PERSONALITY BASED GAMIFICATION: HOW DIFFERENT PERSONALITIES PERCIVE GAMIFICATION Research in Progress Codish, David, Ben-Gurion University of the Negev, Beer-Sheba, Israel, [email protected]
Ravid, Gilad, Ben-Gurion University of the Negev, Beer-Sheba, Israel, [email protected]
Abstract Personality is known to moderate task performance, learning styles, and gaming preferences. With the inclusion of game elements into non-game contexts such as education and workplaces, it is essential to develop models that will help explain and predict the influence of personality in environments that are, unlike the existing research, both ludic and utilitarian at the same time. In this study which is the first in a series of studies, we apply game elements to an undergraduate information systems course and examine the students perception of playfulness (n=102). Students completed a five factor personality test and answered questions about their preference for specific game elements and their overall playfulness from the gamification applied. T-tests and Partial Least Squares (PLS) analysis were performed comparing low and high measures of each personality trait revealing a higher preference level for badges by introverts and higher levels of preference from progression by agreeable personalities. Several significant differences in the overall playfulness were found in the relations of the path model and are presented. The significance of these results comes from the understanding that playfulness from game elements is not always granted and depends among other factors on personality. Keywords: Gamification, Personality, Game mechanics, Playfulness
The inclusion of hedonic elements in utilitarian information systems is becoming commonplace as a means of engaging users and increasing system acceptance (Huotari, Hamari 2011, Deterding et al. 2011b, Zichermann, Cunningham 2011). Traditionally, hedonic and utilitarian information systems were treated and researched as separate entities (Van der Heijden 2004, Brown, Venkatesh 2005, Lowry et al. 2012) but in the past years they are converging into a field called gamification which is defined as the use of game design elements in non-game contexts such as, but not limited to, workplaces (Deterding et al. 2011a). Utilitarian systems are focused on ease of use and usefulness (Davis, Bagozzi & Warshaw 1992) whereas hedonic systems focus also on enjoyment, curiosity, and immersion (Lowry et al. 2012). The combination of these two distinct system characteristics requires finding the fine line between them; increasing system usage through hedonic motivation while maintaining the utilitarian nature of the system. System acceptance and usage has been a key focus in IS research and literature. Theories of system acceptance such as the Technology Acceptance Model (TAM) (Davis 1989) and its enhancements (Venkatesh, Thong & Xu 2012, Venkatesh et al. 2003) are rooted in psychology
based theories such as the Theory of Reasoned Action (TRA) and the Theory of Planned Behavior (TPB) (Ajzen 1985). These theories propose a model for predicting behavior and pose that the attitude toward the behavior, subjective norms, and perceived behavior control are the key antecedents to the behavior intention and the behavior itself. TAM is based on TPB and defines ease of use and usefulness as the antecedents for attitude with usefulness being the dominant factor. Subjective norms of behavior control were not included in the original model but were later included in the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et al. 2003). These theories focus on work-related systems where usefulness and performance are important. In a study on acceptance of hedonic systems, ease of use and enjoyment have been shown to be more dominant than usefulness (Van der Heijden 2004) indicating that people indeed relate to hedonic and utilitarian systems differently. A study combining hedonic and utilitarian systems in the consumer application domain (Venkatesh, Thong & Xu 2012) has taken the UTAUT and updated it with elements from the hedonic system acceptance model showing that hedonic motivation has a significant effect on behavior intention. However, gamification in work-related environment may have different characteristics than consumer applications thus the model proposed in the updated UTAUT requires further validation. The gamification business is expected to reach $5.5B in 2018, up from a mere $100M in 2011 (marketsandmarkets.com 2013), and Gartner is predicting that gamified services for consumer goods marketing and customer retention will become as important as Facebook, eBay or Amazon and will be implemented by more than 50% of companies (Gartner 2011, Gartner 2012). At the same time, Gartner predicts that 80% of gamification initiatives will fail due to poor design and implementation. Game elements included in games are also referred to as game mechanics and dynamics. Game mechanics are defined by the gamification wiki as "constructs of rules and feedback loops intended to produce enjoyable gameplay. They are the building blocks that can be applied and combined to gamify any non-game context" (Gamification.org Wikipedia 2012). Most common game mechanics are Points, Badges, and Leaderboards (PBL) (Werbach, Hunter 2012, Zichermann, Cunningham 2011, Narasimhan, Chiricescu & Vasudevan 2011, Antin, Churchill 2011) but there are many additional mechanics (SCVNGR 2010) that exist in games and can be designed into systems and processes. Dynamics are the run-time behavior of the mechanics acting on player inputs and each other's outputs over time (Hunicke, LeBlanc & Zubek 2004) and can be viewed as the pattern of play that is generated by the application of specific mechanics and in response to other player interactions or expected interactions (Brathwaite, Schreiber 2009). For example, the use of a leaderboard mechanic in a game might generate a competition dynamic between the players whereas the use of a chat box and user profile mechanics may generate a cooperative dynamic. Dynamics cannot be programmed into a gamified solution but the use of the right mechanics can improve the chance of the dynamic occurring. Typical dynamics found in games are constraints, emotions, narrative, progression, and relationships (Werbach, Hunter 2012). There is an increasing number of case studies and research dealing with gamification in different context as shown by the meta-analysis done by (Hamari, Koivisto & Sarsa 2014). While the majority of these studies report overall positive outcomes as a result of adding game elements, not all have exhibited these results. Differences between studies can be explained by design and context, but even within studies there were differences in how different individuals were impacted by the gamification which can be explained by personality differences (Hamari, Koivisto & Sarsa 2014, Hamari 2013). The claim of our study is that although overall positive results may be achieved, different people will be motivated differently and perhaps even negatively by the inclusion of game elements as seen in a study comparing the effect of gamification in a mandatory and voluntary setting (Mollick, Rothbard 2012). Our study will try
to also answer the question of how exactly these different gamification mechanics are perceived by people with different personalities. Some personality theories that focus on personality types while others focus on needs. In this study we chose to focus on the Five Factor Model (FFM) which has been proposed as trait theory (McCrae, Costa 1989) and is widely acceptable. FFM posits that individuals have the following five traits: neurotism, extraversion, openness, agreeableness, and consciousness (McCrae, John 1992). Using personality which is a relatively stable measure to explain behavior has advantages over existing theories mostly based on perception (McElroy et al. 2007) and has been used in studies aimed at explaining a behavior such technology acceptance (Devaraj, Easley & Crant 2008), internet usage (Swickert et al. 2002, Landers, Lounsbury 2006, McElroy et al. 2007, Ross et al. 2009) and user interface preferences (Nov et al. 2013). There is a plethora of literature on the moderating effects of personalities on work oriented task performance (Barrick, Mount & Strauss 1993, Colbert et al. 2004), learning styles (Kolb, Kolb 2005, Furnham 1992, Furnham, Jackson & Miller 1999, Yeung, Read & Schmid 2012), and gaming preferences (Holbrook 1984, Teng 2008, Yee 2006). These studies highlight the importance of personality traits in daily tasks such as working and learning, as well as leisure activities such as computer game playing. The majority of these studies examine personality with relation to a specific domain with the exception of Furnham, Jackson, and Miller who looked at training combined with work activities. As gamification combines the worlds of work and play, there is a missing body of research on the interaction of these domains with personality. This study contributes to this missing link by understanding how different personalities perceive playfulness in a non-game situation such as an academic course, using game elements. Once these relations are understood, information systems can be designed and developed to address these findings and optimize system acceptance by ensuring maximum engagement based on user personality. Our research questions are 1) how do specific game elements influence playfulness, 2) what are the relations between these game elements, and 3) how does personality moderate these relations? The model tested (Figure 1) includes three feedback mechanics: points, rewards, and badges, and two presentation mechanics: leaderboards and progress bars. Points are operationalized as the preference to receive feedback in the form of grade points as opposed to verbal qualitative feedback, rewards are the motivation derived by the existence of a physical rewards and badges are operationalized as the desire to receive virtual rewards. The difference between the two feedback presentation mechanics is that leaderboards are a form of comparative feedback and progress bars refer to feedback against a personal goal. The feedback mechanics of points, rewards, and badges are the atomic level of feedback (i.e., for a specific action points are granted). Feedback is an essential element of games and is present in any gamified system. However, the way this feedback is presented may vary. In some cases, the feedback is presented in progress bars and in some in leaderboards, which is not to say these are the only feedback or presentation mechanics available. Rewards are the only physical and tangible form of feedback and is what students are familiar with. The relations between points and badges which are virtual to rewards, are aimed at assessing how closely students relate virtual feedback to something physical. The way feedback is received and presented influences the perceived playfulness of the entire system. Different personalities will perceive playfulness differently for any gamification implementation thus the moderating effect.
The ultimate goal of this research is to apply gamification into a utilitarian IS context and analyze the effect it has on different personalities. As a first step, gamification was applied to an academic course using a pen and paper information system which is easier to apply, requires
less technical resources and is more controllable. Results from this first step, will serve as the basis for the automation and system development of a fully automated information system. In this first step, data was collected from undergraduate students in their third year of four in the software analysis and design course, with more than 95% of the students majoring in industrial engineering and management. Throughout the course, students are required to complete a project in which they select an organization of their choice, identify a business problem and go through the analysis, design and development phases for their selected organization. The gamification of the project was done in such a way that students had to hand their projects over to a random team and receive a project from a different team. Doing so, mimics a real life situation where a single person typically is not responsible for a project from analysis through development. Students received feedback from the course staff in the form of grades (points) and verbal and written (badges) recognitions in the form of appraisal letters from the teaching staff. Teams performing extra work received extra bonus points (rewards). Four surveys were administered throughout the course at different phases and which were voluntarily completed in exchange to half a credit point for each survey completion. Personality was measured using the revised FFM questionnaire (Goldberg 1992) which includes 50 items. Each dimension was divided into high and low based on the mean value for that measure since there is no guidance as to what accounts for a high or low measure. Game mechanics preferences were measured based on a developed questionnaire. Finally, perceived playfulness was measured using a nine item scale adapted from Moon and Kim measuring the state of playfulness derived from the specific situation (Moon, Kim 2001). Perceived playfulness is a multi-dimension construct including focus, curiosity, and enjoyment from an activity and is distinct from the playfulness trait which is context independent as appears in other studies (Atkinson, Kydd 1997, Webster, Martocchio 1992).
Partial Least Squares (PLS) structural equation modeling was selected as the approach to test the model. This selection of PLS is a) due to the exploratory nature of the research and the relatively small sample size (Henseler, Ringle & Sinkovics 2009), and b) due to the interaction nature of the model which is best tested with PLS methods (Endler, Parker 1992). The software used was SmartPLS Version 2.0M3 (Ringle, Wende & Will 2005) which is freely distributed. Descriptive statistics of the measured values are presented in Table 1. The course included 133 students; 102 completed all four surveys and were included in final analysis (76.7%). Of the students responding, 58 were female and 44 were male. Students were all in the age range of 2330 years.
Points Leaderboard Progress Badge Playfulness Reward Extraversion Agreeableness Conscientiousness Emotional stability Intellect
1.00 1.33 2.20 1.50 1.00 1.00 16.00 28.00 20.00 15.00 13.00
5.00 5.00 5.00 5.00 4.00 5.00 49.00 50.00 47.00 50.00 29.00
2.37 3.35 3.95 3.16 2.29 3.40 35.01 41.16 35.78 33.31 21.94
1.05 0.92 0.64 0.87 0.80 0.76 7.01 4.84 5.90 7.56 3.32
Internal and convergent validity indices have been examined for the full model. All Cronbach alpha values are above the desired 0.7 index with the exception of the points construct that is 0.67 and is deemed acceptable by us due to the explorative nature of this study. To test the first proposition, a t-test was performed on the game mechanics across each personality trait comparing the means of the low and high personality trait. In all cases but two (Table 2), there was no significant difference.
Personality Extraversion Agreeableness
Mechanic Mean Low High Low High
T-test for game mechanics mean differences by personality
To measure how such a model would behave, we first tested it on the entire user base (n=102). We later, examined the model twice for each personality type using the low and high values of that personality type in our sample. A comparative table showing the total effects between the mechanics and playfulness for each model are presented in Table 3. A few things to notice are the impact of points on playfulness between low and high conscientiousness personalities that shift from strong positive to strong negative, the negative impact of leaderboards on extroverts and the lack of evidence that leaderboards have a positive effect on playfulness. The significance of the differences between low and high personalities was tested through a multi-group analysis t-test (Andreev et al. 2009, Keil et al. 2000). The results of this test are presented in Table 4 and highlight some areas where group differences exist. The analysis is conducted as a comparison between groups regardless of the significance of the total effect simply indicating there is a difference. Last, the predictive relevance of the different models was tested using blindfolding method in SmartPLS. ܳ ଶ values were calculated for each of the models and the results are summarized in Table 5. ܳ ଶ values above 0 indicate predictive relevance; values above 0.02 provide low predictive relevance, values above 0.15 provide medium relevance and above 0.35 values are considered high. In most cases, the predictive capability of the models is low thus the models should not be used as predictive models.
Extroversion L H 0.63 0.42 *** *** 0.14 ** 0.66 0.39 *** **
Agreeableness L H 0.54 0.47 **** **** 0.16 ** 0.52 0.59 **** **** 0.25 **
Leaderboard Progress Playfulness
0.25 ** 0.20 **
0.33 ** 0.14 **
0.35 ** 0.15 ** -0.17 * 0.43 ***
0.23 ** 0.09 **
0.49 **** 0.16 ***
Leaderboard Progress Playfulness
Leaderboard Playfulness Progress
Conscientiousness L H 0.5 0.44 **** **** 0.20 **** 0.53 0.55 **** ****
0.15 * 0.42 **** 0.36 **** 0.26 ***
Intellect L H 0.6 0.44 **** **** 0.2 *** 0.53 0.53 **** **** 0.08 ***
0.36 **** 0.31 ***
0.37 **** 0.15 *** -0.26 *** 0.41 ****
Emotional Stability L H 0.60 0.40 **** **** 0.34 **** 0.44 0.6 **** **** 0.11 *
0.48 **** 0.16 *** 0.21 **
0.38 *** 0.15 ***