Chee Siang Ang & Panayiotis Zaphiris SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
A social network analytic perspective
In this study, we investigated the social roles that emerged from the users’ behaviour and interaction (as opposed to pre-defined formal roles) within a virtual community of massively multiplayer online role playing game. Adopting the social network analysis perspective, we identified and analysed three major social roles with distinct interaction styles and network properties. It was found that core members of the community were players who contributed to the community expansion by providing help and fostering group cohesiveness through friendly social interaction. On the other hand, some periphery members were freeloaders whose sole purpose of joining the guild was only to utilize the guild resource. Keywords
MMORPG; SNA; social roles; virtual community
(Received 8 March 2009; final version received 18 August 2009)
Introduction Like its real life counterpart, a virtual community does not consist of only homogeneous people with similar interaction patterns. For instance, some people contribute to the growth of the community by actively providing help to newcomers, and others by creating a friendly atmosphere and hence a sense of belonging through friendly behaviour. Some people, on the other hand, operate at the periphery, enjoying observing the community activities and only engaging in the community activity very rarely, if any. From the design point of view, such diversity is one of the key factors for a virtual community to be successful (Anderson 2004), and thus the challenge lies in the sociability design of tools/social policies that can facilitate and coordinate various types of interaction styles. We believe that this can be achieved by Information, Communication & Society Vol. 13, No. 4, June 2010, pp. 592 –614 ISSN 1369-118X print/ISSN 1468-4462 online # 2010 Taylor & Francis http://www.tandf.co.uk/journals DOI: 10.1080/13691180903266952
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
analysing how different groups with different social roles work together. This allows us to understand how their diverse interactions can be leveraged. Although quite a considerable amount of literature exists in this area, most studies have focused mainly on the attributes (e.g. age, gender, education level, number of messages they post, etc.) of the individuals, groups or communities as the basis of analysis. In this study, our focus was on the relational level of interaction between individuals and groups, i.e. who interacts with whom. We analysed social roles in a massively multiplayer online role playing game (MMORPG), namely, World of Warcraft (WoW) (Blizzard Entertainment 2004), by focusing on the structure and relation of these individuals, through the social network analysis (SNA) perspective. MMORPGs are a particularly interesting virtual community to study due to their explicitly pronounced role-play element. The main aim of this study was to investigate the social roles in the community of WoW. Specifically, we aimed to: .
. .
identify social roles that emerge from individuals’ interaction in the community; examine the interaction styles and network properties for the social roles; identify the relationship between different social roles;
A social role allows the person holding it to enjoy some forms of privileges, and at the same time, it constrains his/her behaviour. Therefore, categorizing social roles is important in order to understand and to a certain extent predict the behaviour of the people holding the roles.
MMORPG as virtual communities Playing games is essentially a social experience, and it is argued that it is impossible to play a game in isolation in a meaningful sense (Ang et al. 2005). With the inception of the Internet in games, the scope of social interaction has expanded to involve a large number of players playing simultaneously through the use of diverse computer-mediated communication (CMC) applications. Perhaps, the most prominent example of social game play is MMORPGs. These virtual worlds represent the persistent social and material world, which is structured around narrative themes (usually fantasy), where players are engaged in various activities such as slay monsters, attack castles, scavenge goods, trade merchandize, etc. Therefore, MMORPGs can be seen as a virtual community that takes place in a three-dimensional virtual world. Virtual communities have become an important topic of research that attracts scholars from different disciplines (Anderson 1991; Wellman 1997). Various definitions of virtual communities have been proposed, and we find Rheingold’s (1993) definition particularly relevant in providing a background to our study:
593
594
INFORMATION, COMMUNICATION & SOCIETY
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
[virtual] communities are social aggregations that emerge from the Net when enough people carry on those public discussions long enough, with sufficient human feeling, to form webs of personal relationships in cyberspace. (Rheingold 1993) In the context of MMORPGs, users’ activities go beyond simply public discussions and include a whole range of collaborative activities. Therefore, another definition, known as communities-of-practice, is useful in linking MMORPGs to virtual communities. From the perspective of communitiesof-practice, it is assumed that the essence of a community lies in the common activities. In other words, it is the common actions and procedures that make a community (Lave & Wenger 1991). Participation in MMORPGs is characterized by the shared experience and the collaborative nature of players’ activities (Jakobson & Taylor 2003). Indeed, several scholars have begun analysing MMORPGs as a type of virtual communities (Bruckman 1997; Muramatsu & Ackerman 1998; Ducheneaut et al. 2004; Steinkuehler 2004; Ang et al. 2007; Yee 2005). Understanding the patterns of interaction in these game communities is important for design, as these virtual communities function as a major mechanism of socialization of the players. Ducheneaut et al. (2004), for instance, discussed in depth how observing and analysing social interactions in an MMORPG, Star War Galaxies (Sony Online Entertainment 2003), can lead to better sociability designs of such playful communities. One of the most evident examples is the concept of ‘guilds’. Guilds are a fundamental component of MMORPG culture for players to run a virtual association, which has formalized membership and rank assignments to encourage participation. Each guild usually has a leader, and several guilds could team up in a battle. This involves complicated leader– subordinate and leader – leader relationships. Therefore, divisions of roles tend to arise to mediate the coordination of these complicated collective actions.
Social roles Just like in physical life, people spend much of their time online in groups and often take different positions within the groups. These positions, usually known as social roles, are a set of expectations and norms that define how people holding the roles should and would behave in a social situation (Stark 2007). Ashforth (2001, p. 29) defined that ‘a role is a set of prescriptions defining what the behaviour of a position member should be’. Roles have been traditionally seen as an important element in explaining the relationship between individual and society or between person and social system. For instance, Mead (1934) assumed that a social system is composed of
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
interactions, and these interactions result in the development of social roles. This symbolic interaction approach emphasizes that roles are formed on the subjective will of the individuals. The functionalistic perspective, on the other hand (Parsons 1951), stressed that society determine roles. This approach suggests the existence of objective role structures that determine the individuals’ behaviour. Fundamentally, social roles put people in a certain position, which is linked to tasks and functions. Based on Herrmann et al. (2004), a role has the following four characteristics: . .
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
.
.
position: the social strata in a society or the hierarchy in an organization; function/tasks: roles imply special functions and tasks in the form of explicit and documented expectations and responsibility; behaviour expectations: implicit and informal agreements of individual behaviours, which are expected of the roles; social interaction: roles are the result of a social interaction and negotiation between the actor and those with whom he/she interacts.
This phenomenon can also be observed in virtual communities. Online users assume different social roles such as administrators, lurkers, contributors, etc. These roles shape the tasks and functions of the social position thus what the individuals should do. Apart from these, users participating in virtual communities should also follow certain norms (i.e. to behave in some expected ways) such as non-verbal communication like the emoticons. It is important to study social roles because they can explain how people interact, collaborate and work together to cultivate community building and growth. It has also been found through experiments that social roles increased the participants’ awareness of their social interaction and thus the efficiency of group tasks (Strijbos et al. 2003). Furthermore, the forming of communities, which share interests, values and most importantly common activities, is positively influenced by social roles. Social roles can have a significant influence on the structure, coordination and support of the community (Herrmann et al. 2004). Basically, at least two types of social roles can be identified: formal and informal (Golder & Donath 2004). Formal social roles are pre-defined and clear-cut divisions of roles, which are defined by official, legal, governmental and employment means. Some examples include the role of the prime minister, lawyer, manager, teacher, etc. These roles are usually obvious and explicitly ascribed to the people holding the roles. Individuals with specific roles have certain sets of skills, responsibilities and expectations. Furthermore, we can also presume different interaction styles for individuals interacting with people with different roles. In some cases, these responsibilities, functions and expectations of the social roles are formalized and legitimized into rules and norms. On the other hand, informal roles are not well defined, ambiguous and flexible. These roles are usually unspoken, and they ‘emerge’ through interaction, negotiation and agreement. For instance, in a group of teachers
595
596
INFORMATION, COMMUNICATION & SOCIETY
(formal role), one might emerge as the ‘organizer’ (informal role), who is particularly good at planning field trips. Unlike formal roles, people do not tend to have explicit expectations on informal roles, although in many cases, they gradually become ingrained into the community. For example, the teacher who planned a couple of field trips successfully might be expected to also organize the next one. The major difference between formal and informal roles is that formal roles result in the elicitation of certain behaviours (e.g. the teacher is responsible for teaching), while informal roles are the result of the antecedent behaviour (e.g. the teacher is recognized as a good organizer because of his/her ability to organize).
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Social roles in MMORPGs In the case of MMORPGs which are essentially a form of role play, the emergence of (formal and to some extent informal) social roles is explicitly facilitated by the games. Formal roles can be easily observed since the players are asked to choose a class (e.g. warrior, mage, hunter, etc.) at the outset. Each class is designed to have some skills and limitations. For instance, a mage has the privilege of ranged attack, but is extremely vulnerable to physical attack. Conversely, a warrior has high defence but can only perform proximity attack. Informal roles, although not clearly pronounced, can also be identified. Within the MMORPG community, some common informal roles have emerged and formalized into rules and even written in the official game manual. ‘Tank’, for instance, refers to a role that engages in melee attack with the enemies and withstands damage. ‘Puller’ refers to the player who is responsible to draw a monster to the player group. Most MMORPGs feature a mode of play known as the guild. A guild is a formal organized system, in which tasks and formal roles are normally assigned explicitly to the players. During the social interaction, however, informal roles are developed, assigned or taken over. Although communities usually have only few formal roles but a lot of dynamically developing informal roles, the stability of a guild community is based on explicit formal roles and less on informal, emerging roles. In this kind of setting, informal roles are pretty much influenced by knowledge and expectations the players already have about each other through existing formal roles. As Smith (1966) noted in observations of an Antarctic work group, informal or emergent roles will often replace or supplement more formal roles.
Social Network Analysis (SNA) approach to social role research An early study on text-based multi-user dungeons (often referred to as the predecessor of MMORPGs) has been conducted by Bartle (1996) in order to
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
identify the typology of roles. His typology consists of four categories: killers who like to annoy other players, socializers who like casual social interaction with other players, achievers who aim to master the game and explorers who enjoy exploring the game world. Newer research has also been undertaken in the context of graphical MMORPGs by Yee (2005), and he identified six categories of play (relationship, grief, immersion, escapism, achievement and analyse). It is noticed that most of the previous work in this area assumed the ‘conventional’ social science analytical perspective. In other words, they studied ‘who the users are’ (gender, age, etc.) and ‘what the users do’ in the community rather than the relations between them. For instance, most studies focused on the contents of the users’ posting or activity. Although such approaches have been useful, we believe that adopting an SNA (Wasserman & Faust 1994) standpoint, focusing on relations (with whom the users interact) and structures, would cast new light onto social roles in virtual worlds. The importance of structural and relational analysis is particularly pronounced in researching social roles as people obtain their roles through multiple social networks in which they are involved (Wellman & Berkowitz 1994). The formal role as a teacher only takes on its meaning in relation to the students; while the informal role as an organizer to the peers in the group. Similarly, the formal role as a warrior has no meaning outside the MMORPG environment; while the informal role as a ‘tank’ can only exist in relation to other roles such as ‘caster’ (someone who casts magic to the enemies while the tank receives damage). In a nutshell, a role is relational to the group with whom one interacts. Although rare in the context of play, the interest of social roles in SNA is gradually growing in computer-supported collaborative work. Koku and Wellman (2004), for instance, examined a scholarly network using block-modelling (White et al. 1976) to identify role structures in the community. They found four blocks of users: managers, scholars, experts and periphery members. Haythornthwaite and Wellman (1998) in their study about work, friendship and media use for information exchange identified three types of emerging roles in a university research group: work-oriented, computerprogramming-oriented and social-oriented. Similar work has also been done in CMC through the SNA perspective. Turner et al. (2005), for example, investigated the hierarchies and social networks of a Usenet newsgroup, in which they attempted to describe the network properties of each role through relational and structural analyses. Following this, Welser et al. (2007) carried out a confirmatory analysis on a specific role of ‘answer people’ with logistic regression in order to predict the likelihood of a user being an answer person based on their network structures. In the study reported in this paper, we examined social roles of guild community in an MMORPG through SNA, complemented by traditional thematic and content analysis.
597
598
INFORMATION, COMMUNICATION & SOCIETY
Methods
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Data collection In this study, we analysed a guild community in WoW. Before collecting the data, we dwelt in the game for a week to familiarize ourselves with the general practice and culture around the game. Once we became comfortable with the game, we joined a relatively large guild (with 76 members at the time of observation). We chose to study a guild community because of its clear-cut boundary of actor membership, thus enabling us to collect whole network data (Wasserman & Faust 1994). Using the in-game chat-log function, we managed to keep a record of guild members’ chat activities throughout the study. We also observed the game activities and kept a field note in order to reflect on our observations. A total of 1944 guild messages were collected in 30 hours of observation. Interactions in MMORPGS are typically rich and dense; therefore, 30 hours of observation gave us substantial amount of data for analysing the structure of social roles. In addition, we believed that the guild we chose for our analysis was a mature and stable one as the structure of formal roles within the guild had been established and the number of members had stabilized. This stability was needed to enhance the validity of our results. Then, we categorized the messages using thematic and content analysis (Krippendorff 1980). An initial set of categories was identified through the field note. Using these initial categories, the data were examined in detail to revise and strengthen the categories. This process identified seven types of interactions through which guild players were related to each other: . group management: activities that are related to organizing the groups including soliciting/responding to invitations, identifying group members, deciding meeting points, (re)structuring the group and leaving the group; . coordination: task coordination, involving coordinating group member activities towards task completion such as pointing out targets, coordinating actions, looting, discussing strategies and trading; . ask for help: asking for help or asking questions. It usually involves situations that need immediate solutions such as the solution of quests, equipments, asking for game items and money; . give help: giving help and answering questions. It could also be attempts to help/answer questions; . friendly remarks: messages that express friendliness towards others such as apologizing, greeting, laughing, being polite (e.g. saying thank you), smiley and non-verbal communications such as ‘dancing’; . game chats: casual chats that are not directed towards achieving game goal or completing game task. These include small talks, talking about WoW/other games, telling jokes, being sensitive to others and being supportive and encouraging;
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
.
real life chats: chats that reveal the member’s real life identities (real life gender, nationality, etc.) and chats about real life topics such as their work/college life.
In order to validate the category, a focus group was run with three experts in this area. We started by explaining the category scheme to them. Then, each of them was required to apply the scheme to check whether it was able to explain all the data. Feedback was gathered and any discrepancy was discussed and resolved in order to come up with a robust category set. Using these categories, we coded all the messages. In order to establish inter-coder reliability, we used a second coder. Inter-coder reliability was 0.7069 using Cohen’s kappa coefficient (Cohen 1960), indicating substantial agreement between the two coders (Landis & Koch 1977). We then tabulated the messages into socio-matrices for SNA by identifying ‘who talked to whom’ relationships in the observation data. Seven socialmatrices were generated representing the networks for each interaction type. We wanted to study social roles based on the similar pattern of interaction the players were engaged. Therefore, an affiliation matrix was constructed. The affiliation matrix shows the interaction type each individual was engaged with. The affiliation matrix was then transformed to obtain a square player – player matrix. Note that unlike conventional socio-matrix, this matrix does not show the ‘who-talk-to-whom’ relationship. Instead, each cell contains the ‘minimum number of interactions of the same type’ two players were engaged in. For instance, two players might be engaged in two interactions of the same types, but they did not necessarily interact with each other. The minimum method examines the entries for two actors at each activity type and selects the minimum value to sum up. This approach is commonly used when the data are measured as valued. The tie between the two actors is equal to the weaker of the ties of the two actors to the activity type. In this way, we can ensure that both actors have at last the minimum number of shared activities under each activity type.
Data analysis The analysis was carried out in three phases. At the first phase, we explored different techniques of block-modelling in order to identify categories of players with different structural positions to infer their social roles in the guild community. Convergence of Iterated Correlations (CONCOR) (Breiger et al. 1975) block-modelling procedure was carried out on the transformed affiliation matrix. It will be discussed in-depth in the finding section. As aforementioned, our approach is relational. Therefore, in identifying social roles, we did not impose the players’ position based on their attributes. The most commonly used SNA technique for this purpose is CONCOR. It
599
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
600
INFORMATION, COMMUNICATION & SOCIETY
has been applied successfully in various contexts of studies including learner’s roles in online community (Koku & Wellman 2004), online discussion boards (Kuwashima 2006), etc. However, due to the obscure mathematical properties of the CONCOR method, it has been advised that it should be used with great caution. Therefore, we employed an iterative approach to make sure that our solution yields the results with sound interpretations. More specifically, we run a number of CONCOR analyses on the data (i.e. with two, four and eight clusters) and then we examined the activities patterns for each separate CONCOR analysis. We found that CONCOR analysis with four clusters yielded the best interpretation. Furthermore, we also obtained the best r2 value with four-cluster CONCOR analysis. Block-modelling categorized players into different blocks according to their structural equivalence. It does not, however, tell us the ‘content’ of the interaction. Therefore, at phase 2, the interaction styles and ego-network properties of each block were investigated in order to further understand these blocks. At phase 3, we compared different blocks and identified the intra-relation within and inter-relation between blocks. At the end of this phase, a model of interaction was developed based on the results obtained from the analysis. SNA software packages, UCInet (Borgatti et al. 2002) and SIENA (Snijders et al. 2007) in StOCNET (Stokman et al. 2004), were used to assist the analysis.
Results and findings Table 1 shows the general network properties for the overall interaction network and the networks for each interaction type. TABLE 1
General descriptions of the guild network. no. of actors
no. of observed ties
density (SD)
no. of messages
Overall interaction
74
457
0.0869 (0.2818)
1902a
Ask for help
40
52
0.0333 (0.1795)
250
Give help
60
117
0.0331 (0.1788)
279
Friendly remark
63
258
0.0661 (0.2484)
509
Game chat
43
123
0.0681 (0.2519)
441
Real life chat
16
12
0.0500 (0.2179)
36
Group management
32
62
0.0625 (0.2421)
274
Coordination
16
24
0.1000 (0.3000)
113
a
This number is lower than the total number of messages collected (1944) because some
messages were categorized as miscellaneous and were not included in the analysis.
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
About 85.14 per cent of the players were involved in ‘friendly remark’ interaction and 81.08 per cent in ‘give help’ interaction. Only 21.62 per cent chatted about real life topics and was engaged in coordination. The density for all interaction types was relatively low, indicating that not everyone was directly connected to each other. ‘Ask for help’ and ‘give help’ had particularly low densities (0.0333 and 0.0331, respectively), and we believe that this was mainly because players often asked for help from and gave help to a specific set of people, causing them to be connected to only a certain group of people rather than to everyone. Table 1 also shows that most of the interaction types focused on ‘friendly remark’ and ‘game chat’ and a considerable amount of interaction was about ‘ask for help’, ‘give help’ and ‘group management’. In the following sections, we present the results in three parts, each addressing the research questions: types of social roles, network properties of social roles and interaction of roles.
Results 1: types of social roles We carried out the CONCOR block-modelling procedure on the player – player matrix, which we obtained by transforming the affiliation matrix. CONCOR is a top-down block-modelling technique, and it partitions the players into non-overlapping mutually exclusive sub-groups. In a conventional ‘who-talk-to-whom’ socio-matrix, partitioning is based on the similarity of choice made and received by all players. In the transformed affiliation matrix used in this study, however, partitioning is based on similar patterns of interaction the players have with each other within the same block and with players from other blocks. In other words, players who tend to exhibit similar behaviour towards each other within the group and players in other groups are blocked together. CONCOR produced a four-block partition of the guild players (R2 ¼ 0.409). One of the blocks consisted of only two isolate players and was thus excluded from further analysis. For the sake of readability of the paper, let us now call the three blocks produced by CONCOR, blocks A, B and C. The partition resulted from the CONCOR block-modelling was used to extract three sub-graphs from the ‘who-talk-to-whom’ socio-matrix in order to examine their general network properties. Figure 1 summarizes the three blocks. We can see that block A was internally highly dense, while block C was almost not connected at all internally. It is also quite interesting that 24.4 per cent of the players were connected directly to each other in block A, and this block also constituted 50 per cent of the ties of the whole network. It is worth mentioning that Figure 1 shows only ties within the blocks, and there were a total of 202 ties between blocks (ties of the overall interaction network – ties in block A – ties in block B – ties in block C).
601
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
602
INFORMATION, COMMUNICATION & SOCIETY
FIGURE 1
Socio-grams of the three blocks.
Results 2: interaction styles and network properties of individuals within the blocks Interaction styles. Having partitioned the players into three groups, we set out to examine the characteristics of each block. We would like to know the interaction styles of individuals in the blocks based on the content of interaction. For instance, did group A predominantly give help? The line chart (Figure 2) shows the proportion of ‘interaction types’ frequency the players of each block participated in. From this, we can identify some distinct differences between blocks. First, we notice that members of block C were engaged mostly in ‘ask for help’ (mean ¼ 0.48) and ‘friendly’ (mean ¼ 0.49) interactions, meaning that 48 per cent of the interaction was about asking for help and 49 per cent of the interaction was about friendly remarks. However, they did not participate at all in ‘game chat’ and ‘give help’ interactions. Members of block A
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
FIGURE 2
Proportion of interaction frequency for each interaction type.
participated almost equally in ‘friendly’ (mean ¼ 0.24), ‘game chat’ (mean ¼ 0.23) and ‘give help’ (mean ¼ 0.25) interactions, but they hardly asked for help (mean ¼ 0.07) in the guild community. Members of block B participated quite actively in ‘friendly’ (mean ¼ 0.35) and ‘game chat’ (mean ¼ 0.20) interactions. Although not actively, block B did engage both in ‘ask for help’ (mean ¼ 0.18) and in ‘give help’ (mean ¼ 0.08) interactions. Table 2 shows the mean of the proportion of all interaction types across three blocks. An analysis of variance (ANOVA) with permutation tests revealed that the interactions of the three blocks were significantly different from each other. TABLE 2
The mean table of interaction types in each block.
block
A
N
31
Friendly remarka
B 25
C 18
0.2383
0.3529
0.4907
Game chat
0.2348
0.2025
0.000
Real life chata
0.0157
0.0631
0.000
Ask for helpa
0.0681
0.1813
0.4796
Give helpa
0.2472
0.0844
0.000
Group management
0.1457
0.0795
0.0111
Coordinationb
0.0362
0.0068
0.000
a
a
a
The permutation ANOVA test with 5000 iterations shows that the three blocks were
significantly different at p , 0.001. b
The permutation ANOVA test with 5000 iterations shows that the three blocks were
significantly different at p , 0.05.
603
604
INFORMATION, COMMUNICATION & SOCIETY TABLE 3
Comparison of interaction type for each block.
interaction type
higher
moderate
lower
Friendly
C
B
A
Game chat
A, B
Ask help
C
B
A
C
Give help
A
B
C
Group management
A, B
C
Interpretation of the table: post hoc test shows that for friendly interaction, C was significantly higher than B and B was significantly higher than A. For game chat, C was significantly
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
higher than A and B, but no significance was found between A and B.
Post hoc tests were run with ‘friendly’, ‘game chat’, ‘ask help’, ‘give help’ and ‘group management’ categories to further explore the differences between blocks. Post hoc tests were carried out using permutation t-tests (10,000 iterations). Table 3 summarizes the post hoc test results that show the ranking of the three blocks in each interaction type under questions. Two blocks were put together if there was no statistical significance between them. Based on these results, we can infer that block A represented the ‘core members’ of the guild community as they provided instrumental support by acting as a source of knowledge (answering questions), giving practical help and managing groups. Apart from these, they also supported the guild through non-instrumental activities such as game chats. This nurtures a friendly environment within the guild community. Block C, on the other hand, was mainly asking questions in the guild and sending short friendly messages. We suspect that block C consisted of members whose main goal was to get help from the guild, but not to get involved in the guild community itself. Their purpose of joining the guild was simply to use the ‘resources’ of the guild to achieve their personal goals. It was also noted that their involvement in the social relation of the guild was limited to short and brief friendly remarks. In this case, the social interaction might have been just a means of instrumental play, e.g. obtaining help, since merely saying ‘hello’ did not consume a lot of their time. We would like to call them ‘periphery members’. As for the members of block B, they participated moderately in ‘give help’, ‘ask help’ and ‘friendly’ categories. They were not as active as the core member block in giving help, but were equally active in their involvement in game chat. Compared to block C, block B contributed more to the growth of the guild as they not only asked for help, but also gave help to other members. We would like to call them ‘semi-periphery members’.
Network properties. Having examined the interaction styles of each block, we turned our interest to their network properties. In other words, we wanted
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
to find out how the block members were positioned and embedded within their ‘neighbouring’ network (called ego network). To achieve this, we conducted ego-net analysis on each individual to take a closer look at their local social structure and position. Before we present the results, we would like to provide the definitions of some SNA terms: ego, the focal actor of the ego network on which our analysis focuses; degree, the number of ties of ego; density, the strength of ties in each ego network; reciprocity, the proportion of pairs having a reciprocated tie between them; reach efficiency, how much secondary contact the ego gets for each unit of primary contact. If reach efficiency is high, the ego is able to reach a wider network for each unit of effort invested in maintaining primary contact; betweenness, being in between each other pairs of actors. The index of the percentage of all shortest paths from neighbour to neighbour that pass through ego; closeness, being able to reach other actors at shortest path lengths.
. . . .
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
.
.
.
Table 4 shows that the three blocks were significantly different in terms of their degree, density, reach efficiency and closeness, while no significance was observed for reciprocity and betweenness. The post hoc test (with permutation t-test) results are summarized in Table 5. Two blocks were put together if there was no statistical significance between them. From these results, it appears that core members (block A) tended to interact with more other individuals and thus not surprisingly were able to reach more actors within the shortest path. However, they had unexpectedly low density, implying that although they were connected to a huge number of actors, they only interacted frequently with a small subset of them. Periphery members (block C) did not interact with a lot of players in general, resulting in low degree, density and closeness. They had, nevertheless, TABLE 4
The mean ego-net measures of each block.
block Degreea Densityb Reciprocity Reachb
A 2.125324 35.02667 0.119355 36.24269
B 0.575632 44.88 0.1164 52.36246
C 0.14304 39.5825 0.196471 76.24004
Betweenness
27.57447
27.31555
23.07692
Closenessb
20.04763
19.01916
18.11476
a
The permutation ANOVA test shows significant difference at p , 0.001.
b
The permutation ANOVA test shows significant difference at p , 0.05.
605
606
INFORMATION, COMMUNICATION & SOCIETY TABLE 5
The comparison of ego-net measures for each block.
type of activity
higher
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Degree centrality
moderate
lower
A
B, C
Density
B
Reach efficiency
C
B
A
A, C
Closeness
A
B
C
the highest reach efficiency among the three blocks, meaning that they mainly interacted with members who had a lot of connections. In other words, this implies that they did not interact with other periphery members since periphery members tended to have low degree centrality (see next section for further explanation). As opposed to block A, semi-periphery members (block B) had higher density but lower degree centrality. This means that although they were not connected to many other players, they interacted frequently with most of the players they were connected to. Furthermore, unlike block C who had both low density and degree, block B had a few connections but the connections were stronger. Tables 3 and 5 show some interesting relationships between the interaction types the players were engaged in and their ego-net measurements. For instance, we can notice the relation between reach efficiency and ask for help. Moreover, closeness seemed to be related to give help. Therefore, a permutation correlation analysis (with 5000 iterations) was carried out (Table 6). Unsurprisingly, it was found that a player’s ego-net degree was positively correlated to ‘game chat’ interaction and negatively correlated to ‘ask help’ activity, implying that players who were connected to many others tended to participate in ‘game chat’ interaction and were less likely to ask for help. That reflects the interaction styles and network properties of block A members. On the contrary, higher reach efficiency was positively correlated with ‘ask TABLE 6
The correlation between interaction type and ego-net measures. friendly remark
Degree Density Reciprocity Reach efficiency Betweenness Closeness a
group game chat
ask for help
0.279a
20.233a
0.272
0.134
20.1
20.006
20.041
20.18 a
a
0.143 a
give help
management
0.114
0.22
20.068
20.04
20.065
20.011 20.166
20.282
0.302
0.086
20.342a
0.084
0.134
0.084
0.169
0.218
0.233a
0.116
0.179
0.16
Correlation coefficient with p , 0.05.
20.18
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS TABLE 7
Density table. block A
block B
block C
Block A
0.314
0.146
0.057
Block B
0.146
0.07
0.016
Block C
0.057
0.016
0.007
for help’, but negatively correlated with ‘game chat’. This characterizes the network properties of block C members.
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Results of phase 3: inter-block interaction So far we have looked at the characteristics of each block by analysing the intermember interaction within the blocks. The density table (Table 7) allows us to explore the internality and externality of the interaction flow across the three blocks. The main diagonal shows the density within blocks. Generally, block A had very high internal density and therefore most ties were mainly internal within the block. Conversely, block C had internal density lower than their external density, suggesting that their ties were mainly external to other blocks. Block B was connected to block A, although compared to its tie to C, it had denser internal ties. The density table, however, does not tell us the direction of the interaction. In order to investigate the direction, the external –internal (EI) index (Krackhardt & Robert 1988) was calculated for each interaction type. EI index compares the numbers of ties within blocks and between blocks, and it ranges from 21 (all ties are internal within the group) to +1 (all ties are external to the group). Table 8 shows that the interaction direction of block A was predominantly internal in all interaction types, while block C was exclusively external in all interaction types. The magnitude of the EI index of block A was not as high as block C, suggesting that some ties from block A were also external to the block. Similarly, although block B interaction was predominantly external, the moderate magnitude of density implies the presence of internal ties. Assuming that the density lower than 0.05 (refer again Table 7) as being weak ties, we can describe the strong relationship of the three blocks in Figure 3. Although the core members interacted with the members from all three blocks, they interacted largely within the block. They formed a ‘self sufficient’ group, which did not rely on others to maintain their local network. They knew each other well and interacted very frequently with members known to them. Although they were also connected to many members from other blocks, the interaction was not as frequent as internal interaction. This possibly
607
608
INFORMATION, COMMUNICATION & SOCIETY TABLE 8
EI index table.
block
A
B
C
overall
Overall interaction
20.336
0.481
0.902
20.050a
Ask for helpb
20.421
0.619
1
20.137a
Give help
20.307
0.592
1
Friendly remarkb
20.366
0.509
0.833
20.076a
Group management
20.611
0.529
1
20.378a
Game chat
20.683
0.467
1
20.477a
0.018
a
The permutation test with 5000 iterations shows that the EI index was significant at p , 0.05.
b
’Ask help’ and ‘friendly’ messages were often posted to the whole guild community instead
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
of to each individual. Since our analysis excluded this aspect, ‘ask help’ and ‘friendly’ might have had higher EI index than that reported here.
FIGURE 3
The inter-block interaction model.
results in high degree but lower density index for this group. Although the EI index shows that core members generally interacted internally across different activities, they were particularly strongly connected internally in game chat and group management categories and least internally connected in give help category. Semi-periphery members interacted both within the block and with core members. The EI index suggests that they were more strongly connected to core members than to members from the same group. Although in general their interaction was external to the group, they tended to interact with core members the most through ‘ask for help’ and ‘give help’ categories. Their
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
interaction was, however, the least external when it comes to ‘game chat’ activity, implying that their internal interactions revolved mainly around ‘game chat’. Periphery members only interacted with core members. This was strongly supported by the EI index of 1 for all but one activity, ‘friendly remark’. Since periphery members did not give help nor did they engage in game chat or group management, the EI indices of these categories were meaningless.
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Discussion and conclusions This study identified the structural characteristics of three social roles of a guild community in WoW: a densely connected core members (block A), loosely connected fragments semi-periphery members (block B) and an outer ring of disconnected periphery players (block C). These three blocks of members, located at different structural positions in the guild network, clearly demonstrated distinct levels of participation and thus sense of belongingness to the community. We can conclude that core members were highly connected within the block and moderately connected to other blocks, and thus they were very likely to be players who had been longer in the guild and thus knew each other well. They were actively involved in game chat, group management and give help, but not ask for help. Semi-periphery members were those who were not in the core of the guild but attempted to get involved in the guild community. Apart from seeking help, they also gave help. Furthermore, they had high game chat and group management participation, indicating their active involvement. The periphery block consisted of members whose main goal was to get help from the guild but not involved in the guild community. They had high reach efficiency, meaning that they had access to a lot of other players, thus increasing the chance of getting help. They were mainly using the resource of the guild without contributing to the community. Through the findings, we can infer that the involvement in certain types of interactions hugely encourages the structural movement of the players from the periphery to the core. First, it is obvious that ‘give help’ interaction is a key factor that positions a player in the core network. However, the ability of giving help relies very much on the player’s knowledge and game skills. Thus, the longer the player plays the game, the more likely the player will take up the role as a core member. Interestingly, being knowledgeable is not the only characteristic of core members, as shown by the results, they were also active in ‘game chat’. Compared to ‘give help’, ‘game chat’ is less dependent on the player’s skills and knowledge and therefore can be enjoyed by anyone who wishes to do so. In fact, we observed some lower level players being categorized as core members due to their active participation in social-oriented tasks such as game chat.
609
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
610
INFORMATION, COMMUNICATION & SOCIETY
Therefore, there could be two types of core members: knowledge players and social players. Knowledge players are those who contribute by providing help and developing the guild as a pool of resource that would assist other players to complete their quests. This could be a factor that attracts more members into the guild. Therefore, it is reasonable to believe that knowledge players contribute to the expansion of the guild community. Social players, on the other hand, contribute by nurturing a friendly and welcoming atmosphere within the guild community. While this certainly encourages more players to join the guild, social players play a more important role in maintaining the cohesiveness of the guild community. As shown in Table 6, social interaction (i.e. friendly remark and game chat) was positively correlated to degree, density and closeness. Therefore, the more the social players in the guild, the more connected, the denser and the closer they are. It was also noted, through the observation, that although periphery members consisted of mostly new players with lower level, that was not always the case. Some higher level players were so located at the periphery, due to their lack of participation in other forms of interactions apart from ‘asking for help’. Therefore, two types of periphery members could be identified: ‘newbie’ and ‘freeloaders’. Newbies are players who are new not only to the guild but to the game in general. They need help on basic issues such as the game –user interface, basic rules, etc. Although quite similar in terms of their interaction patterns, what differentiate a freeloader from a newbie are their skills and the types of help they ask. A freeloader is a player with generally higher level who asks quest-related questions. They do not feel belonged since they are using the guild only as an instrumental tool for their task interaction. Newbies might gradually move towards the inner social circle of the guild community as they are gaining more experience and skills. As the newbie moves into the semi-periphery circle, they will start forming ‘game chat’ clusters and giving help to others. ‘Ask for help’ interaction drops dramatically while a player moves from periphery to semi-periphery to core block as we can see from Figure 2. However, some of the newbies might also turn into a freeloader. From the design point of view, we want to prevent players from becoming freeloaders by encouraging them to participate.
Future work This work has also some significant implications on other areas of virtual communities such as computer support collaborative work and learning. Similar approaches can be used to study the social roles in such communities in order to increase productivity and learning outcome. One possible way of extending this work is to study a bigger guild for a longer period of time to identify more detailed sub-roles within each block. A more specific typology of social roles derived from relational data and analysis
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
can be constructed to better understand players’ behaviour in the community. Another area of potential expansion is the observation of ‘virtual actions’. Instead of just observing the ‘chat interaction’, we can record the non-verbal interaction carried out by the players, such as killing monsters together. This will allow the collection and analysis of richer interaction between players. For instance, a player who gives a powerful sword to another player might be considered a much stronger tie of ‘giving help’ than ‘telling the direction’. This also gives us an opportunity to contrast networks that arise from verbal and non-verbal interactions. Finally, like Ducheneaut et al. (2004, 2006), we also strongly believe that game design, especially game rules/mechanics and formal guild rules, can have a significant impact on the structure of the social networks that arise and thus the social roles that form. The study by Ducheneaut et al. (2006) showed that successful MMORPGs are a result of careful game design as well as community design. They noted that MMORPGs have a carefully designed reward structure, which reinforces player commitment throughout the game play. Therefore, a future research direction can be centred on establishing the relationship between the design of game/community rules and the emergence of social rules and how they contribute to the engagement of the sociability experience.
References Anderson, B. (1991) Imagined Communities: Reflections on the Origin and Spread of Nationalism, Verso, London. Anderson, B. (2004) ‘Dimensions of learning and support in an online community’, Open Learning, vol. 19, no. 2, pp. 183–190. Ang, C. S., Zaphiris, P. & Wilson, S. (2005) ‘Social interaction in game communities and second language learning’, paper presented at the 19th British HCI Group Annual Conference, Edinburgh, UK, September. Ang, C. S., Zaphriis, P. & Mahmood, S. (2007) ‘A model of cognitive loads in massively multiplayer online role playing games’, Interacting with Computers, vol. 19, no. 2, pp. 167–179. Ashforth, B. E. (2001) Role Transitions in Organizational Life, an Identity-based Perspective, Lawrence Erlbaum Associates, Mahwah, NJ. Bartle, R. (1996) ‘Hearts, clubs, diamonds, spades: players who suit muds’, Journal of MUD Research, vol. 1, no. 1, [Online] Available at: http://www.mud.co. uk/richard/hcds.htm (16 March 2010). Blizzard Entertainment (2004) ‘World of warcraft’, [Online] Available at: http:// www.worldofwarcraft.com/ (10 March 2010). Borgatti, S. P., Everett, M. G. & Freeman, L. C. (2002) ‘Ucinet for Windows: software for social network analysis’, [Online] Available at: http://www. analytictech.com/ucinet/ (10 March 2010).
611
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
612
INFORMATION, COMMUNICATION & SOCIETY
Breiger, R., Boorman, S. & Arabie, P. (1975) ‘An algorithm for clustering relational data, with applications to social network analysis and comparison with multidimensional scaling’, Journal of Mathematical Psychology, vol. 12, no. 3, pp. 328–383. Bruckman, A. S. (1997) Moose Crossing: Construction, Community, and Learning in a Networked Virtual World for Kids, PhD dissertation, MIT. Cohen, J. (1960) ‘A coefficient of agreement for nominal scales’, Educational and Psychological Measurement, vol. 20, no. 1, pp. 37–46. Ducheneaut, N., Moore, R. J. & Nickell, E. (2004) ‘Designing for sociability in massively multiplayer games: an examination of the “third places” of SWG’, paper presented at Other Players conference, Copenhagen, Denmark, December 2004. Ducheneaut, N., Yee, N., Nickell, E. & Moore, R. J. (2006) ‘Games and performances: “alone together”? Exploring the social dynamics of massively multiplayer online games’, paper presented at the SIGCHI Conference on Human Factors in Computing Systems CHI ‘06, April Canada. Golder, S. A. & Donath, J. (2004) ‘Social roles in electronic communities’, paper presented at the Association of Internet Researchers (AoIR) Conference Internet Research 5.0, Brighton, UK, 19 – 22 September, [Online] Available at: http://web.media.mit.edu/~golder/projects/roles/golder2004.pdf (16 March 2010). Haythornthwaite, C. & Wellman, B. (1998) ‘Work, friendship, and media use for information exchange in a networked organisation’, Journal of the American Society for Information Science and Technology, vol. 49, no. 2, pp. 1101–1114. Herrmann, T., Jahnke, I. & Loser, K. (2004) ‘The role concept as a basis for designing community systems’, in Cooperative Systems Design, eds F. Darses, R. Dieng, C. Simone & M. Zackland, IOS Press, Amsterdam, pp. 163–178. Jakobson, M. & Taylor, T. L. (2003) ‘The Sopranos meets EverQuest: social networking in massively multiplayer online games’, in Proceedings of DAC 2003, Melbourne, Australia, May 2003, pp. 81 – 90. Koku, E. F. & Wellman, B. (2004) ‘Scholarly networks as learning communities: the case of Technet’, in Designing Virtual Communities in the Service of Learning, eds S. Barab & R. Kling, Cambridge University Press, Cambridge, pp. 299–337. Krackhardt, D. & Robert, N. S. (1988) ‘Informal networks and organizational crisis: an experimental simulation’, Social Psychology Quarterly, vol. 51, no. 2, pp. 123–140. Krippendorff, K. (1980) Content Analysis: An Introduction to its Methodology, Sage Publications, Newbury Park, CA. Kuwashima, Y. (2006) ‘The word-of-mouth effect on the web’, MMRC Workshop, [Online] Available at: http://merc.e.u-tokyo.ac.jp/mmrc/dp/pdf/ MMRC79_2006.pdf (16 March 2010). Landis, J. R. & Koch, G. G. (1977) ‘The measurement of observer agreement for categorical data’, Biometrics, vol. 33, pp. 159 –174.
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
SOCIAL ROLES OF PLAYERS IN MMORPG GUILDS
Lave, J. & Wenger, E. (1991) Situated Learning. Legitimate Peripheral Participation, Cambridge University Press, Cambridge. Mead, G. H. (1934) Mind, Self and Society, University of Chicago Press, London. Muramatsu, J. & Ackerman, M. S. (1998) ‘Computing, social activity, and entertainment: a field study of a game mud. Computer Supported Cooperative Work’, The Journal of Collaborative Computing, vol. 7, nos 1 –2, pp. 7–122. Parsons, T. (1951) The Social System, Routledge & Paul, London. Rheingold, H. (1993) The Virtual Community: Homesteading on the Electronic Frontier, Addison-Wesley, Reading. Smith, W. M. (1966) ‘Observations over the lifetime of a small isolated group: structure, danger, boredom, and vision’, Psychological Reports, vol. 19, no. 2, pp. 475–514. Snijders, T., Steglich, C., Michael, S. & Mark, H. (2007) ‘Manual for SIENA version 3.1’, [Online] Available at: http://stat.gamma.rug.nl/sie_man31.pdf (16 March 2010). Sony Online Entertainment (2003) ‘Star War Galaxies’, [Online] Available at: http://starwarsgalaxies.station.sony.com (16 March 2010). Stark, R. (2007) Sociology, 10th edn, Thomson, Wadsworth, CA. Steinkuehler, C. A. (2004) ‘Providing resources for MMOG guild leaders’, paper presented at the MUD Developers Conference, San Jose, CA, March 2004. Stokman, F., van Duijin, M. & Snijders, T. (2004) ‘Stocnet: an open software system for the advanced statistical analysis of social networks’, [Online] Available at: http://stat.gamma.rug.nl/stocnet/ (16 March 2010). Strijbos, J. W., Martens, R. & Jochems, W. (2003) ‘The effect of roles on group efficiency’, in International Conference on Computer Support for Collaborative Learning 2003, eds B. Wasson, R. Baggetun, U. Hoppe & S. Ludvigsen, Intermedia, Bergen. Turner, T. C., Smith, M. A., Fisher, D. & Welser, H. T. (2005) ‘Picturing Usenet: mapping computer-mediated collective action’, Journal of Computer-Mediated Communication, vol. 10, no. 4, [Online] Available at: http://jcmc.indiana. edu/vol10/issue4/turner.html (16 March 2010). Wasserman, S. & Faust, K. (1994) Social Network Analysis: Methods and Applications, Cambridge University Press, Cambridge. Wellman, B. (1997) ‘An electronic group is virtually a social network’, in Cultures of the Internet, ed. S. B. Kiesler, Lawrence Erlbaum, Hillsdale, NJ, pp. 179–205. Wellman, B. & Berkowitz, S. D. (1994) Social Structure: A Network Approach, JAI Press Inc., London. Welser, H. T., Gleave, E., Fisher, D. & Smith, M. (2007) ‘Visualizing the signatures of social roles in online discussion groups’, Journal of Social Structure, vol. 8, no. 2, [Online] Available at: http://www.cmu.edu/joss/content/articles/ volume8/Welser/ (16 March 2010). White, H. C., Boorman, S. A. & Breiger, R. L. (1976) ‘Social structure from multiple networks. I. Blockmodels of roles and positions’, American Journal of Sociology, vol. 81, January, pp. 730–780.
613
614
INFORMATION, COMMUNICATION & SOCIETY
Yee, N. (2005) ‘The psychology of MMORPGs: emotional investment, motivations, relationship formation, and problematic usage’, in Avatars at Work and Play: Collaboration and Interaction in Shared Virtual Environments, eds R. Schroeder & A. Axelsson, Springer-Verlag, London, pp. 187–207.
Downloaded By: [HEAL-Link Consortium] At: 19:30 18 June 2011
Dr C. S. Ang is a Lecturer at the School of Engineering and Digital Arts at the University of Kent. He is interested in human interactions and social tendencies in the virtual world, such as Second Life, from a sociological, psychological and HCI perspective. His main research interests include the psychology and sociology of computer games, virtual worlds or 3D computer-mediated communication, learning theories particularly in gaming, digital media such as interactive narrative and simulation. Address: School of Engineering and Digital Arts, University of Kent, UK [email: csa8@kent.ac.uk]
Dr Panayiotis Zaphiris is an Associate Professor at the Department of Multimedia and Graphic Arts of Cyprus University of Technology (CUT). Before joining CUT, he spent seven years at City University, where he holds the title of Reader. Before City he was a researcher at the Institute of Gerontology at Wayne State University from where he also got his PhD in HCI. His research interests lie in HCI with an emphasis on inclusive design and social aspects of computing. He has a strong interest in Internet-related research (web usability, mathematical modelling of browsing behaviour in hierarchical online information systems, online communities, e-learning, web-based digital libraries and social network analysis of online human-to-human interactions). He has published over 120 publications in prominent journals (e.g. Zaphiris, P. & Sarwar, R. (2006) ‘Trends, similarities and differences in the usage of teen and senior public online newsgroups’, ACM Transactions on Computer – Human Interaction (TOCHI), vol. 13, no. 3, pp. 403– 422, ACM Press) and has presented his work in numerous conferences. For more information, see http://www.zaphiris.org/. Address: Department of Multimedia and Graphic Arts, Cyprus University of Technology, Limassol, Cyprus. [email: zaphiri@soi.city.ac.uk]