensure the validity of the results.

Execution and Analysis
Game designers with academic and professional experience designing purposeful games, identified in lists from gaming communities, were invited to conduct the research.Five participants responded with their availability to participate in the study.The interviews happened between the 6th and 9th of June 2023, with participants interviewed in individual online meetings lasting 40 minutes, where the statements were recorded and transcribed.The interview report can include the questions that guided the interviews and the quotes from the participants used for this analysis (Table 6).We conducted the data analysis procedure using qualitative questions about the SpGD method.Although the GT mentions three types of coding, according to Strauss and Corbin (1990), it is up to the researcher to define which of these codings will be appropriate for the study.Therefore, this research adopted open and axial coding.Since the central idea of the study is unique and well-defined, the researchers did not consider it necessary to apply selective coding.
We read the data obtained in the interviews to implement the coding with the help of the Atlas.tisoftware, highlighting the important excerpts for the research, and assigning codes according to the content (open coding).Subsequently, we made the axial coding, where we grouped the codes into categories, which are sets of codes related to each other at a high degree of abstraction.
The analysis of the data obtained from the interviews resulted in identifying 14 different codes for the main contributions of the statements, grouped into three categories: strongness, improvements, and feasibility of the method.In Figure 11, it is possible to view the relationships among the codes and their respective categories.similar to the Miro7 platform, to optimize the registration of ideas: The idea of [using] Canva would be interesting; it's a cool idea.Even to organize the ideas [in the brainstorming to make the GDD] ... Maybe use a structure similar to Miro's.[Participant #1] Participant #3 contributed to the game assessment model contained in the SpGD method.
According to this participant, it is important to define who will be the professional acting in the role of manager in the stages of the process, highlighting that a possible solution would be to change the term from manager to stakeholder (that is, an individual who has an interest in or influence on the project).

Viability
The indications regarding the method's viability are based on the statements where the participants respond that they would use SpGD if they needed to produce a game for safety training in the industry.
The simplicity of the method and the possibility of validation with managers, trainers, and the target audience were cited as points that favored its use, in addition to the well-defined steps and functions.It is worth highlighting the comment by Participant #2, who suggested that the method could be expanded to various professional training, not just those focused on safety.
Participant #2 comments that support the method's viability: "I believe it helps.The steps in the method are very specific.Each step is well described.I believe so [

Reflect
Based on the need to search for more modern and immersive alternatives to innovate in safety training sessions, this research presents the SpGD method as a proposed intervention in this problem context.The method aims to support the game design process for safety training, guaranteeing the traceability of training and game design elements throughout the competition process.For this, we built a training game prototype, which allowed reflections on the perception of the game experience and safety training and the possibility of interviewing game designers concerning the feasibility and positive and negative points of the method.

Did the players/trained positively perceive the game experience and safety training (learning and risks) through the gameplay?
Some authors argue that purposeful digital games can be useful for training sessions in risky situations.They allow the active involvement of trained in complex situations, immersing them in problem-solving as long as such games provide a good gaming experience (Kwegyir-Afful and Kantola, 2020).In this study, based on the game assessment by the trained, enabled us to perceive indications that the game experience contributed positively to immersion, attention, perception of the relevance of the training, and even to the fun and satisfaction of the trained (Table 4).These are indicative variables of the gameflow state (Berube, 2021), as participants' engagement and focus.
This immersion is ideal for learning during purposeful gameplay (Teichmann et al., 2020).
We evidenced immersion in this study by identifying a strong correlation between the game experience and training perceptions, involving learning specific content related to the safety training contexts and identifying the risks involved in the presented situations.These observed values align with research by Mayer et al. (2013).In their study, the game is stressed by simulating risk situations and their consequences, leading people involved in the training to an experience of learning and reflection.
Therefore, it is possible to affirmatively answer that there is identification that the trained had a positive gameplay experience with the digital game for safety training purposes and that the game provided learning and risk identification within its training context.Since it was developed based on the SpGD steps, the method supported the game design process by monitoring and tracking the training in the game elements.

Was the SpGD viable to support the game's design for safety training purposes?
Although the benefits of games for training purposes are known, building them is a challenging and often unfeasible task due to the complexity involved when trying to translate the context of security training into the elements of digital games (Forbes, 2022; Wolf et al., 2022; Rufino Júnior et al., 2022).When conducting interviews with game designers experienced in building purposeful games, they pointed out that a method for building games for safety training is welcome, as long as they have well-defined steps that help understand the training.
However, a significant concern when building these games is precisely presenting the training without information being misunderstood or interpreted (Rufino Júnior et al., 2023).It is one of the main reasons SpGD supported training assessment models and aligned them with models or frameworks for game elements, such as the Kirkpatric X MDA association.As pointed out by the interviewees in validating the method, this type of association allows not only support for game design but also to track training throughout the game creation process and, finally, validate the game results through the training evaluation model itself.
Moreover, some interviewees said that, although very useful in creating games for security training, attention is needed when carrying out Kirkpatric X MDA mapping.This suggestion, although it guarantees training traceability, can limit creativity and influence the composition of game mechanics.
In general, although there are suggestions for improvements to the method, the game designers considered it viable due to its simplicity, well-defined steps, and the involvement of managers, trainers, and game designers.Although existing, the risk of including poorly conveyed training information in the game is minimized.
In summary, there is evidence that the SpGD method is viable.It supports the design of games for safety training purposes, providing a good gaming experience that culminates in learning about training and identifying the risks involved in this context.Thus, it gives a viable option for contributing to safety training sections.

Final Remarks
This research aims to create a more systematic and simple game design for safety training in the industry.It presents the SafetyPlay Game Design method, which offers a structured and targeted approach to the context of risk situations in the industry.
We demonstrated the method by developing a game where fire safety training was translated into game elements, showing that designing this game using its proposed steps is possible.The game assessment was done by industry volunteers focusing on the perception of the game experience and safety training when learning concepts and identifying risks.After that, we assessed the method with game designers who attested to the technique's feasibility, all of whom highlighted that they would use it if necessary due to the simplicity and the presented theoretical-technical basis.
Additionally, the same professionals contributed by highlighting strongness and suggesting improvements for future method versions.
In this sense, when analyzing the results of the evaluations, it is understood that there is evidence that the SpGD method is viable to support the process of building games for safety training.The research objective of supporting the industry's design of safety training games was achieved.
However, we can point out some limitations of the study.We recognize that we conducted quantitative and qualitative assessments with few participants in a controlled setting.In the quantitative assessment of the game, although the participants were volunteers with positions in the industry, it would be necessary to evaluate in the real context of a training section so that the results could be better generalized.We can say the same about the method in the interview.The game designers had contact with SpGD.They did not use it in practice in a game project for safety training purposes.Thus, the limitations and insights obtained from the assessment served as learned lessons and improvements for the subsequent investigation cycle.
In future work, we suggest that the variables investigated in the game and which had negative perceptions and the statements about suggestions for improvements to SpGD be analyzed, thought about, and implemented in the method to attest whether the changes added positive points to the process.
We also recommend that new games for safety training that use the SpGD method go through validation steps with managers and evaluations with the target audience for these types of safety training.Furthermore, the development of a computational tool is being studied to help systematize the alignments between the elements and in the other steps of the method.
With participants, we can validate the game by separating it into two experimental groups: one with prior knowledge about the training context and the other without.It will be considered in future assessments because we can understand the level of learning of the participants in each game session.
We have opportunities to examine the continued evolution of these games.In this context, it is interesting that we evaluate the method in the long term because we can constantly improve people's formations.
The continuous formation takes into account the people's technological boundaries.The industrial staff can not know about the technology being used.
Finally, we understand that the study's contributions go beyond the academic field, offering real opportunities for the gaming industry and for startups and/or game houses to get involved in developing innovative training solutions in the industry, leveraging this promising market niche.
(2022) investigated trainers' perception of safety training in the industry.The research results show problems such as low levels of engagement in training that used traditional methods (such as videos and presentations) and point to evidence that participatory approaches and scenario-based training can increase the effectiveness of security training.Additionally, the work conducted a literature study identifying contexts and techniques on how the industry has been using games for training situations.

Figure 2 .
Figure 2. Instantiated research based on Action Research steps.

Figure 2
Figure 2 depicts the stages of this research concerning AR.Thus, the main challenge is making safety training sections more motivating and engaging, with the proposal of using digital games for this purpose being a problem diagnosed in Rufino Júnior et al. (2022, 2023) with industry professionals.At the same time, the research problem arises: how to support the design of safety training games in the industry, mitigating the erroneous transmission of training and ensuring that players/trained playfully and engagingly can learn from the training and reflect on the risks linked to work performance.As an investigation plan, we conducted a literature study to identify how organizations have used purposeful games in training contexts and what elements are necessary for this, correlating them with training evaluation proposals (Rufino Júnior et al., 2023).Still, during planning we proposed the SpGD method to systematize these games' designs.As intervention/action, we used the SpGD method to construct a digital safety training game to solve the research problem.In this way, we carried out the evaluation in two moments: i) the game, which was used to demonstrate the feasibility of the SpGD method, in addition to the analysis of experience, learning, and identification of security risks by the players/trained, and ii) the method itself, submitted to evaluation by game designers through the exemplification of its application in the creation of the game, collecting perceptions of the feasibility of use, strongness and weaknesses on the part of game designers.Finally, the assessment results led to reflection on the research artifact about solving the problem.Our research presented two reflection questions in line with the evaluation moments: 1) Did the players/trained positively perceive the game experience and safety training (learning 1) Understanding the training, 2) Mapping the training to game elements, 3) Performing brainstorming, 4) Developing the game, 5) Validating the game, 6) Assessing the game and 7) Delivering the game (Rufino Júnior and Classe, 2024).Although it is not well-represented in Figure 4, it is essential to mention that all process actors (game design team, managers, and trained people) are involved in all steps of SpGD.The Figure represents the main actor for each task.The first stage (Understanding the training), conducted by the game design team, process managers, and trainers, consists of understanding the training.Its objective is to analyze the details of the training as it is currently performed so that the gameplay reflects it.To this end, Kirkpatrick's assessment model is used to identify information in the training to fill in the questions related to its four levels.With this, we expected a general and organized view of the training.

Figure 6 .
Figure 6.Mapping the training to game elements.

Figure 5 .
Figure 5. Game "Bob Ruff in Deck is on Fire" overview: safety training elements mapped to game elements.

Figure 7 .
Figure 7. Demonstration of mapping the training to game elements.
to make this work clearer and more organized.With this, we describe the objective (goal) as Analyzing the BobRuff in Deck is on Fire digital game developed with SpGD; for the purpose of evaluation; concerning the perception of 1) game experience and 2) safety training (learning and risks); from the point of view of industrial workers; in the context of fire safety training.
the method would help overcome challenges]."[Participant #2].Also, Participants #4 and #1, answered about the viability: [I would use it because the method allows] understanding the training, using Kirkpatrick to get an idea of how the training could be evaluated.And validating with people who are there to minimize the game risk of not doing what it should.[Participant #4] [I would use] yes, it makes sense.It even makes sense to expand a little.It's more about the training line, in general.Not specifically, I haven't seen [another similar method] yet.[Participant #2]

Table 1 .
Data from Understanding the training step.
Table 3 presents the main threats and the treatments to mitigate them.

Table 2 .
Questionnaire items

Table 4 .
Questionnaire Reliability Analysis

Table 6 .
Questions of interviewsID Question Q1 The SpGD method uses Kirkpatrick's model as the base for its steps.How do you perceive utilizing this model to extract essential information from safety training and organize game design elements?Q2 The second step of the SpGD method consists of mapping elements from Kirkpatricks' model to the MDA framework.How do you evaluate this mapping concerning representing safety training adequately in terms of game design?Q3 The third step of the SpGD method aims to discuss and organize the safety training and MDA elements in a GDD.How do you perceive this creativity step concerning the game design process?Q4 The SpGD method predicts a safety game validation with managers and an assessment of the trained to analyze if the game is aligned with the training context.