Increasing Engagement with the Library via Gamification Matthew Barr, Kay Munro, and Frank Hopfgartner University of Glasgow Glasgow, UK {firstname.lastname}@glasgow.ac.uk a positive and meaningful game-based experience that is closely connected to the underlying non-game set- Abstract ting (the library) then it will result in longer-term and deeper engagement between participants, non-game One of the main challenges faced by providers activities & supporting organisations. of interactive information access systems is to We argue that it is important to incorporate users’ engage users in the use of their systems. The context when providing gamified information systems. library sector in particular can benefit signif- In order to study this further, we present a prelimi- icantly from increased user engagement. In nary analysis of users’ engagement with the Library- this short paper, we present a preliminary Tree system [13], a gamified web application that aims analysis of a university library system that to increase engagement with library users by harness- aims to trigger users’ extrinsic motivation to ing gaming techniques to reward elements of library increase their interaction with the system. Re- behaviour and make interactions with the library more sults suggest that di↵erent user groups react fun. LibraryTree is operated by the library of a larger in di↵erent ways to such “gamified” systems. British university. The system was deployed over 12 months ago and is actively advertised on the library 1 Introduction website, as well as on posters and flyers that are dis- Many libraries currently su↵er from a decreasing num- played in the main library. ber of customers, threatening their main purpose to The paper is structured as follows. In Section 2, we serve as a provider of knowledge for mankind. Apart discuss related work. Section 3 briefly introduces the from the rise of the Internet as a challenging source LibraryTree system. Preliminary results of a transac- of information, an important factor that hinders users tion log analysis covering six months of user interac- from actively using their library is the lack of user- tion is presented in Section 4. Section 5 concludes the friendly graphical user interfaces. As shown in other paper and provides an outline of future work. domains where information access systems are de- ployed, engaging graphical user interfaces play a key 2 Related Work role in motivating users to engage with the content maintained by these systems. This work touches upon two main research topics, Kazai et al. [9] argue that a promising technique to namely the us of gamification to improve user expe- address this challenge is to adopt gamification. Gam- rience and the impact of context on users’ behaviour ification refers to the “use of game design elements while interacting with information access systems. In in non-game contexts” [5]. In fact, various studies the remainder of this section, we first provide an have been performed that showcase the benefit of gam- overview of gamification and then present related work ification, e.g., in the field of document annotation in the field of context-based retrieval and recommen- [19, 17], relevance assessment [11] or item recommen- dation. dation [10, 1]. Nicholson [14] argues that if users have 2.1 Gamification for Improving User Experi- Copyright c by the paper’s authors. Copying permitted for ence private and academic purposes. In: F. Hopfgartner, G. Kazai, U. Kruschwitz, and M. Meder Deterding et al. [5] suggest that “gamification is an (eds.): Proceedings of the GamifIR 2016 Workshop, Pisa, Italy, informal umbrella term for the use of video game ele- 21-July-2016, published at http://ceur-ws.org ments in non-gaming systems to improve user expe- rience (UX) and user engagement”, while Kapp [8] specific content, and the level of expertise of users. defines gamification as “using game-based mechanics, Bhavani et al. [3] report that di↵erent search strate- aesthetics and game thinking to engage people, mo- gies are required when retrieving domain-specific con- tivate action, promote learning, and solve problems” tent. Similarly, Meij et al. [12] study di↵erent language while it is ”not the superficial addition of points, re- models to improve domain-specific retrieval. Focusing wards, and badges to learning experiences.” The Li- on recommender systems, Zhang et al. [21] perform braryTree system described here aims to avoid such community topic mining to improve domain-specific superficial adornments, with rewards tied instead to recommendation. All studies are based on the assump- meaningful engagement with the library’s services, and tion that the domain to which documents or items be- particular attention paid to highlighting under-utilised long to is an important contextual factor that needs or little understood aspects of the library’s function. to incorporated when building an information access Kapp also points out that many, if not all, of the ideas system. now associated with gamification have been used suc- Focusing on users’ expertise in using search engines cessfully in some form or another in classrooms before as contextual factor, Halvey et al. [6] observed that this they were ever assembled under this umbrella. We, context plays an important role in a retrieval task. A and other proponents of gamification, would argue similar study is performed by Scott et al. [18]. Both that the tried-and-tested nature of these techniques studies indicate that users’ expertise needs to be con- actually demonstrates their utility - it is merely the sidered when developing graphical user interfaces to context in which they are being applied that is novel. support their information seeking task. This observa- A concern that is often overlooked when designing tion is also considered in the field of human-computer gamified systems is that of user preference and per- interaction where a specific focus is set on the di↵er- sonality, and the context in which the player interacts ence between novice and expert users. For a detailed with the system. Bartle [2], for example, famously survey, we refer to Cockburn et al. [4]. identified four types of personality, each with di↵er- Concluding from these studies, it is evident that ent motivations for playing the Multi-User Dungeon users’ context, e.g., their expertise or the domain they (MUD) games which Bartle pioneered. For a game to are interested in, directly influences their search be- appeal to all four player types (Killers, Explorers, So- haviour, thus indicating the need for context specific cialisers and Achievers), it must o↵er features that sat- information access systems (e.g., [7]). Consequently, isfy each of these various motivations. A game which treating gamification as a novel method to enhance appeals only to Killer type players is unlikely to appeal user interfaces of information access systems, we argue to players of the Socialiser type, for example. that further research is required to study the impor- There is also evidence to suggest that di↵erent tance of context for the development of a successfully genres of game appeal to di↵erent player personality gamified system. types. For example, Peever et al. [16] found strong relationships between game genre preference and per- 3 LibraryTree sonality types, as measured by the five-factor model of personality. While Park et al. [15] found no such In order to study the role of context for gamified sys- correlation between personality type and game genre tems further, we introduce the LibraryTree system. preference, they noted that players’ di↵erent motiva- The system allows students to gain points & badges tions for playing correlated with the personality traits (referred to as stamps) for entering the library build- associated with the five-factor model. ing, borrowing & returning books, accessing an e- Games, and by extension, gamified systems, are resource or sharing a review of an item they have read therefore likely to engage di↵erent users in di↵erent with friends and classmates. Users can share this in- ways, and to varying degrees. formation with their friends via a web interface or a mobile app. Figure 1 depicts the graphical user inter- face, the home page, of LibraryTree. 2.2 Users and Context While LibraryTree is available to all students, play- In recent years, various studies have been published ers must opt in to play the game. Registering to play that indicate that users’ context play an important the game is a simple process that utilises the univer- role in the way users interact with an information ac- sity’s existing user authentication system, making it as cess system. The definition of context, however, di↵ers straightforward as possible for students to register and based on the research questions or hypotheses that are subsequently log in. The registration process is kept as studied in literature. In this section, we focus on the brief as possible, but students are required to config- main contextual factors that are most important in the ure a range of privacy options that dictate the visibility context of library systems, namely access to domain- of their library interactions and LibraryTree progress. type - progress towards this goal is clearly indicated on the visual representation of the stamp, as shown in Fig- ure 3. Each stamp is associated with an appropriate number of points. For example, a player adding their first friend is awarded five points, while recommending ten items grants the player 25 points. Figure 1: The LibraryTree Graphical User Interface. Figure 3: Progression towards stamp achievement. Information relating to books and other items, library visits and LibraryTree stamps acquired may be made viewable by all players, or may be restricted to the 4 System Usage players’ friends or to the player only. Once registered, During the first six months of operation, 1751 play- players’ interactions with the library are automatically ers registered to use LibraryTree. During this pe- tracked by the LibraryTree system, requiring no fur- riod, 10072 stamps were awarded, the most commonly ther action on the part of the player. The balance awarded of which were related to physically visiting struck between ease of use and granularity of privacy the library building, or borrowing and returning items. control has, thus far, proved satisfactory. For example, 1323 players earned the “First of Many” A key component of LibraryTree is the progression badge for borrowing their first item, while 1118 players visualisation, displayed in the form of a tree - inspired were awarded the “Let’s have lunch” badge for visiting by that depicted in the university’s coat of arms - on the library building five times during lunch hours. the right hand side of the interface. As the player gains Engagement with LibraryTree (and, by corollary, points, the tree on their profile page is seen to grow. the library management system) varied significantly by Moreover, leaderboards show how the di↵erent colleges college, as illustrated by the points awarded to play- rank against each other. They show the overall points ers from each of the four colleges (Arts, Science and scored as well as weekly high scorers. Engineering, Social Science, and Medical, Veterinary In order to trigger users’ extrinsic motivation to in- and Life Sciences). A snapshot of activities is depicted teract with the system, LibraryTree allows users to in Figure 4. The mean number of points awarded to collect stamps and displays them on their profile page. students across all four colleges in this period was 173, Example stamps are depicted in Figure 2. The sys- with a median value of 60. Students based in the Col- tem supports a total of 107 stamps covering general lege of Arts and College of Social Science, on average, activities such as adding friends, rating books, or vis- scored significantly higher, with means of 268 and 235 iting the library building, but also more topic-specific points, respectively. The mean points awarded to stu- badges such as borrowing books associated with indi- dents in the College of Science and Engineering and vidual subjects. the College of Medical, Veterinary and Life Sciences were almost identical, at around 109. This disparity in apparent engagement with the Li- braryTree system can, in part, be explained by the nature of the subjects taught in each of the four col- leges. The study of the Arts and Humanities and the Social Sciences tends to rely more heavily on a broad range of textbooks and monographs, while the various Sciences are, perhaps, more concerned with lab work and online resources that do not require physical ac- cess to the library. These di↵erences exemplify the Figure 2: Examples of stamps awarded in LibraryTree. e↵ects of user context, of which the player’s College or Where a stamp is awarded based on a series of inter- discipline is an important aspect. actions - for example, borrowing five items of a certain The nature of individuals’ engagement with the Li- Figure 4: Weekly leaderboard and high scorers in LibraryTree. braryTree game is also variable, leading to speculation As future work, we will perform an in-depth anal- that users of the system may be aligned with one or ysis of the users’ behaviour when interacting with the more of the player types associated with traditional system. In particular, we aim to identify di↵erent con- video games, such as those identified by Bartle [2] and textual factors that influence user behaviour such as Yee [20]. In terms of Bartle player types, for exam- age and sex of the students or requirements to use the ple, one might classify those players with the highest library system to further their study. The primary scores as Achievers and those with the broadest range means by which these factors will be explored is an of stamps as Explorers. Socialisers would be expected online questionnaire, aimed at collecting data from as to have the greatest number of friends in the game, large a number of LibraryTree players as possible. The and to have obtained a higher proportion of stamps main motivation of our research is to study whether associated with friend interaction. The identification di↵erent player types, e.g., as introduced by Bartle [2] of Bartle’s Killers, however, may be more problematic and Yee [20] can be derived from their interaction. in a gamified system such as this, which limits the scope for ’griefing’ and other killer-like behaviour. In References this regard, Yee’s motivations for play may provide a more readily applicable scheme for classifying Library- [1] Sam Banks, Rachael Rafter, and Barry Smyth. Tree players. Yee’s components of motivation eschew The recommendation game: Using a game-with- ’killer’ type behaviour almost entirely, and subcompo- a-purpose to generate recommendation data. In nents of motivation for play that fall under the larger Proceedings of the 9th ACM Conference on Rec- ’Achievement’ component (e.g. competition) may bet- ommender Systems, RecSys 2015, Vienna, Aus- ter explain the motivations at work in LibraryTree. tria, September 16-20, 2015, pages 305–308, 2015. [2] Richard Bartle. Hearts, clubs, dia- 5 Conclusion monds, spades: Players who suit muds. In this short paper, we presented a preliminary anal- http://www.mud.co.uk/richard/hcds.htm, 1996. ysis on the role of user context of a gamified library [3] Suresh K. Bhavnani. Domain-specific search system. We analysed the transaction logfiles over a pe- strategies for the e↵ective retrieval of healthcare riod of six months to determine typical user behaviour and shopping information. In CHI ’02 Extended patterns, as well as to analyse acceptance of the system Abstracts on Human Factors in Computing Sys- by their users. The results suggest that while the sys- tems, CHI EA ’02, pages 610–611, New York, NY, tem has been broadly welcomed by the student body, USA, 2002. ACM. there is significant variation in the manner in which players interact with LibraryTree, in much the same [4] Andy Cockburn, Carl Gutwin, Joey Scarr, and way as players’ motivations to play traditional video Sylvain Malacria. Supporting novice to expert games varies. transitions in user interfaces. ACM Comput. In particular, we observed significant di↵erences be- Surv., 47(2):31:1–31:36, 2014. tween di↵erent students from di↵erent colleges of the university. We argue that this indicates that users’ [5] Sebastian Deterding, Dan Dixon, Rilla Khaled, context, e.g., the subject they are studying, influences and Lennart Nacke. From game design ele- the way they use a gamified system. ments to gamefulness: defining ”gamification”. In Proceedings of the 15th International Academic [14] Scott Nicholson. A user-centered theoretical MindTrek Conference: Envisioning Future Media framework for meaningful gamification. In Environments, MindTrek 2011, Tampere, Fin- Games+Learning+Society 8.0, Madison, WI, land, September 28-30, 2011, pages 9–15, 2011. 2012. [6] Martin Halvey and Joemon M. Jose. Bridging [15] Jowon Park, Yosep Song, and Ching-I Teng. Ex- the gap between expert and novice users for video ploring the links between personality traits and search. IJMIR, 1(1):17–29, 2012. motivations to play online games. Cyberpsychol- ogy, Behavior, and Social Networking, 14(12), [7] Frank Hopfgartner, Jinlin Guo, David Scott, 2011. Hongyi Wang, Yang Yang, Zhenxing Zhang, Li- juan Marissa Zhou, and Cathal Gurrin. Helping [16] Nicole Peever, Daniel Johnson, and John Gard- the helpers: How video retrieval can assist spe- ner. Personality and video game genre prefer- cial interest groups. In Advances in Multimedia ences. In Proceedings of The 8th Australasian Modeling (MMM), pages 493–495, 2013. Conference on Interactive Entertainment: Play- ing the System (IE ’12), 2012. [8] Karl M. Kapp. The gamification of learning and instruction: game-based methods and strategies [17] Massimo Poesio, Jon Chamberlain, Udo Kr- for training and education. Wiley, 2012. uschwitz, Livio Robaldo, and Luca Ducceschi. Phrase detectives: Utilizing collective intelli- [9] Gabriella Kazai, Frank Hopfgartner, Udo Kr- gence for internet-scale language resource cre- uschwitz, and Michael Meder. ECIR 2015 work- ation. TiiS, 3(1):3, 2013. shop on gamification for information retrieval (gamifir’15). SIGIR Forum, 49(1):41–49, 2015. [18] David Scott, Frank Hopfgartner, Jinlin Guo, and Cathal Gurrin. Evaluating novice and expert [10] Michael Meder, Till Plumbaum, and Frank Hopf- users on handheld video retrieval systems. In gartner. Daiknow: A gamified enterprise book- MMM’13, pages 69–78, 2013. marking system. In Advances in Information Re- trieval - 36th European Conference on IR Re- [19] Luis von Ahn and Laura Dabbish. ESP: labeling search, ECIR 2014, Amsterdam, The Nether- images with a computer game. In Knowledge Col- lands, April 13-16, 2014. Proceedings, pages 759– lection from Volunteer Contributors, Papers from 762, 2014. the 2005 AAAI Spring Symposium, Technical Re- port SS-05-03, Stanford, California, USA, March [11] Olga Megorskaya, Vladimir Kukushkin, and Pavel 21-23, 2005, pages 91–98, 2005. Serdyukov. On the relation between assessor’s agreement and accuracy in gamified relevance as- [20] N Yee. Motivation for play in online games. Cy- sessment. In ACM SIGIR 2015, pages 605–614, berpsychology and Behavior: The Impact of the 2015. Internet, Multimedia and Virtual Reality on Be- havior and Society, 9(6):772–775, 2006. [12] Edgar Meij, Dolf Trieschnigg, Maarten de Rijke, and Wessel Kraaij. Conceptual language models [21] Xi Zhang, Jian Cheng, Ting Yuan, Biao Niu, and for domain-specific retrieval. Inf. Process. Man- Hanqing Lu. Toprec: Domain-specific recommen- age., 46(4):448–469, July 2010. dation through community topic mining in social network. In WWW’13, pages 1501–1510, New [13] Kay Munro and Ciaran Talbot. Librarygame – York, NY, USA, 2013. ACM. evaluating gamification as a means of increasing customer engagement. In 38th ELAG Conference, 2014.