A Systematic Mapping on Player’s Profiles: Motivations, Behavior, and Personality Characteristics

Digital games have become part of the daily life of a large part of the world population, reaching audiences of different ages, genders, and cultures. Games are also becoming an increasingly explored research topic in Human­ Computer Interaction (HCI), and several studies have sought to deepen the knowledge about players, identifying individual differences. Although the literature is rich in works that typify and classify players, the lack of objective comparisons can make it difficult to adopt such types to support game design or future research. Thus, this research investigates players’ taxonomies and typologies regarding their motivations, behavior, and personality character­ istics, analyzing how they explore these traits. We conducted a systematic mapping of the literature and analyzed 19 studies that propose or update players’ types, observing how they explore the above mentioned traits. The main contribution of this paper is to offer an overview of the identified taxonomies and typologies, comparing them and mapping their attributes and applications. Such knowledge is a tool for designing games and gamified systems and can support game designers to promote engagement and motivation in complementary ways in their games. It also allows researchers and practitioners to compose a multidimensional view of the different players’ types.


Introduction
In recent years, digital games have become part of the rou tine of a large part of the world's population, reaching audi ences of different ages, genders, and cultures (Rozen, 2020). Consequently, games have become an increasingly explored research topic in academia, both globally and nationally. In HumanComputer Interaction (HCI) and related disciplines, such as Games User Research (GUR), for example, several studies investigate various issues related to games and their potential for the most diverse applications, which go beyond the single focus on entertainment.
Examples of this are studies that deal with the use of games in health  (for example, for reha bilitation (Adisusilo et al., 2020; Alankus et al., 2010), in tourism (Shen et al., 2020) and in education (Mendoza andBaranauskas, 2019; Miranda et al., 2019), with special em phasis on edutainment (Aksakal, 2015; Wang andNunes, 2020) and gamebased learning (Bahrini et al., 2020), in addi tion to the growing adoption of gamification, which applies game elements in different types of applications and contexts (Bitrián et al., 2021; Toda et al., 2020. Much of the success of purposebuilt games like the ones mentioned above depends on player engagement (Hookham and Nesbitt, 2019; Orhan Göksün and Gürsoy, 2019; Adis usilo et al., 2020. Engagement is a quality of user experience characterized by the depth of an actor's investment when in teracting with a digital system (O'Brien and Cairns, 2016), which is more than being satisfied with such a system (Cairns, 2016). Affective engagement -which includes a sense of pleasure, immersion, and spontaneous flow -is fundamen tal for players to follow the gameplay complexity so that fun and highlevel engagement correlate to the gaming experi ence itself (Gee, 2005; Voulgari andKomis, 2010).
No wonder this construct has aroused increasing interest in the HCI community, and many studies have explored its nature and definitions and ways to increase and measure it, inside and outside the gaming context (O'Brien et al., 2018; Wiebe et al., 2014; Martins et al., 2020; de Souza Filho et al., 2019; Vasconcelos et al., 2018. Studies investigating more general issues surrounding games are common, such as their effects, strategies, and evaluation methods, and which ele ments make up a successful game -for example, (Santos et al., 2015; Evelin et al., 2016; Silva et al., 2020; Carneiro et al., 2019; Borges et al., 2020.
However, despite the increasing knowledge production in this domain, making a good game is not a trivial task. Game design is something intrinsically complex (Rozen, 2020). The study and creation of highquality games capable of promoting engagement and achieving their purposes require knowledge and understanding of players, which leads to the need to understand their behaviors, motivations, and charac teristics (Drachen et al., 2018). Comprehending the user and their behavior is an essential part of the HCI processes (Bar bosa et al., 2021), and the same is true for the study and de sign of games (Drachen et al., 2018; MacKenzie, 2013. Several studies have identified and classified players' traits and types according to their motivations, behavior, and emotions while playing (Bartle, 1996; Yee, 2007; Yee et al., 2011; Bateman et al., 2011; Brühlmann et al., 2020; Nacke et al., 2014. Bartle (1996) was one of the first to document specific behaviors in different types of players. Since then, many other researchers have invested efforts in detecting mo tivational profiles and how they impact playing. Some re searchers have even searched for typological classifications in other areas, such as Psychology, Anthropology, and Soci ology, that can be used to understand the players' profiles (Paulin, 2013). More recently, researchers have started in vestigating scientifically sound approaches to identify player types, demonstrating that game effectiveness may be corre lated to player types (Van Gaalen et al., 2022).
Although the literature presents several works that typ ify and classify players through different criteria, to the best of our knowledge, no research compares these proposals in terms of the differences and similarities of the behavioral and motivational profiles described in the several existing classi fications. Thus, the present research aims to investigate the taxonomies and player typologies proposed in the literature regarding their motivations, behavior, and personality char acteristics -elements directly related to the player's engage ment and experience -and survey how they explore these traits comparing them.
To this end, we carried out a systematic mapping (SM) of the literature (Petersen et al., 2015), identified and ana lyzed 19 classifications that propose or update types of play ers based on the parameters mentioned above. The main con tribution of this work lies in offering an overview of players' taxonomies and typologies. We compare their target genre, organization, classification strategy, and categories and map their attributes and applications. Such knowledge helps pro mote engagement and motivation in complementary ways and allows researchers and practitioners to look at the player with a multidimensional view of their engagement and moti vations.
We highlight that our purpose is not to propose a classifi cation, discuss the effects of games on certain types of play ers, or demonstrate any relationship or inferences related to types of players and their context and behavior in real life. We present a systematization of knowledge about player pro files that, although not exhaustive, can inform researchers and practitioners studying human factors in digital games, helping to form and refine concepts and identify underlying dimensions related to engagement and motivation. In addi tion, it also aims to assist game designers in customizing the player's experience, in making decisions for game design or even in segmenting and targeting their games to specific au diences, supporting analytical activities during the game de sign and evaluation process, including classifying mixed pro files, identifying player needs and associated behaviors, and mapping engagement techniques with the game's objectives and its interactive and narrative elements.

Background
In this section, we discuss concepts related to studying types of players and investigating their differences. We approach user engagement, a fundamental construct for the success of digital games and gamified systems, regarding the three as pects that enable the knowledge and typification of players: motivation, behavior, and personality characteristics, which together make up the focus of this work. Before, however, it is necessary to establish the difference between taxonomy and typology, two central terms for the research.

Is it a Typology or a Taxonomy?
Despite being treated as synonyms, typologies and tax onomies refer to different theoretical approaches and modes of information organization (Bailey, 1994). Part of this con fusion is attributed to the fact that a typology is a theoreti cal approach rather than a classification system which is of ten misinterpreted and wrongly developed (Doty and Glick, 1994). For Bailey (1994), typologies are conceptual classi fication modes, which presuppose arrangements for the ele ments of a set in multidimensional structures (Bailey, 1994; Smith, 2002. Units that form a typology are called types and bring to gether concepts. They do not have an empirical basis for their proposition and are more linked to the notion of an ideal type that models something (Weber, 1949). Hence, the propo sition of a typological classification is an a priori, and de ductive approach (Da Silva, 2013). Using typologies can be helpful for understanding complex scenarios and proposing generalizations as a previous step in establishing hypotheses. However, some disadvantages must be considered, such as the fact that the types are neither exhaustive nor mutually exclusive; the arbitrariness of the criteria often adopted; and its descriptive rather than explanatory or predictive character (Bailey, 1994; Smith, 2002. In the context addressed by this work, the types proposed by Nacke et al. (2014) constitute an example of typology.
Taxonomies, in turn, are empirically based ways of classi fication (Bailey, 1994). The taxonomy construction is done inversely compared to typology since the taxonomies derive from sample observations and measurements and an asso ciative arrangement of the elements to be classified as pos sible. According to ANSI/NISO Z39.192005 (R2010), tax onomies consist of controlled vocabularies whose terms es tablish a relationship with each other and have some hierar chy level (NISO, 2010).
A taxonomy can be understood as a knowledge organiza tion system. Its use can support one, or a combination of more than one, of these three main functions: indexing, retrieval or organization, and navigation (Hedden, 2010, p.55). Since the terms result from an observation of the general data set, one of the advantages of building and adopting a taxonomy is the consistency of the classification (Hedden, 2010). An example of a taxonomy related to player classification is in the work of Tondello et al. (2017), who proposes a taxonomy related to player preferences.
This disambiguation of terms is essential for the analysis of classificatory studies. It establishes a basis for interpreting their nature and underlying concepts, allowing the compari son of different studies. Some of the papers analyzed in this work confuse this terminology, resulting in methodological issues: sometimes, what was called a typology was, in fact, a taxonomy, for example.
The work of Bartle (1996) exemplifies the term confusion problem since it is easily observable that most of the works that reference it call their types a taxonomy, while countless others define them as typology. Once the terms typology and taxonomy are not synonymous, we use the word "classifica tion" as an umbrella term to refer to both throughout this pa per.

Study of player types
Besides being an essential tool for the design of games and gamified systems, the knowledge of specific types of play ers can support game designers and researchers. It offers a basis for the study of the player's experience, motivational factors, and specific goals, for example, learning or seg menting strategies (Fortes Tondello et al., 2018). Several re searchers have made efforts over time to propose and up date player models, typologies, and taxonomies -e.g., (Yee, 2006; Nacke et al., 2014; Bateman et al., 2011; Tondello et al., 2019.
The study of player types began with the work of Bartle (1996Bartle ( ), still in 1996 proposed what is usually consid ered the first typology of players based on behaviors or ap proaches presented during the game. He built the work from the observation of discussions of experienced MUD play ers 1 about motivations for playing such games. The author analyzed the players' comments and identified four distinct groups of motivations, relating them to types of players: the achievers, players who pursued winning points and leveling up their character in the game; the explorers, whose objective was to explore the game in search of understanding its me chanics and finding possible bugs; the socializers, whose in terest is focused on socializing with other players; and finally, the killers, which mainly aim to attack other players and de stroy their characters. Such types were based on four play ers' interests: interacting, acting, players, and game world. Each type identified was related to at least two of these inter ests -for example, achievers are interested in acting in the game world. In contrast, socializers are interested in interact ing with other players. A weakness of Bartle's work, how ever, is that the assumptions adjacent to the types have not been empirically tested, which makes this an informal model (Yee, 2006; Nacke et al., 2014. Despite the lack of validation, the Bartle model became seminal for fostering the first investigations into this field in later years, as its types became widely known and were the basis for several other studies (Nacke et al., 2014; Hamari and Tuunanen, 2014; Tondello et al., 2017. Yee (2006), for example, built quantitative measures on the foundations of Bartle's qualitative work to create an empirical model for MMORPG 2 which revealed ten motivation subcompo nents grouped into three components: achievement, social and immersion. Among other relevant contributions, Yee's work made it possible to raise questions about the generaliza tion that is usually made of Bartle's types, constructed from the comparison between specific scenarios subject to biases (Nacke et al., 2014).
Subsequently, studies were undertaken on other grounds, aiming to overcome possible limitations derived from Bar tle's original work, as is the case of Tondello et al. (2017), who propose a taxonomy of player preferences based on game elements and styles to play. The correlation with per 1 MultiUser Dungeons (MUD) is a game genre that resembles a chat room in which users incorporate roles within a virtual world imagined by players.
2 MassivelyMultiplayer Online RolePlaying Games player motiva tions, i.e., games that admit a massive number of players who assume the role of characters in a persistent virtual world in constant evolution, and can freely interact with each other. sonality profiles already mapped in studies in the field of psy chology was also used for the categorization of players: for example, Yee et al. (2011) used the model Big Five (Gold berg et al., 1999) to insert the players in profiles. The model in question establishes five personality factors, with which the author relates the players, namely: extroversion, affabil ity, conscientiousness, emotional stability and openness to experience. Through game data and questionnaires, Yee pro posed a way to study them through a set of archetypes. An other strand of study makes efforts to create tools that help the automated identification of different player profiles. Al though not the focus of the present work, the relevance of such research is highlighted, e.g., (Gow et al., 2012; Drachen et al., 2009).
Thus, it is possible to notice different strategies for iden tifying player types and profiles. However, until the conclu sion of the present research, no works were identified that systematically addressed the composition of an overview of existing studies, except for the work by Klock et al. (2016), who performed an SM and identified ten classifications of players.
Although it shares similarities with the research reported here, the work of Klock et al. (2016) differs in relevant points, especially: the focus on identifying studies that used player classifications, while this one focuses on studies that propose classifications, in addition to a different focus on searches -by using a string and more general selection criteria, the authors admitted works with different purposes (for exam ple, the proposition of computational clustering techniques), while we restricted the search to classifications describing different player dimensions, a perspective that seems to have expanded the scope of the results. Furthermore, Klock's work was conducted in 2016, which opens a large window for new publications. The works identified by Klock et al. that were not among our results were included in the filtering process of the present work, as reported in Section 3.

Players' engagement, motivation, and pro files
User engagement is an essential element of interaction, man ifested by a range of individual states such as attention, in trinsic interest, curiosity, and motivation (Chapman, 1997; Laurel, 2013; O'Brien and Cairns, 2016. Engagement can be seen as the "emotional, cognitive and behavioral connec tion that exists, at any point in time and possibly over time, between a user and a resource" (Attfield et al., 2011). This construct is directly related to the level of investment some one employs to interact with a digital system including behavioral, temporal, cognitive, and emotional investment (O'Brien and Cairns, 2016). Thus, as a quality of user experience characterized by the depth of such investment, engagement is more than "mere" satisfaction: it is believed that the ability to engage and main tain engagement in digital environments can generate posi tive results. for several areas, such as citizen participation, electronic health, elearning and others Cairns, 2016; O'Brien et al., 2018). Thus, there is a broad under standing that designing engaging experiences are necessary for any interactive design to be considered appropriate and successful (Doherty and Doherty, 2018).
In the gaming industry, engagement metrics are widely used. Large worldrenowned companies have already pub lished whitepapers on the subject. Facebook, for example, presented the results of a survey where it highlights the power of engagement to maintain the use of games and, conse quently, increase revenue (Facebook IQ, 2019). Google sim ilarly linked engagement with the "inclination to spend" in apps of this genre (Google, 2019). Activision Blizzard also highlighted that engagement means earnings for the game (Blizzard, 2021).
Engagement is a concept of interest in playergame inter action, as it contributes to different constructs related to the game and the player experience (Borges et al., 2020; O'Brien andToms, 2008), being closely related to other constructs, such as immersion and presence (Cheng et al., 2015; Lessiter et al., 2001. Several researches have investigated engage ment through typologies -e.g., (Brühlmann et al., 2020; Shen et al., 2020; Calegari and Celino, 2018. It is common to fraction the study of engagement to trace factors that cooper ate to engage players, or that can determine how they interact and relate to a game, noting especially the player's motiva tions, his behavior and personality traits or characteris tics (including preferences). Such factors -which constitute the focus of this work -stand out for cooperating individu ally and collectively for user engagement and can shape their experience with a game or gamified system.
O'Brien and Toms conducted a literature review and iden tified a certain equivalence of definitions about engagement. According to the authors, engagement is a cognitive, affec tive, and behavioral state of interaction with a computer ap plication that makes the user want to be there (interacting) (O'Brien and Toms, 2010). Thus, engagement occurs in three steps: engagement point, which is the beginning of the inter actional process; engagement which comprises the interac tion itself; and disengagement, which comes with the end of interaction. Game designers usually seek to keep the user in the moment of engagement for large amounts of time and minimize the process of disruption (disengagement). Although necessary for the success of a game, if abused, this resource can lead to an exaggerated engagement, which can cause serious issues such as Gaming Disorder 3 . Under standing engagement in the universe of digital games is im portant not only to avoid and denounce abusive and unethi cal practices in game design, but also to reveal the various parameters that guide human motivation itself, in a positive way.
The study of motivation is relevant to understand engage ment because making players feel inclined to return to a game and continue playing is critical to its success (Melhart et al., 2019). Motivation -a central theme in the study of player game interaction -is widely considered a determining factor 3 Gaming disorder is defined in the 11th Revision of the Inter national Classification of Diseases (ICD11) as a pattern of gam ing behavior ("digitalgaming" or "videogaming".) characterized by impaired control over gaming, increasing priority given to gam ing over other activities to the extent that gaming takes precedence over other interests and daily activities, and continuation or esca lation of gaming despite the occurrence of negative consequences. The full description is available at: https://icd.who.int/browse11/l m/en/http://id.who.int/icd/entity/1448597234 of player preferences and their Brühlmann et al. (2020) ex perience and is also related to user behaviors towards games. Thus, the study of motivation allows establishing associa tions between personality and behavior of users within the game -as done in the classifications of Bartle (1996) and Yee et al. (2011), for example. The first is based on motivations, and the second goes further by also relating personalities and behaviors.
The player's behavior, in turn, is directly related to what types of stimuli are presented to him during the experience (Calegari and Celino, 2018). At the same time, personality is a fraction of the engagement that strongly influences the player's involvement with the game. By identifying players' personalities and associating other factors, it is possible to de sign for a better and more personalized experience (O'Brien and Cairns, 2016).
Psychology studies have correlated motivation with dif ferent aspects of player profiles while engaged in gaming, including their personalities, gameplay behavior, and enjoy ment (Liu et al., 2021). For example, achievement, affilia tion, and power motivations -which are influential in motiva tion psychology -can be matched with existing player types (Liu et al., 2017). Because of that type of relationship, dif ferent works in the literature have been using a range of sub jective and objective techniques for identifying player mo tivation, seeking to identify the profile that best describes a given player, which may impact the game design (Nacke et al., 2014).
Given the above, the present work contributes to the game user research, focusing on the overview of players' classifica tions according to their motivations, behavior, and personal ity characteristics. This perspective helps to understand the relevance and nuances of studies related to motivation and engagement in digital games, and how they can be applied to game design.

Methodology
To reach the proposed objective for our research -that is, to investigate and systematize the knowledge about the play ers' classifications proposed in the literature regarding moti vations, behavior, and personality characteristics -we con ducted a systematic mapping (SM) of the literature (Petersen et al., 2015). We used the PICOS characteristics (Population, Intervention, Outcome, and Study) (Stone, 2002) as eligibil ity criteria to help structure the review, select relevant ques tions and avoid unnecessary searches.
The SM consists of a secondary study method to review primary studies and build an overview of a given area of re search, identifying opportunities and research gaps, for exam ple, Petersen et al. (2015); Kitchenham et al. (2010). The SM process followed can be summarized in three phases: plan ning, conduction and results reporting (Figure 1), which are described throughout this section.
In the planning phase, we composed a research proto col according to the guidelines proposed by Petersen et al. (2015). The protocol comprised the study objectives, the re search questions, the search string and bases, and the study selection criteria (Table 1). We listed the following research We specified the selection criteria for the review protocol and reviewed them among the researchers. We used a think aloud protocol where one researcher described the reasoning of inclusion and exclusion criteria by applying them to one study. After that, we piloted the criteria and determined the level of agreement.
We used four sources to carry out searches for papers: Sco pus, PubMed, Science Direct, and Web of Science. We se lected these databases because together, they index the ten most relevant journals and proceedings on the topics "Game design" and "Player engagement", according to the quality score calculated by the Microsoft Academic tool 4 . From the most relevant venues identified, we selected four control pa pers (Yee, 2006; Si et al., 2017; Shen et al., 2020; Kahn et al., 2015, based on their quality and relevance to our study. To compose the search string, we listed keywords from terms and their synonyms identified in a set of articles rele vant to the topic (Bartle, 1996; Kahn et al., 2015; Yee, 2006; Busch et al., 2016; Barata, 2014; Xu et al., 2012. These arti Table 1. Main selection criteria (inclusion and exclusion) used in the filtering process

Inclusion Criteria
The study proposes or updates a classification related to different: • motivations of digital game players • personality characteristics of digital game players • behaviors of digital game players • motivations of users of gamified systems • personality characteristics of users of gamified systems • behaviors of users of gamified systems

Exclusion Criteria
• The full text is not available • The paper does not describe the methodology/experiment, the proposed types and/or their descriptions • The study does not propose or update a taxonomy or typology • The study does not refer to the context of games or gamified systems • The study proposes a subcategorization restricted to a specific type of player • The study does not propose a classification, it only applies one already proposed in the literature, without modifying it cles were defined by searches on the topic and expert recom mendations. The string was iteratively tested and reviewed by three evaluators, using the control papers to ensure its quality. After the revisions, we reached the final version (Fig  ure 2), which captured all the control papers when applied to the four selected bases. In the conduction phase, we submitted the search string to the four bases and obtained 397 works, including dupli cates. Of these, 33 works (8.33%) came from Science Di rect, four (1.01%) from PubMed, 86 (21.72%) from Web of Science, and 273 (68.94%) from Scopus. To select relevant works, we filtered the resulting set of studies in a threestep process, represented in Figure 3, applying the selection cri teria established in the research protocol (Table 1). Zotero 5 was used to organize references and remove duplicates.
The first filter (F1) consisted of removing duplicates, which reduced the set of 397 works to 309 (77.83%) -Zotero automatically identified 83 works, and we manually removed another five, totaling 88 duplicate items (22.17% of the total). Then, three independent researchers performed the second filter (F2), applying the inclusion and exclusion criteria after reading the title, abstract, and keywords of the 309 remain ing studies. This resulted in the exclusion of 264 (66.5%) works. In addition to the exclusion criteria listed in Table 1, we excluded (i) secondary or tertiary studies, (ii) works that were not written in English or Portuguese, (iii) works that had three pages or less, and (iv) works that were not published in peerreviewed venues -here one should note that the work of Bartle Bartle (1996), despite being widely used in several studies, were excluded from this mapping because it did not meet this last requirement.
Aiming to verify the reliability of the filtering process, we randomly selected a sample of 40 works (10.1% of the to tal). We asked three other researchers to analyze them in dependently, applying the inclusion and exclusion criteria. We then calculated the level of agreement between the three evaluators and between these evaluators and the initial score. We used the Pearson's correlation coefficient (PCC) 6 , and all tests showed a strong positive agreement between raters, with r ranging from 0.408 to 0.720 between raters (p≤0.05).
Once attested the reliability of the filtering process, we per formed the third filter (F3) in 45 (11.33%) studies, which consisted of reading the full text of these articles and reap plying the selection criteria. In this step, two evaluators in dependently analyzed the works and later discussed and con solidated their evaluations. In cases where the evaluators did not reach a consensus (six works), a third evaluator analyzed them separately to validate the applied criteria. Thus, we excluded 29 (7.3%) studies in this filter, which resulted in the final set of 16 accepted articles, namely: Bontchev et al. Seven additional studies were analyzed using the same fil ters. They were selected not from the basis but from the only related literature review identified in F2 (Klock et al., 2016), as the present research did not include secondary studies. That work is an SM on player classifications, as discussed in Section 2. We then performed filters F2 and F3 on them.
Four of the seven additional works did not meet our inclu sion criteria and, therefore, were excluded. This procedure resulted in the inclusion of three new papers (Bateman et al., 2011; Schuurman et al., 2008; Drachen et al., 2009, totaling our final set of 19 studies for the data extraction stage. Finally, three evaluators extracted the data from the se lected works, using a form prepared with the Google Forms tool, containing questions elaborated according to our re search protocol. According to the respective authors' per spective, the extraction form allowed the identification of various qualiquantitative information about each article. The extraction form items included: basic information about the article, such as its reference and type, year and place of publication; information about the classification, including whether the proposed classification is a typology or taxon omy, its name, description, organization criteria, quantity, levels, and list of taxonomic categories, scope and target game genre; previous classifications used as a basis (if any); characterization of the methodology to propose the classifi cation; ways to apply the classification and additional obser vations about the work.
The data were organized in spreadsheets and analyzed quantitatively and qualitatively, considering the research questions. In the following section, we conclude the report ing phase.

Results
In this section, we present the results obtained by analyzing the data extracted from the accepted studies as answers to the research questions (RQ) listed in Section 3.

Classifications of player profiles (RQ1)
The accepted works propose or update classifications of play ers' profiles in games or gamified systems, using their moti vations, behaviors, personality characteristics, or preferences as a parameter. Fifteen (78.9%) of the 19 works propose new classifications, and four (21.1%) update or modify an exist ing classification in the literature. Eighteen (94.74%) classi fications focus on games and one (5.26%) on gamified sys tems, more precisely, on applications for gamified travels (Shen et al., 2020).
Regarding the entities or objects of classification explored in the works, we observed that most classifications deal with behavior (7 works -36.84%), and the least explored entity is personality characteristics, with two (10 .53%) works only (Bateman et al., 2011; Nacke et al., 2014) -as shown in Fig  ure 4. Considering the difference between typology and tax onomy (as discussed in Section 2), it is worth noting that all 19 studies accepted present taxonomies once they all used empirical methods to gather data to ground the proposed pro files. Table 2 presents a summary of the identified classifica tions. In general, the analysis of the 19 articles revealed the ex istence of three types of player classification: (1) specific to a game, (2) specific to a game genre or mechanics, and (3) generic.
In the first case, gamespecific, the classification refers ex clusively to the context of a specific game, not corresponding to a broader generalization beyond it, due to the absence of evidence to prove this. An example of this type is the work of Rodrigues and Brancher (2018), in which both the clas sification and the used data relate only to the game "Space Math".
In the second, specific to a genre or mechanics, the def inition of each player profile can be extended to an entire genre or game mechanic. A good example of this type of work is (Chandra et al., 2019), which explicitly covers the exploratory process of virtual environments, classifying play ers based on their behavior towards this type of mechanics. Another work that exemplifies this type of classification is the one presented by Shen et al. (2020), which typifies the behavior of participants in gamified travel.
Finally, the generic classifications apply to human behav ior patterns and human preferences regarding games and are not limited to a specific game or genre. An example is "The Trojan Player Typology" by Kahn et al. (2015), which is broadly based on a player's preferences. Another example would be Tondello et al.'s taxonomy (Tondello et al., 2017), which, similarly to the previous one, considers the prefer ences of individuals in the classification.

Publication types (RQ2)
We observed that the year of publication of the obtained works varies from 2006 to 2020, with 2017 and 2018 be ing the years with the highest number of publications (four works, or 21.05%, in each), as shown in Figure 5. No works published from 2007No works published from , 2010No works published from , 2012No works published from , 2014No works published from , and 2016 were iden tified. We highlight that we did not include Bartle's work in the results because it did not meet the selection criteria. Hence, the oldest work was Yee (2006), published in 2006, and which builds its proposal on the work of Bartle when ex ploring quantitative measures to analyze the original types -a study that provided the basis for only three of the four motivations originally pointed out (achievement, social and immersion), allowing to check the type killer proposed by Bartle. Regarding the publication types, 10 (52.6%) classifica tions were published in proceedings, while nine (47.4%) were published in journals, as indicated in Table 2. The jour nals Computers in Human Behavior and Entertainment Com

How classifications are proposed (RQ3)
The SM we conducted aims to bring light to the application of existing classifications for the design and study of digital games and to outline a characterization of the research that gave rise to the classifications analyzed. In that regard, all the papers reported employing applied research. Most of them re lied on quantitative approaches -12 (63.16%). One (5.26%) study focused on qualitative research, and six (31.58%) oth ers followed mixed approaches, combining quantitative and qualitative measures to deepen and enrich their results.
All studies report descriptive research. However, some of them also describe exploratory activities, such as (Ton dello et al., 2017) and (Tondello et al., 2017). The most used method was surveying, applied in nine (47.37%) stud ies. In most cases, the surveys generated data to identify and extract player profiles through advanced statistical anal ysis techniques or machine learning. Another common ap proach was game metrics, such as hours played, number of wins and losses, and areas of the map most explored. Six (31.58%) studies analyzed this data type to extract their pro posed classifications. Some works relied heavily on biblio graphic research, building and complementing their classi fications from related theories, as did Fortes Tondello et al. (2018), whose research also stands out for having an explana tory bias. Some studies have combined the three types of re search in complementary phases or steps.

Strategies to obtain classifications (RQ4)
We also searched for information on methods used by re searchers to identify the players' profiles and propose their classifications. We observed that all studies rely on empirical research, showing that all analyzed studies aimed to propose taxonomies. Most studies (15 78,94%) used surveys with players as the primary way to gather data to identify patterns and group players in profiles according to each study's fo cus. Four (21,05%) works used other methodologies involv ing clustering of game metrics. Besides the studies that used surveys, the work of Shen et al. stands out for using a spe cific methodology, called Q Methodology, to identify their players' profiles (Shen et al., 2020). According to the authors, the Q methodology is a quali tative research approach developed in 1935 that combines qualitative explanation with quantitative statistical analysis and overcomes some drawbacks of exploratory factor analy sis. In the words of Shen and coauthors, the Q methodology reveals groups of individuals who have ranked statements in the same order and categorizes them under each factor. Hence, it changes the focus from variables to respondents to explore subjectivity (Shen et al., 2020). Brühlmann et al. (2020) and Si et al. (2017) combined surveys and game met rics - Si et al. (2017) also conducted interviews with play ers. The work of Benlamine et al. (2017) was the only study that also analyzed eyetracking and physiological data. This work stands out because the authors collected multimodal data from the player's body and face (visual and physiolog ical signals) to analyze their affective and mental states and produce a machine learning model to predict players' moti vational goal orientations.
Once all works report empirical research, user partici pation is a constant in their methodologies. As mentioned above, all studies report using data from players but with dif ferent origins. Some studies collect ingame data from play ers (sometimes in retrospect, as in (Drachen et al., 2009)'s work, that studied patterns of playing behavior in the com mercial game Tomb Raider: Underworld), others conduct surveys or interviews.
Regarding the sample size in the analyzed works, we ob served the number of participants varying from 21 (Ben lamine et al., 2017) to over 50,000 players (Fortes Tondello et al., 2018; Nacke et al., 2014. For the sake of organization, we grouped the studies in zones according to the sample size, being: • Strict user participation (n < 50): three studies fit this category (Bicalho et al., 2019; Benlamine et al., 2017; Si et al., 2017. These studies offer an initial understanding or more general models of players' behaviors or moti vations, but the limited user participation compromises these proposals' generalization. • Wide user participation (50 ≥ n ≤ 1000): Seven stud ies (Bontchev et al., 2018; Shen et al., 2020; Rodrigues and Brancher, 2018; Brühlmann et al., 2020; Calegari and Celino, 2018; Tondello et al., 2017, 2019 fit this range of participants. Interestingly, only two studies had national samples (i.e., they did not compose their sam ple with people of different nationalities). This sample heterogeneity contributes positively to generating more comprehensive models and classifications. • Massive user participation (1000 ≤ n): Nine stud ies fit here (Kahn et al., 2015; Fortes Tondello et al., 2018; Nacke et al., 2014; Chandra et al., 2019; Yee, 2006; Vahlo et al., 2017; Bateman et al., 2011; Schuur man et al., 2008; Drachen et al., 2009). The two works (Fortes Tondello et al., 2018; Nacke et al., 2014) that presented the most extensive sample (50,423 partici pants) performed tests on a base test of the game Ev erQuest II -both shared the same database. Five of these studies in this range recruited participants, and the four performed their tests based on existing databases.

Organization of the classifications (RQ5)
To better understand the proposed classifications, we noted which criteria were used to classify players into types and organize them. For this, we used the work of Hamari and Tuunanen (2014), which lists four organization criteria: ge ographic (i.e., divides players into groups based on where they live); demographic (i.e. the organization is based on de scriptive characteristics such as age, gender, education or so cial status); psychographic (i.e., grouping players based on their attitudes, interests, values and lifestyles) and behavioral (ie, grouping based on patterns of behavior with or in rela tion to the game ). Sometimes, the organization may involve more than one criterion, combining them to generate more "detailed" types. Thus, it is more appropriate to indicate what seems to be the main criterion used without discarding the others. The criteria behavioral (12 papers 63.16%) and psychographic (7 papers 36.84%) were the main criteria used. As shown in Table 2, many studies also considered demographic fac tors to investigate whether descriptive characteristics such as gender, age, or even education significantly affect the types identified, as done by Rodrigues and Brancher (2018).
The categories used in the analyzed classifications are listed in Table 2. The number of categories proposed in each work -an indication of the granularity of each classifica tion -varies from three to 14 categories, as shown in Fig  ure 7. The works with the fewest categories are Yee (2006) and Fortes Tondello et al. (2018), with three categories each. The most frequent number of categories among the analyzed group was four, observed in 10 works.
Furthermore, the work of Tondello et al. (2017) stood out for having the largest number of categories, 14. However, this "leap" is explained by the authors using a different per spective to compose their taxonomy, which is the combina tion of nine game elements preferred by players and four "play styles". Only three classifications have sublevels for the identified types: (Yee, 2006), (Calegari and Celino, 2018) and (Tondello et al., 2017). The Table 2, presented above, in dicates the number of categories for each work and lists each one of them.

Domains of the classifications (RQ6)
Almost all the classifications analyzed focus on game stud ies (18 out of 19, i.e., 94.74%), and only the work by Shen Again, the classification proposed by Shen et al. stands out for studying pervasive systems, i.e., that blend the boundaries between the real and virtual world, which allows us to con sider them mixed media.
In the 18 works that focused on games, 12 (66.67%) clas sifications aim to cover the domain in general without point ing out distinctions between types or genres of games. Four (22.22%) works have a more restricted scope, as they were developed for specific games or types of games (Drachen et al., 2009; Brühlmann et al., 2020; Calegari and Celino, 2018; Rodrigues and Brancher, 2018. Drachen et al. (2009) and Brühlmann et al. (2020) focused on classifying players of specific games, the first on Tomb Raider: Underworld, and the second on League of Legends. Calegari and Celino (2018) focused on games with a purpose or GWAP (Games With A Purpose), which are games that encourage users to perform tasks with an entertainment reward. Rodrigues and Brancher (2018) focused on a classification for educational games.
Some works, however, make distinctions between "seg ments" of games, for example, multiplayer and online, as is the case with Yee (2006) and Chandra et al. (2019) who built their rankings by focusing on MMORPG games. Regarding the genres of the games targeted by the classifications, most works (eight -44.44%) were not tied to a specific genre. In contrast, the others were equally divided among other genres, as shown in Figure 8.
The work of Kahn et al. (2015) stood out for targeting two genres, MMO 7 , and MOBA 8 . The definition of genres and types of games is a frequent cause of confusion and disagree ment, and there is no standard followed in the industry, and the studies of games (Grace, 2005). Therefore, the division used here was based on the complementary views of Grace (2005) and Apperley (2006).
Regarding their purpose, most of the papers focused on games for entertainment, as shown in Figure 9. The work of Nacke et al. (2014) stands out for embracing games for all purposes, which seeks broad applicability of the typology. 7 Massively Multiplayer Online Game -games capable of supporting large numbers of players simultaneously and connected 8 Multiplayer Online Battle Arena -a type of game in which the player controls a character in a battle between two teams, to defeat the enemy base  Calegari and Celino (2018) work with GWAPs, which may have varied purposes within the spectrum of games that aim to get player collaboration on something.

Use of game elements in the classifications (RQ7)
We observed if and how the studies considered game design elements to compose their classifications to gain perspective on how they explore games' particularities. We noted that only two (10,53%) of the 19 works do not directly incorpo rate game elements in their classifications (Benlamine et al., 2017; Bateman et al., 2011. That can be explained by the fact that Bateman et al.'s work offers a view centered more on players' subjectiveness, focusing on players' personality factors and play styles. However, the work conclusions inspired a new player sat isfaction model, the BrainHex (Nacke et al., 2014), based on game elements. Regarding the work of Benlamine et al. (2017), this study used a different approach to obtain data to compose their classification (monitoring players' visual and physiological signals) and thus explored game elements indirectly, in the form of game scenes and how they affect players' motivations.
Considering the other 17 studies, we noted that we could group the game design elements considered by the classifi cations into two types: conceptual and parametric. We con sidered conceptual elements as the qualitative aspects ex tracted from theories or mechanics that make up conceptual aspects of games -for example, game genres, mechanics (e.g., cards, strategic management, roleplaying or puzzles), and motivations (e.g., surprise, socialization, progression, ac cumulation).
Fourteen ( (2008). The percentage of studies in this group suggests a slight preference for exploring this element type.
On the other hand, parametric elements would be the quan titative elements used to detect players' behavior patternswe can point as an example of these quantifiable elements: game time, win rate, score, number of shots, and lives, among others. Nine studies (47,37%) considered this type of el ement to determine their profiles: Bontchev et al.

Tools proposed to identify players profiles (RQ8)
Aiming to offer some additional information that could help other researchers and practitioners use the classifications identified here, we noted which studies proposed tools to help other researchers apply their classifications to identify which profile fits a specific player or group of players. Eight (42,10%) papers present, besides a classification, an instru ment to this end (Bontchev et al., 2018; Shen et al., 2020; Benlamine et al., 2017; Vahlo et al., 2017; Tondello et al., 2019; Bicalho et al., 2019; Bateman et al., 2011; Kahn et al., 2015. In general, the proposed instruments were derived from studies that used surveys to gather data on players' traits. It generated selfreport tools (questionnaires or scales) that present a set of questions that help to relate players with pro files regarding behavior, preferences, and motivations. Ta ble 3 indicates these studies, specifying the entity their in strument address, the types they identify, and the number of items/questions each one presents. It is worth noticing that not all of the instruments are validated, and, in some cases, they still need further studies to prove them.
One of the other works (Shen et al., 2020) that do not pro pose a specific instrument gives clear recommendations on using the methodology used in the study to obtain their clas sification (the Q methodology) to identify other players' pro files. This methodology, briefly discussed above, is widely used in the social sciences and humanities to seek a more quantitative bias to investigate beliefs, attitudes, behaviors, and opinions (Herrington and Coogan, 2011). Applying this methodology to identify players' profiles is a differential of this work and shows an alternative path to others who want to explore another perspective to this process.
Regarding the studies that relied on game metrics cluster ing with machine learning techniques, it was unclear if some proposed, in fact, a tool or algorithm for general use to iden tify profiles given other datasets. However, the description of their methodologies, processes, and lessons can undoubtedly support other studies that aim to do the same.

Discussion
The analysis of the results allowed us to draw an overview of the players' classifications available in the literature and highlighted relevant points to be considered in the mission to make these results more beneficial for the design and evalu ation of games. Therefore, in this section, some insights and concerns that emerged during the conduction of the research are discussed, aiming to provoke reflections and contribute to game studies.

Lack of coherence in the use of terms
In games research, the same phenomenon previously de scribed by Doty and Glick (1994) occurs: many works use the terms typology and taxonomy interchangeably, although they are not synonymous. We agree with Doty and Glick that such confusion of terminology can impact the work method ology.
However, we also observed that this misinterpretation makes understanding the classification and its application in game design and evaluation challenging. In addition, the in coherent use of terms makes an objective comparison of the classifications difficult since it becomes hard to identify, and it is necessary to analyze the terms used in the context of each work.
This issue echos the analysis of Kultima who affirms that the lack of conversation between game studies and general design research is visible yet historically explainable (Kul tima, 2015). We also agree with the author when she states that, considering the field's maturity, incorporating more sound design research frameworks could alleviate this epis temic gap, which we find to be not only between the practice and academia but also inside game theoretical research.
The lack of standardization in using these terms makes it difficult to correctly identify the proposed classifications and the possibility of applying the terminologies for subse quent studies. This issue may come from the growing ef fort to solidify the theoretical framework on games, which is also manifested in the debate and adequacy of using dif ferent terms in this area. For example, Darin and Carneiro (2020) and Borges et al. (2020) discuss the lack of consensus on what the term "player experience" means, what dimen sions it encompasses, and which human characteristics and market practices it impacts.
We highlight that the use of technical and scientific terms must be done carefully, as it can generate confusion and in consistencies in research, design, and game evaluation. The misuse of terms makes it difficult to systematize and de velop mature research on human factors in games and diffuse poorly founded knowledge. Thus, it is necessary to discuss and identify ambiguities about terms and concepts related to player behavior and game elements, aiming to solidify them, contributing to the maturation of practices and research in games and interfaces.

Relationship between different player classifications
When surveying the different classifications of players, one could think that seeking a correlation between the elements of the classifications towards a unified model would be desir able. However, in this work, we did not look for those rela tionships. As we see it, one needs to recognize that classifica tions are an abstraction of the complexity of human behavior, emphasizing some characteristics, to the detriment of others, to group individuals into types. A unified model would look for similarities between ab stractions. Therefore, a new abstraction must be created, looking for characteristics between different classifications. That "abstraction of previous abstractions" increasingly di minishes the ability to represent the complexity of human behavior, which is likely to move further and further away from the real motivations of individuals.
We demonstrate our point by analyzing the attempt at uni fication by Stewart (2011), in which the author proposes a unified model for personality and play styles. The work be gins by approximating the four types of players proposed by Bartle (1996) with the four types of temperaments by Keirsey (1998). Based on each author's descriptions of types, Stewart Stewart (2011) proposes that Keirsey types are supersets of Bartle types and correlate the models. After defining the four types of players from the union of the theories of those two researchers, the author uses different known classifications, relating their types to the now unified elements.
In the approximation created between the Killer type (Bar tle) and the Artisan type (Keirsey), the author emphasizes the common characteristic between the types as manipulation (the author prefers to call the Killer type Manipulator). When describing the Killer, the author points out the player's main characteristics: the desire to impose themselves on other play ers and demonstrate their superiority over others. As for Arti san, the author emphasizes the player's desire to have power over everything in their world.
Stewart then presents some expressions from the original works related to their types to determine that correlation. If we tabulate and observe the terms the author uses to correlate the two classifications (Table 4), one can realize that some elements can be directly related. Still, many others do not have correspondence or would need a greater abstraction to be associated, removing a type's characteristic so it can relate to the other one. Table 4. Summary of analysis of correlation between characteristics of the Artisan (Keirsey, 1998) and the Killer (Bartle, 1996)  Thus, a unified model becomes a new abstraction. The sense that it more comprehensively represents players' moti vations turns out to be just a new way of classifying themperhaps even less representative because it is an abstraction of several abstractions. However, executing a unification is not unproductive since it makes us reflect on the characteris tics of the types of each classification.
Still, we believe that it is more productive for designers, game creators, and researchers to have a broader knowledge of the possibilities of organizing players' motivations. Then they can select the one that best helps in developing a spe cific project, considering the criteria used for the classifica tion organization. For that, we believe that this mapping can contribute to this choice.

Players profiles and game elements
The players' profiles and motivations can impact how a group engages with a game. Hence, the analysis of player motivations can inform game designers and researchers on the behavioral patterns of game players, and they may use those behavioral patterns to drive eventual player engage ment through game elements. For example, achiever players engage with tangible rewards for achievement, like coins and badges, while explorer players engage by exploring the rules and bounds of the game environment (Stefan et al., 2017).
However, our analysis showed that classifications that seek to explain the behavior and motivations of different player profiles -in most cases -do not directly relate the identified types to specific game elements. In the context of games, the term "element" can represent different ways of di viding or understanding the parts of a game. Such elements have been identified in the game design literature at differ ent levels of abstraction. Schell, for example, defends four elements: mechanics, narrative, aesthetics, and technology (Schell, 2020).
Another famous approach is the framework MDA, which proposes the division into mechanics, dynamics, and aes thetics (Hunicke et al., 2004). The possibilities are end less since the elements that make up the games can still be grouped in terms of components, environment, players, context, rules, mechanics, theme, interface, and information (Järvinen, 2009). Or -as Bjork and Holopainen presentcan be defined by design patterns exceeding 100 (Bjork and Holopainen, 2004).
Although this diversity in understanding the elements that make up a game derives from seeing which characteristics are essential to be observed, depending on the purpose, it is not easily comparable -and the same happens to players' classifications. One can perceive the different perspectives on game elements in how they are delimited by different def initions, in different levels of granularity, views, and forms, making it difficult to compare which game elements from one work relate to the elements from another. The same hap pens to how players' profiles can relate to the different cate gories of game elements, ultimately confusing their applica tion in practice.
However, the work of Tondello et al. (2019) stands out in discussing that analyzing player types in a way that is directly related to game elements makes the application of types more direct and practicable. The authors translate the game ele ments into activities that players engage in while gaming, such as progression, action and roleplaying, resource man agement, exploration, or combat.
In the present work, we confirmed this perspective by iden tifying that most approaches to studying player types seem to ignore the relationship between such types and game ele ments. We agree with the authors that much work focuses on higherlevel factors such as immersion or achievement, mak ing applying such classifications difficult. Although we iden tified some preliminary initiatives to relate the behavioral profiles of players with the components of games (Paulin, 2013), it is necessary to deepen the study of these elements and their correlation with the motivational and behavioral profiles of players.

Gaps and research opportunities
As detailed above, the results of this SM indicate a research gap for carrying out experimental studies that map game ele ments to player profiles. Our findings also indicated the need to investigate personality characteristics further -an essen tial factor for engagement but still little explored. Further more, some expanding domains, such as serious games and gamified systems, were timidly addressed in the classifica tions (only in one work each).
As recent research indicates, the effectiveness of serious games appears to be correlated to the degree to which play ers like the game (Van Gaalen et al., 2022). Hence, using a player classification in the games user research can help game designers to identify more clearly interpretable patterns and show how players perceive play to design more com pelling games.
It is also necessary to highlight the scarcity of Brazilian re search on the subject -in formal and informal searches, only two works Brancher, 2018; Bicalho et al., 2019) were identified in a national event. There is an opportu nity of developing more research that seeks to investigate the profiles of Brazilian players and users, aiming to understand cultural particularities.
In this sense, we agree with the conclusions drawn in the work of Miranda et al. (2021) when they question the use of game research tools made from and for other cultural and so cial characteristics, which are only assumed to be valid for lo cal users. We underline the work of Kahn et al. (2015) which stood out for validating its typology in two different cultural contexts (with American players and, in other ways, with the Chinese audience), which is an excellent example to be fol lowed in future research.
Finally, it is necessary to emphasize that there is still a wide range of games to be examined. Works that explore online games and their direct variations (such as MMO, MMORPG, and MOBA, among others) are common, while few works deal with offline or singleplayer games, for ex ample. As pointed out in the work of Fortes Tondello et al. (2018), we also identified that research on player classifica tions has tended to fall into the same group. Observing our findings, we have the same understanding as Hamari and Tu unanen (2014) when their work suggests that it can compro mise the generalizability of the results.

Research Limitations
Apart from the contributions to the literature in this domain, this study is not free from limitations characteristic of most systematic reviews and mappings. First, due to the nature of the process of selecting, filtering, and extracting data from articles, it is possible that relevant studies have not been an alyzed. In addition, only four bases were selected. Although evaluating the bases' quality was decisive in the choice of these four as representative of the research scenario in na tional and international games, the authors understand that there may be works that are relevant to the objectives of this research that are not indexed in them.
Furthermore, another limitation can be found in the fact that the searches did not capture some relevant works due to the wide variety of terms papers use for classifications (which are not always synonymous, as discussed in Section 5). Such lack of standardization makes it difficult to build a complete and farreaching search strings. To mitigate such limitations, relevant works identified in another SM were manually included and properly filtered and analyzed, as de scribed in Section 3.

Conclusion
Although the literature is rich in works that typify and clas sify players, the lack of an objective analysis can make it difficult for researchers and practitioners to decide how to employ them to support game user research and game design. Thus, this research investigates players' taxonomies and ty pologies regarding their motivations, behavior, and personal ity characteristics, analyzing how they explore these traits.
Our results indicated various categories to describe play ers, different ways to propose and validate such categoriza tions, and tools to assist researchers and practitioners in iden tifying players' profiles. They can be used to help customize the player's experience and increase the engagement and mo tivation of specific groups with the interactive and narrative elements most suitable to them. They can also be employed in experimental studies that analyze physical, psychological, and social factors impact on player profiles.
Our future work focuses on analyzing the relationship be tween different interface and interaction elements in games with the level of motivation and engagement of different player profiles. Attempts were made to relate players' behav ioral profiles with game components, but still, superficially (Paulin, 2013). Research on game resources is pretty frag mented, and experimental studies are needed to map game resources to player engagement (Boyle et al., 2016), relating them to their motivational and behavioral profiles. Thus, in the future, this research will address the search for a relation ship between player profiles and their level of engagement with the game, given specific elements of digital games.