Personalized Gamification: A Model for Play Data Profiling Dorina Rajanen Mikko Rajanen Interact Research Unit, Interact Research Unit, University of Oulu, Finland University of Oulu, Finland dorina.rajanen@oulu.fi mikko.rajanen@oulu.fi gamified system whose usage produces the expected outcomes. Motivation and fulfilling goals are in strong Abstract relationships with individual differences. It was shown already in the 70s that personality traits of the users This paper proposes model for introducing should be considered when designing information personalized game-design elements in a systems [2]. Since then, the design informed by gamification system. The model is based on personality and human values has been proposed by user-centred design, human values theory, and many scholars (see e.g., [21,22]). Moreover, it has been gamification design framework. The proposed shown that personality traits affect the interaction with model promotes the idea of a baseline game and the response to different technological systems as component that is meant to acquire online, well as the adoption of technology (see e.g., real-time data about the cognitive and [9,12,23,24]). emotional state of the individual, and based on In this paper, we propose a model for introducing the collected data to adjust the gamified personalized game-design elements in a gamification system to the state of the user. This system. The novelty of this model is that it promotes framework, which we name Play Data the idea of a baseline game component that is meant to Profiling (PDP), describes a model of acquire online, real-time data about the cognitive and collecting and processing data before, during emotional state of the individual, and based on the and after the actual use of the gamified collected data to adjust the game (gamified application) application in order to optimize the subsequent to the state of the user. The role of the baseline game user experience and outcome. Implications component is to assess the personality and the current and future work are discussed. state of the user. This information will be processed and classified into predefined player (user) profiles, 1. Introduction which will determine the type of game interface and mechanics to be loaded in the current game session1. Gamification is defined as the use of game elements in This type of game component would ensure an optimal non-gaming systems to improve user experience and user experience and outcome. The proposed framework user engagement [7]. The first design ideas of using fun draws upon established design theories and framework and game for computer-aided learning belonged to such as the user-center design [11] and gamification Malone in 1980s (see [16,17]). The design of design framework [31], as well as on the human values gamification involves introducing game-design theory [30] and research on personality and moods, elements into the software development of the target emotions and affective states. system. Because of its potential benefits, gamification has attracted attention to both researchers and developers and has been explored in many use contexts 1 such as education and use of libraries, usability testing, In our model, gamified system and game are used personal health informatics, risk management, and interchangeably, as well as users and players. In the model, enterprise information systems (see e.g., [1,3, we refer to the gamified application or system by the more 6,18,19,27]). However, the design of a successful generic term “game” as we adopt the conceptualization that gamified-system is still challenging. In particular, two a gamified system is built with the assumption that the challenges exist: 1) to design a gamified system that system will be perceived by users in the same way a game motivates people to use the system, and 2) to design a is perceived by players by invoking similar psychological experiences as in gameplay (see [10]). Thus, in our Copyright © by the paper’s authors. Copying permitted for private and conceptualization, a gamified application or system has the academic purposes. characteristics of a game, but it is used in a non-game In: M. Meder, A. Rapp, T. Plumbaum, and F. Hopfgartner (eds.): context such as work or education to stimulate user Proceedings of the Data-Driven Gamification Design Workshop, engagement. Therefore, we also refer to users as players. Tampere, Finland, 20-September-2017, published at http://ceur-ws.org The paper is structured as follows. Section 2 games and can be associated with intrinsic and extrinsic describes the theoretical background and related work. motivation (see [13]). Werbach and Hunter [31] Section 3 describes the proposed framework. Section 4 approach the user by guiding the development to define discusses the implications and future work. the business goal, target behaviors, and players’ characteristics and types. Nevertheless, in the 2 Theoretical Background gamification domain, the gamified solution has to be both engaging and fit to the organizational purpose; therefore an integration of UCD and a gamification 2.1 User-centred and gamification design framework would ensure that usability and UX User-centred design (UCD) is an approach used requirements are fulfilled by using a range of methods throughout the entire system development cycle to and techniques centered around usability and UX at ensure the developed system fulfils the usability each step in the gamification process (see [25]). requirements [11]. This means that the developed system matches the user profile, goals, and needs (see 2.2 Personality, values and emotions e.g., ISO 9241-210 [11]). The UCD process is iterative and incremental; different designed solutions are Both the UCD and the gamification design models created and tested [26] by employing usability stress the importance of understanding and defining the knowledge and methods. The design activities specified user characteristics. In addition, the UCD provides the in ISO 9241-210 are: 1) understand and specify methodology to ensure that these characteristics are organizational requirements; 2) understand and specify well understood and taken into account in the design, context of use; 3) produce design solutions; 4) evaluate thus, complementing the gamification design design against requirements. framework by providing actionable guidelines to ensure Usability is very important for the game to be well satisfactory usability and UX. One of the guidelines received and successful [8] in terms of satisfaction, refers to defining all relevant dimensions of players in efficiency, and effectiveness [5]. In addition, user the context of use and describing the players in terms of experience (UX) is a dimension of user satisfaction that personas (i.e., user representatives) (see [4]). should be taken into account when designing and Personality, psychology, and behavioral phenomena evaluating games. User experience is defined as being should be addressed when profiling the players [7] and the sum of an individual’s “perceptions and responses these profiles should be taken into account in the that result from the use or anticipated use of a product, design. system, or service” [11]. UX is thus associated with the However, studies focusing on individual differences internal state of the user when interacting with a define the individual characteristics in several ways, for product in different stages of use (before, during, and example, by demographics (e.g., age, gender, after) and it is believed to affect the overall satisfaction education), personality type (e.g., the Big Five with the product (see e.g., [5]). personality types [9,12]), behavioral-disposition traits UCD approach is applied successfully in various (approach and withdrawal motivation [23,24]), human contexts of use of information systems; however, in the values [29]. Of particular interest is the human values game development, the application of UCD is more theory by Schwartz [30] which posits that “(1) values challenging because the entertaining nature of the are concepts or beliefs, (2) pertain to desirable end games makes more difficult and complex to design and states or behaviors, (3) transcend specific situations, (4) guide selection or evaluation of behavior and events, assess the fun of the game and the user engagement. and (5) are ordered by relative importance.” The human In the gamification context, there are several values can be therefore seen as individual gamification frameworks proposed in the literature that characteristics that guide and motivate people in their address the user dimensions and propose design life, and examples of such value types are self- recommendations to ensure the fun in the game (e.g., direction, hedonism, and achievement. This theory has [7,31]). Deterding et al. [7] identify as the most been also used in system design to cluster users by effective game design elements that elicit user motivational values (see e.g., [29]). Moreover, recent engagement the so-called points, badges, and research shows that mood, cognitive and emotional leaderboards; these elements appear in most of the states affect the user experience and responses to interaction with a system, product or service. However, method in UCD is to identify and define persona satisfaction and affective states are complex concepts to profiles based on individual characteristics such as define and measure as they are multidimensional and personality, demographics, roles, and needs. However, time-dependent [28]. Scherer [28] classifies the these profiles identified during the system development affective states based on two dimensions (duration and cover only partially the characteristics of the users, intensity) into several constructs such as: personality namely the ones that are relatively stable such as age, traits, attitudes, interpersonal stances, mood and gender, and personality traits. On the other hand, users emotions. Emotions themselves can be defined and have different moods, values, and psychophysiological categorized in several ways such as discrete (anger, states (e.g., high level of stress as indicated by heart happiness, etc.), bi-dimensional (along the valence and rate or electrodermal activity) that are fluctuating over arousal axes) (see [14]). a certain period of time and which influence the user Given the diversity of users and user dimensions, the experience. profiling of players and the clustering of the players by Therefore, we propose that in addition to profiling meaningful profile characteristics is important for the the target users during the development cycle, a success of the game according to UCD and gamified system should be built in such a way that it gamification principles (see [11,31]). However, the collects on-line, real-time play data based on which clusters/profiles are not necessarily stable in time, but profiling continues also after the system development. they change over time (i.e., one player can be Thus, the proposed Play Data Profiling (PDP) model categorized according to his/her characteristics and states that the gamification elements can be adapted behavior as belonging to one profile cluster at a certain and personalized based on the interaction and/or time, but later his/her profile can change). Moreover, psychophysiological data collected before, during and there are individual characteristics that are inherently after each gameplay session in order to fit best to the fluctuating such as mood, tiredness, and emotions. current state of the player (user). The assessment of the Table 1 illustrates different types of individual player (user) or game session at a particular time may characteristics classified by the degree of variation determine a transition from one profile cluster to along time. As they influence and are being influenced another or an update in the player profile, which in turn by the gameplay, system developers should take them may determine a change in the game interface and into account when developing a game or a gamified gameplay. Thus, this model proposes that a system. In the next section, we propose a model of gamification system is composed of a set of alternative gamification system that takes into account this designs corresponding to different profiles and player variation and transition of player profiles. (user) states. These alternative designs are pre-built based on the UCD guidelines and gamification design Table 1: Individual characteristics of players by degree principles. During the play, on-line evaluations of the of variation along time game sessions and the player are performed using built- in game analytics which provide new information for Small or no Moderate High variation the profiles and the current state of the player. The play variation variation data profiling is then used to personalize the gameplay Gender Attitudes Mood and the game interface. Nationality Values Emotions PDP model presents a gamification system as Personality traits Socio-cultural Cognitive load consisting of three parts: 1) the pre-play data profiling component, 2) gameplay, and 3) post-game analysis Education level Experience Psychophysiological (see Figure 1). The first component is meant to provide a baseline interaction with the system of a short duration (e.g., 3 3 Play Data Profiling (PDP) Model min) during which the player is assessed; for the player it acts like a “warm-up” session before the actual play. UCD and gamification design models focus on system For example, different stimuli related to the business development, namely they provide guidelines to design, objectives of the gamified system and to the overall development, and evaluation of a gamified system, categorization of the player persona profile categories software or service that fulfills the needs of the players are presented, while the system collects and analyzes and organization. For this purpose, one widely used different interaction events (e.g., mouse movements, choices, time taken) and, if possible, integrated). After the baseline is created, the new psychophysiological measures (eye tracking, heart rate, information obtained during pre-play session is fed skin conductance, etc.) to measure stress level, back into the player profile, and thus some updates in emotional states, and cognitive load. During and after the player profile are possible and enabled by the this “warm-up” session, a machine learning based pre- system (thus, the system is constantly learning the play data profiler evaluates the user interaction events player profile and the baseline is created not only based and the psychophysiological measurements in order to on the real-time data, but also on historical data and the identify the best player persona profile category for this player profile by using machine learning techniques). individual player. The system uses this categorization To build this component both the UCD and to personalize and tailor the gamification elements gamification design principles are employed; the (target behaviors, activity loops, elements of fun, and assessment of the profile at this stage is automatic, but tools) in the gamified system in next step in order to it has to include knowledge of the users, their needs, maximize the player engagement, fun, and fulfillment and characteristics, as well as behavioral data (such as of the business objectives. user selections) collected during the baseline. The second component in the PDP model represents the actual gameplay or system use, personalized so it matches the player profile and the current state Baseline play evaluated by the first component. While the user profiling interacts with the gamified system, the system collects logs of interaction events (and psychophysiological measures) until the play session ends. This component Create baseline based on profile Personalize game Update profile based on baseline also utilizes elements from both the UCD and based on profile gamification design principles; the designers of the system must identify the business objectives, define the target behaviors, activity loops, elements of fun, and Game(play) Player the available tools employing the UCD process and profile methods. The third component in the PDP model represents a machine-learning component for processing the play data of the user, his/her current mood, values, Data collection and analysis Update profile emotional and cognitive states and the success of the gamification. Were the business objectives and target behaviors successfully fulfilled? Was the gameplay Post-play session fun to the user? These assessments as well as profiling the identified play patterns are fed back into the profiling of the player, and the update is used in further sessions of the same player or other players. Along with the conceptual model of collecting and Figure 1: Play Data Profiling model. The baseline processing play data in different stages of use of a gamified system (Figure 1), we illustrate the model “warm-up” component collects and analyses pre-play from the perspectives of a designer and of a machine data, based on which a personalized game is provided. learning developer. Figure 2 describes the PDP model Player profile is also updated with the new information. from the designer perspective. The UCD process and The gameplay component collects the data during the methods, and the gamification design principles are actual game. The post-play component uses the game employed to define meaningful profiles, to create play data to update the profile. alternative game designs (interface, mechanics, gameplay elements) to match the business objectives Every time a baseline is created at the beginning of a and the player profiles, and to define criteria for the game session, some of the information is taken from the post-play evaluation and profiling. Here designers player profile (i.e., the player profile acts like a schema employ various methods, techniques, and tools in order or template on which pre-play data are contrasted and to collect and analyze the data such as heuristics and usability engineering [20], psychophysiological learning algorithms and methods should be developed measurements [15], and other user testing methods. and utilized in order to obtain timely information about the current player and categorize it in a meaningful profile. Moreover, at post-play stage the huge amount of log data (as well as physiological data) requires also Baseline play profiling machine-learning methods to make sense of the data and update the profile database with new information about the game session, player, and fulfilled objectives. As in different stages of game development and UCD and game play different data are collected, Table 2 gamification Game(play) Player illustrates different types of data collected during the design profile gamification cycle. The table is not exhaustive and not principles all categories are compulsory (for example, psychophysiological measurements are not always possible to obtain during the actual game play). During the design and development, the range of data Post-play acquisition methods and protocols (e.g., experiments, profiling tests, observations, surveys, expert evaluations, etc.) is limited only by the available resources; on the other hand, during baseline and gameplay the acquisition Figure 2: Play Data Profiling model from the should be carefully implemented so not to disrupt the designer/developer perspective. play experience. Inquiry methods can also be employed if the questions are well integrated into the game interface (for example, at the end of the game session one question can be “Are you satisfied with the game session today?”, or depending on the game domain and business objective a more concrete, context-specific Baseline play question or a question to describe the mood of the profiling player). Table 2: Gamification cycle and data acquisition Machine- Player learning Game Design and Game(play) Baseline and Game play (2) profile methods Development (1) User inputs (e.g., mouse clicks, Observational data scrolling) Inquiry data (questionnaires, User choices or selections surveys, focus groups) Post-play Heuristics Time profiling User testing using various methods Psychophysiological measurements (see column 2) including eye tracking 4 Discussion Figure 3: Play Data Profiling model from the machine- learning perspective. This paper proposed a play data profiling model for data-driven personalization of gamified systems. The Figure 3 describes the PDP model from the machine- model is based on the user-centred design model, learning perspective. As the amount of the gathered human values theory, and gamification design during the baseline and actual play is very big, and the framework. The proposed model introduces the concept processing needs are in real-time, advanced machine- of a baseline game component that acquires online, real-time data about the cognitive and emotional state between different player profiles identified during the of the individual and based on the collected data, the design phase. system adjusts the game interface and elements to the current state of the player. This model, named Play Data Profiling (PDP), describes a process of collecting 4.2 Future Work and processing data before, during and after the actual play in order to optimize the subsequent user In the future, the PDP model might be empirically experience and the outcome from both the user and the evaluated through an evaluation prototype such as a business perspectives. proof-of-concept system. This prototype could be a simple, small-scale gamified system, for example a website or app with educational goals. To implement 4.1 Implications the model in a real system requires managing the The PDP model has implications to research and following challenges: 1) availability of computational practice. First, it provides the researchers and resources for data storage and processing, 2) data practitioners a model of personalized gamified system security, 3) data complexity, 4) design complexity. that utilizes behavioral, physiological, psychological, However, these complexities can be tackled by environmental (context of use, business objectives), employing an incremental and iterative approach which and social data as well as machine learning (data starts with a simple system and adds new features over mining, statistics, and AI) techniques to provide time. The idea is to design the gamified system in a way tailored game elements to users with different that it allows the system to adapt as it learns the users’ characteristics. This model can be empirically tested behaviors and profiles. and further refined and expanded. Second, it provides researchers and practitioners an iterative model for References continuous player persona profiling before, during and after gameplay. Third, the gamified systems [1] P. Bajdor and L. Dragolea. The gamification personalized through PDP model have an impact on as a tool to improve risk management in the users, since it can be assumed that personalizing the enterprise. Annales Universitatis Apulensis gameplay experience to suit the current player profile Series Oeconomica, 13(2), 2011. and psychophysiological state of the user will make [2] M. L. Bariff and E. J. Lusk. Cognitive and each gameplay experience more engaging, fun, personality tests for the design of management positive, and productive for the user. Therefore, information systems. Management Science, designing gamified systems using PDP model can also 23(8) (pp. 820-829), 1977. help the system to fulfill its business objectives, since [3] M. Barr, K. Munro and F. Hopfgartner. they will have a positive impact on target behaviors. Increasing engagement with the library via However, the play data acquisition, analysis, and profiling for personalizing the gamification carries gamification. In Proceedings of the GamifIR some significant ethical implications and challenges 2016 Workshop, Pisa, Italy, 2016. that the designers have to take into account. The users [4] R. Bartle. Hearts, clubs, diamonds, spades: should be made aware that their interaction with the Players who suit MUDs. Journal of MUD system and/or their psychophysiological state will be research, 1(1), 19, 1996. logged in particular ways for data acquisition, analysis, [5] N. Bevan, J. Carter and S. Harker. ISO 9241- and player profiling. Ideally, the user should be given a 11 revised: What have we learnt about choice regarding the degree of personalization they are usability since 1998?. In International comfortable with. This degree of voluntary Conference on Human-Computer Interaction personalization could range from no personalization at (pp. 143-151). Springer International all, to full personalization, and various degrees of Publishing, 2015. privacy in between. In case of user not giving his/her [6] P. Brauner, A. C. Valdez, U. Schroeder and consent to any personalization or data collection, the M. Ziefle. Increase physical fitness and create gamified system should have a pre-defined generic health awareness through exergames and player profile, which should be the best compromise gamification. In Human factors in computing and informatics (pp. 349-362). Springer, research. Journal of Gaming & Virtual Berlin, Heidelberg, 2013. Worlds, 3(3), (pp. 181-199), 2011. [7] S. Deterding, M. Sicart, L. Nacke, K. O'Hara [16] T. W. Malone. What makes things fun to and D. Dixon. Gamification: using game- learn? Heuristics for designing instructional design elements in non-gaming contexts. In computer games. In Proc. of the 3rd ACM CHI'11 Extended Abstract. on Human Factors SIGSMALL symposium on Small systems (pp. in Computing Systems (pp. 2425-2428). ACM, 162-169). ACM, 1980. 2011. [17] T. W. Malone. Heuristics for Designing [8] M. A. Federoff. Heuristics and usability Enjoyable User Interfaces: Lessons from guidelines for the creation and evaluation of Computer Games. In Human Factors in fun in video games. Doctoral dissertation, Computer Systems. (pp. 1-12). Norwood, NJ: Indiana University, 2002. Ablex Intellect Books, 1984. [9] S. Halko and J. A. Kientz. Personality and [18] M. Meder, T. Plumbaum, and S. Albayrak. Persuasive Technology: An Exploratory Study Learning gamification design – An usability on Health-promoting Mobile Applications. In first approach for the enterprise infoboard Proceedings of the 5th International experiment. In Proceeding of theGamifIR Conference on Persuasive Technology 2016 Workshop, Pisa Italy, 2016. (PERSUASIVE’10). Springer-Verlag, Berlin, [19] C. I. Muntean. Raising engagement in e- Heidelberg, (pp. 150–161), 2010. learning through gamification. In proceedings [10] K. Huotari and J. Hamari. Defining of the 6th International Conference on Virtual gamification: A service marketing perspective. Learning ICVL (pp. 323-329), 2011. In Proc. of the 16th International Academic [20] J. Nielsen. Usability engineering. Academic Mindtrek Conference, Mindtrek ’12, (pp. 17– Press, Boston, 1993. 22), New York, NY, USA, 2012. ACM. [21] O. Nov and O. Arazy. Personality-targeted [11] ISO 9241-210. Ergonomics of human system Design: Theory, Experimental Procedure, and interaction-Part 210: Human-centred design Preliminary Results. In Proceedings of the for interactive systems. International 2013 Conference on Computer Supported Standardization Organization (ISO), 2010. Cooperative Work (CSCW ’13). ACM, New [12] Y. Karanam, L. Filko, L. Kaser, H. Alotaibi, York, NY, USA, (pp. 977–984), 2013. E. Makhsoom, and S. Voida. Motivational [22] A. Pommeranz, C. Detweiler, P. Wiggers, and Affordances and Personality Types in Personal C. M. Jonker. Self-reflection on Personal Informatics. In Proc. of the 2014 ACM Values to Support Value-sensitive Design. In International Joint Conference on Pervasive Proceedings of the 25th BCS Conference on and Ubiquitous Computing: Adjunct Human-Computer Interaction (BCS-HCI ’11). Publication (UbiComp ’14 Adjunct). ACM, British Computer Society, Swinton, UK, UK, New York, NY, USA, (pp. 79–82), 2014. (pp. 491–496), 2011. [13] I. Karatassis and N. Fuhr. Gamification for [23] D. Rajanen, M. Salminen and N. Ravaja. WebSAIL. In GamifIR@ SIGIR (pp. 15-20), Reflections on the Use of Psychophysiology in 2016. Studying Reading on Digital Media. In [14] J. M. Kivikangas. Emotion and Social Context Association for Information Systems AIS in Digital Game Experience (Doctoral Electronic Library (AISeL) IRIS Selected dissertation). University of Helsinki database, Papers Issue Nr 6, Paper 3. AIS, 2015. HELDA, 2015. http://aisel.aisnet.org/iris2015/3 [15] J. M. Kivikangas, G. Chanel, B. Cowley, I. [24] D. Rajanen and M. Weng. Digitization for Fun Ekman, M. Salminen, S. Järvelä and N. or Reward? A Study of Acceptance of Ravaja. A review of the use of Wearable Devices for Personal Healthcare. In psychophysiological methods in game Proceedings of AcademicMindtrek 2017. Tampere, Finland. ACM, 2017. [25] M. Rajanen and D. Rajanen. Usability benefits in gamification. In Proceedings of the 1st GamiFin Conference, Pori, Finland (pp. 87- 95), 2017. http://ceur-ws.org/Vol- 1857/gamifin17_p12.pdf [26] Y. Rogers, H. Sharp and J. Preece. Interaction design: beyond human-computer interaction, 2011. [27] R. Saha, R. Manna and G. Geetha. CAPTCHINO-A Gamification of Image-based CAPTCHAs to Evaluate Usability Issues. In Computing Sciences (ICCS), 2012 International Conference on (pp. 95-99). IEEE, 2012. [28] K. R. Scherer. Emotion as a multicomponent process: A model and some cross-cultural data. In P. Shaver, ed. Review of Personality & Social Psychology, Chapter 5 (pp. 37-63), 1984. [29] H. Schneider, K. Moser, A. Butz and F. Alt. Understanding the mechanics of persuasive system design: a mixed-method theory-driven analysis of freeletics. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems (pp. 309-320). ACM, 2016. [30] S. H. Schwartz. Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. Zanna (Ed.). Advances in experimental social psychology (Vol. 25),(pp. 1–65). New York: Academic Press, 1992. [31] K. Werbach and D. Hunter. For the win: How game thinking can revolutionize your business. Wharton Digital Press, 2012.