Computers in Human Behavior 26 (2010) 218–225

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier .com/locate /comphumbeh

User goals in social virtual worlds: A means-end chain approach

Yoonhyuk Jung a,*, Hyunmee Kang b,1

a Department of Information Systems and Decision Sciences, Louisiana State University, Baton Rouge, LA 70803, USAb The Manship School of Mass Communications, Louisiana State University, Baton Rouge, LA 70803, USA

a r t i c l e i n f o a b s t r a c t

Article history:Available online 1 November 2009

Keywords:Virtual worldsSocial virtual worldsUser goalGoal structureMeans-end chain analysis

0747-5632/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.chb.2009.10.002

* Corresponding author. Tel.: +1 225 578 2126; faxE-mail addresses: yjung1@tigers.lsu.edu (Y. Jung

Kang).1 Tel.: +1 225 578 2336; fax: +1 225 578 2125.

The purpose of this study is twofold: first, to investigate user goals in social virtual worlds; second, tointroduce a methodological alternative (i.e., a means-end chain approach) for analyzing user goals incyberspaces. The data were acquired from a web survey, and were analyzed by means-end chain analysis(MECA), which produces users’ goal structure in reference to a hierarchical system of interrelated goals(Olson & Reynolds, 1983). The results show that people come to social virtual worlds to satisfy their socialand hedonic needs, and to escape from real world constraints, as do virtual community members and vir-tual gamers; they also pursue unique activities, such as creating virtual objects and selling them. On theother hand, by clarifying relations among users’ goals, MECA provides a richer explanation for user goalsthan prior research which only offers separate user goals for cyberspace users without explanation ofrelationship among goals.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Virtual worlds (VWs) refer to computer-simulated, graphic-based, virtual environments that are accessed by multiple users.They have grown dramatically over the last decade as supportingtechnologies (e.g., virtual reality, broadband) have been advanced.World of Warcraft, which is the largest gaming virtual world(GVW), had over 10 million subscribers as of the end of 2008(MMOGchart.com, 2008). Virtual gaming even became a profes-sional sport with leagues, sponsorships, and spectators in SouthKorea (Castronova, 2005). Recently a new type of VWs that stresssocial interaction and user empowerment has appeared, namelysocial virtual worlds (SVWs) (e.g., Second Life, There.com). GVWsare characterized by a pre-defined structure and quest-drivenbehaviors, whereas SVWs have emergent structures that are cre-ated by users under minimum constraints (Juul, 2005).

SVWs offer their users an opportunity to determine their expe-riences in the virtual worlds for themselves (Dreyfus, 2008). Thisautonomy makes the worlds places that are filled with diverseactivities such as socializing, learning, sprouting virtual business,entertainment, and so on. The increased number of users reflectsSVWs’ popularity. For instance, Second Life, a popular SVW, an-nounced that its subscribers exceeded 12 million as of February2008 (SLOG, 2008). The amount of real-money trading in SVWshas also been increasing exponentially. User-to-user trading of vir-

ll rights reserved.

: +1 225 578 2511.), hkang2@tigers.lsu.edu (H.

tual goods in Second Life was estimated to be 360 million dollar in2008 (Linden Lab, 2009). In addition to SVWs’ quantitative growth,their potential as a marketing channel and a collaboration tool hasattracted both managers and researchers’ attention. Several corpo-rations, such as Nike, Adidas, Levi’s, GM, and Toyota, create virtualproducts, which simulate real products, and enable SVW users tohave virtual use experiences. Also, IBM trains new employees inorganizational culture and work processes, holds distant meetings,and operates customer service centers in SVWs (Hobson, 2007).Regarding SVWs as a future platform for e-learning, many educa-tional organizations including Harvard University have beenincreasingly using SVWs.

Despite renowned interest in SVWs, there is currently littleempirical research on SVWs, particularly on users’ perceptionsand behaviors in SVWs. SVWs’ distinctive characteristics maycause users to access and behave in such environments differentlythan in other cyberspaces, such as text-based virtual communitiesand even GVWs. The initial step to study SVW users may be a ques-tion of ‘why do people come to SVWs?’; that is, user goals foraccessing SVWs. Knowledge of these goals would benefit busi-nesses that are considering the strategic use of SVWs, as well asSVW operators in terms of attracting and retaining more users,and also the researchers that are begging to investigate the natureof SVW users.

In order to comprehend user goals for accessing SVWs, thisstudy employed means-end chain analysis (MECA), which is con-sidered an effective method for eliciting people’s goal structure(i.e., goals and their relations) when they approach an object oran event (Olson & Reynolds, 1983). Prior studies that investigatedusers’ goals for participating in virtual communities (Ridings & Ge-

Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225 219

fen, 2004) and gaming virtual worlds (Bartle, 2003; Yee, 2006),produced sets of separate individual user goals but offer littleexplanation about the relations among goals. Studies of relationsamong goals, or a goal structure, can provide more informationabout users’ goal-oriented behavior than those of isolated goals(Pieters, Baumgartner, & Allen, 1995). Because it can clarify goalstructure, a means-end analysis is expected to offer richer informa-tion about user goals for participating in SVWs. This study thuscontributes to research on cyberspace users, in terms of offeringa methodological novelty, and demonstrates an empirical studyof an emerging cyberspace.

2. Theoretical background

2.1. Social virtual worlds

Despite its diverse definitions, the term virtual world has beencommonly used to indicate a computer-simulated persistent spatialenvironment that supports synchronous communication amongmultiple users who are represented by avatars (Holmstr}om &Jakobsson, 2001; Jakobsson, 2006). This concept of VWs is similarwith the notion of MMOGs or massive multiplayer online games,such as World of Warcraft. However, MMOGs have been mainlyused to specify one type of VW that has a pre-defined theme andplot and clarifies users’ performances (e.g., level-ups). Althoughthey (or GVWs) still occupy the majority of VWs, the other distinc-tive VWs, where users create their experiences for themselves andhave diverse social interaction, have dramatically increased. In thepresent study, we call these types of VWs social virtual worlds(SVWs) and differentiate them from gaming virtual worlds(GVWs). In GVWs, user activities are based on pre-defined themesand plots which the game designers imagined and produced. Theiractivities usually aim at a quest or level-ups rather than social inter-actions. On the other hand, the lack of predetermined storyline ofSVWs makes them distinct from GVWs, and this openness has at-tracted various kinds of users (Warburton, 2009). SVWs endowusers with the ability to personalize their virtual experiences,which induce various social interactions under minimum con-straints. As a result, SVWs’ nature of minimum rules leads to diver-sity in members’ behavior (Juul, 2005). Furthermore, as someSVWs (e.g., Second Life, There), supports transactional systems(e.g., virtual currency, sanctioned virtual market), and allow users

Table 1User goals in cyberspaces.

Type of cyberspaces Purposes Description

Virtual communities Informationexchange

Attain and transfer

Social relations Get involved with oPsychologicalsupport

Attain and offer em

Entertainment Attain fun

Gaming virtualworlds

Socializers Form groups and coExplorers Seek new placesAchievers Pursue the gradual aControllers Compete with and d

Gaming virtualworlds

Achievement Advancement Gain power and accMechanics Analyze the underly

performanceCompetition Challenge and comp

Social Socializing Help and chat withRelationships Form long-term meTeamwork Be part of a group e

Immersion Discovery Find things that moRole-playing Create a persona wiCustomization Customize the appeEscapism Avoid thinking abou

to create virtual objects, users’ activities are extended to economicactivities, that is, production and real-money trade of virtual ob-jects. This autonomy distinguishes SVWs’ characteristics fromGVWs’.

2.2. User goals in social cyberspaces

User goals for accessing SVWs may be analogous to motivationsfor joining virtual communities (VCs) and playing GVWs. VCs arecyberspaces in which people communicate and form networks ofpersonal relationships (Rheingold, 1993). SVWs that support 3Dinterfaces and real-time avatar interactions can be differentiatedfrom conventional VCs that depend on asynchronous text-basedinteractions. Nevertheless, because of the social nature, SVWs areusually considered an extension of VCs. Prior studies note thatinformation exchange, social relations, psychological support, andentertainment are common goals for joining VCs (Bressler & Gran-tham, 2000; Hagel & Armstrong, 1997; Wellman, 1996). Informa-tion exchange indicates that the members share informationgenerated by other members rather than merely accessing infor-mation that the website operator provides (Hagel & Armstrong,1997). VC members also interact with other members and areembedded in web of personal relationships in VCs; that is, socialrelations (Rheingold, 1993). Additionally, people satisfy psycholog-ical needs, such as emotional support, a sense of belonging, andencouragement in VCs, and also come to VCs for entertainment.Ridings and Gefen (2004) empirically examined users’ motivationto join VCs and confirm that those four goals are the main motiva-tional forces.

Bartle (2003) classifies GVW users into four types according totheir goals: socializers, explorers, achievers, and controllers. Socializ-ers attempt to form groups and complete shared objectives;explorers seek new places in a GVW; achievers pursue the gradualaccumulation of wealth and reputation in GVWs; and controllerswant to compete with and defeat others. Yee (2006) empiricallyexamined Bartle’s GVW user typology through an exploratory fac-tor analysis, and proposes a new framework that consists of threeoverarching goals: achievement, socializing, and immersion. Eachoverarching purpose is composed of sub-motivations. Achieve-ment includes advancement, mechanics, and competition; socializ-ing includes relationships and teamwork; and immersion includesdiscovery, role-playing, customization, and escapism (see Table 1).

References

information about a topic, learn new things Ridings and Gefen(2004)

ther membersotional support

mplete shared objectives Bartle (2003)

ccumulation of wealth and reputationefeat others

umulate wealth or status Yee (2006)ing rules and system to optimize character

ete with othersother playersaningful relationships with othersffortst other players do not know aboutth a background story and interact with othersarance of their charactert real life problems

2 Prior research employing a laddering technique usually has from 40 to 140mples (51 samples in Pieters et al. (1995), 53 samples in Klenosky (2002), 82 in

oss et al. (2007), 133 samples in Bagozzi and Dabholkar (1994)). Thus, based on priorsearch, our number of samples is acceptable.

220 Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225

All prior studies that investigate user goals for inhabiting cyber-spaces provide sets of isolated goals without addressing the rela-tions among those goals. Although the prior studies recognizethat goals do not suppress each other (i.e., a user may have multi-ple goals) (Bartle, 2003; Yee, 2006), they overlook how they areconnected. As a result, the prior studies provide just a set of sepa-rate goals for social cyberspaces. This limit may preclude drawing amore comprehensive picture of user goals because goals need to beexplained within a goal structure that includes a hierarchical sys-tem of interrelated goals (Pervin, 1989). Accordingly, this study at-tempts to clarify a goal structure beyond finding isolated goals byemploying a means-end chain analysis.

2.3. User goals and a means-end chain analysis

A goal is a desired outcome of an action (Locke & Latham, 1990).Many researchers argue that goals exist within a hierarchical sys-tem where a goal is located between its superordinate and subor-dinate goals, and furthermore, each goal is a means to achieve itssuperordinate goal (Kruglanski et al., 2002; Newell & Simon,1972; Pervin, 1989). Means-end chain analysis (MECA) stems fromthe idea of a hierarchical goal system. The analysis posits thatproduct or service attributes represent the means by which con-sumers achieve benefits and important personal values (i.e., ends)(Gutman, 1982; Olson & Reynolds, 1983). In other words, MECAis an approach for discovering the important meanings that con-sumers ascribe to a product or service’s attributes (Voss, Gruberb,& Szmigin, 2007). The analysis assumes that consumer knowledgeis hierarchically organized by levels of abstraction (Reynolds &Whitlark, 1995), and focuses on a product or service’s meaningsat three levels of abstraction: attributions, consequences, and values.Attributes refer to a product or service’s physical or observableproperties: consequences are the benefits attained by the attri-butes; and values imply highly abstract motivation that guidesusage behavior (Klenosky, 2002). An attribute–consequence–valuechain is usually expressed by a hierarchical map, which consists ofnodes (i.e., attributes, consequences, and values) and relationshipsamong them.

A means-end chain analysis typically depends on a ladderinginterview technique, which has been comprehensively used in con-sumer research that tries to understand consumers’ preferences to-ward products or services (e.g., Klenosky, 2002; Reynolds &Rochon, 2001), and organization research, which elucidates anorganization’s strategic values and decision-making structures(e.g., Peffers, Gengler, & Tuunanen, 2003; van Rekom, van Riel, &Wierenga, 2006). For helping respondents elicit lower or higherlevels of abstractions for the concepts, the technique aims tounderstand the way in which the respondent sees the world (Rey-nolds & Gutman, 1988). A laddering procedure typically includesthree questions: the attribute question (What attribute makesthe product (or service) attractive to you?), the consequence ques-tion (Why is the attribute important or desirable to you?), and thevalue question (Why is your response important?). In the firstphase, the respondent is asked to supply the attributes of a productthat affect his or her consumption decision. The respondent is thenasked to explain what benefits he or she attains owing to the attri-butes. Finally, the respondent is asked to offer the reason whythose benefits are important to him or her, namely, a justificationstage. All responses are coded, and then a hierarchical map is final-ly produced by the level of abstraction, that is,attributes ? consequences ? values.

Even though means-end chain studies that employ a ladderinginterview technique usually use the traditional way as statedabove, some studies employ a modified MECA. The main modifica-tion appears to be a way to decide the level of abstraction and atechnique for data collection. Employing network theory (see Scott,

1991), some studies calculate the abstractness of each element(concept) and use it to determine the position of the element ina hierarchical map instead of a strict specification of three levelsof abstraction (i.e., attributes, consequences, and values) (Bagozzi& Dabholkar, 1994; Capozza, Falvo, Robusto, & Orlando, 2003; Piet-ers et al., 1995). This revised method allows researchers to knowthe relationships of elements without having to conduct additionalwork to classify elements into three levels. One critique of the lad-dering technique is that an answer frequently does not correspondto the question (e.g., a consequence or value answer to the attri-bute question). Studies that employ network theory reduce thislimitation because each element has a level according to itsabstractness without any label (e.g., attribute) in the revised meth-od. On the other hand, instead of in-depth interviews, some studiesuse questionnaires: a technique developed by Walker and Olson(1991) (e.g., Botschen & Hemetsberger, 1998; Pieters et al., 1995;Voss et al., 2007). The advantage of a questionnaire version is thatrespondents themselves decide when they finish the ladderingprocess, which may make respondents feel pressure (Botschen &Hemetsberger, 1998). In addition, compared to the in-depth inter-views, a questionnaire version is a cost-effective method for datacollection (Botschen, Thelen, & Pieters, 1999).

3. Methodology

3.1. Data gathering

The target SVW for this study was There (www.there.com),which is a SVW equipped with a 3D environment. The reasonwhy we chose There was that it is closer to our definition of SVWsin terms of users’ autonomy (i.e., diverse user activities includingcreating and selling virtual objects), and it has recently grown rap-idly in terms of membership. The number of its registered mem-bers is over one million as of early 2008. In There, a member hasa personal avatar that represents oneself. A member can manipu-late his or her personal avatar’s face, hair, and body and put it intoclothes. Also, a member can create 3D objects (e.g., chair, building,waterfall) using developer program, and perform virtual tasks withthem or sell them to other members. Controlling their personalavatars, members enjoy synchronous chatting at the park or onthe beach, dancing at night club, or taking a buggy in There.

During two weeks, we solicited Thereians, which There users callthemselves, to participate in the web-based survey. Two ap-proaches were used to recruit participants. First, we sent groupleaders of a wide diversity of groups messages which requestedthem to distribute our solicitation message which included theweb-based survey link. Also, we directly recruited participants invarious places (e.g., beaches, parks, sandboxes) of There. We firstsent longed-in users a message that introduced our survey, and ifthey wanted to participate in the web-based survey, then we sentthe link. Fifty-four Thereians, which There users call themselves, re-sponded to our web-based survey or a questionnaire-based ladder-ing technique.2 Participation in this study was voluntary. Theparticipants were heterogeneous in age, gender, and educationalbackground as seen in Table 2, indicating that There is used by variedsorts of people. The majority of the participants can be regarded ashighly-attached users in that over 60% of the participants logged inThere daily or several times a week and over half of the participantshad used There over one year.

The questionnaire consisted of three questions for a ladderinganalysis and five questions about demographic information. Based

saVre

Table 2Demographics of participants.

Frequency Percent

Age 18–24 10 18.625–34 6 11.135–44 13 24.145–54 11 20.455 or older 3 5.5No answer 11 20.4

Gender Male 21 38.9Female 22 40.7No answer 11 20.4

Education High school 11 20.4Community college 9 16.7Undergraduate 7 13.0Graduate 15 27.8No answer 12 22.2

Logging frequency Once a month 3 5.6Once a week 4 7.4Several times a week 12 22.2Daily 24 44.4No answer 11 20.4

Tenure Less than 6 months 7 13.06 months to 1 year 6 11.11 year to 2 years 11 20.4Longer than 2 years 19 35.2No answer 11 20.4

Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225 221

on prior laddering research, we made three open questions toprobe members’ goals for using There: (1) we first placed the over-arching question (What are your main three purposes for usingThere?); (2) we then asked the downward question (What charac-teristics of There help you achieve each purpose?); (3) and we finallyasked the upward question (Why is each purpose important to you?).In our procedure, the overarching question corresponds to the con-sequence question of a conventional laddering technique; thedownward question corresponds to the attribute question; andthe upward question corresponds to the value question (seeFig. 1). As previously stated, because our focus is on the relation-ships of elements rather than of their classification into attributes,consequences, and values, we used our own terms rather thanthose three labels. Also, based on our pilot study, we ordered thethree questions. In the pilot study, most of our respondents an-swered their consequences or values instead of attributes to theattribute question (the downward question in our study). In thissituation, we had difficulty continuing to probe because they hadalready answered their highly abstractive purposes. For example,when we asked the attribute question (i.e., What characteristicmakes There attractive to you?), many respondents answered justfun or meeting new people which corresponded to consequencesor values rather than attributes. Thus, we had to ask the attributequestion again (i.e., What characteristic of There help you achievethis purpose?). In order to conduct more effective web survey-based laddering procedure, we needed to modify a laddering pro-cedure, and placed the overarching question first. Our procedure,in which an intermediate question (consequence question) isplaced first, is also consistent with thought: humans set a goal atan intermediate level rather than explicitly pursuing a highly ab-stract goal, and then decide operational strategies to attain it (Rif-kin, 1985).

4. Results and analysis

The analysis consisted of two stages: (1) coding and (2) gener-ating a goal structure (a hierarchical goal map).

4.1. Coding

For analysis, the responses from three probing questions werecoded. One of the authors coded the data using an open coding pro-cedure in which codes were not predetermined but rather emergedfrom the data. This resulted in 45 detailed codes present in thedata. In cases where the data contained more than one topic, multi-ple codes were assigned. For example, ‘‘Thereians want to be soci-able and have fun” was assigned two codes – Social relations andFun. A second coder, the other author, independently re-codedthe data using the set of codes identified by the first coder. Thetwo raters were in agreement on 292 of the 363 codes assigned(Cohen’s Kappa = 0.78), indicating an acceptable level of inter-raterreliability (Fleiss, 1981). Inter-rater disagreements were then rec-onciled through discussion. Finally, associated codes were groupedinto 11 topics, as shown in Table 3. This categorization was alsoverified by the same procedure as the detailed coding procedure(Cohen’s Kappa = 0.91).

4.2. Generating a goal structure

Responses to the three questions (the overarching question, thedownward question, and the upward question) generated ameans-end chain, or a ladder of meanings; that is, answers to thedownward question pertain to a means for answers to the over-arching question, and likewise answers to the overarching ques-tion correspond to a means for answers to the upward question.For instance, if a subject responds Technical features to the down-ward question; Social relations to the overarching question; andAmusement to the upward question, two direct linkages are cre-ated: Technical features ? Social relations, and Social rela-tions ? Amusement. We can also consider an indirect linkage; forinstance, eliciting the linkage of Technical features ? Amusementfrom the linkage of Technical features ? Social relations ? Amuse-ment. Ultimately all linkages were summarized in an implicationmatrix which depicts the number of times each topic (code) leadsto each other topic in responses (Klenosky, 2002). As can be seen inTable 4, each topic in the row leads to the other topics in the col-umn. For instance, T5 (Social relations) led to T8 (Escapism) 2 times;T11 (Technical features) led to T10 (Amusement) 5 times.

Typical means-end chain studies classify responses into attri-butes, consequences, and values and then produce a hierarchicalstructure of attributes ? purposes ? values. In order to mitigateclassification errors, the current study employed an alternativemethod proposed by Bagozzi and Dabholkar (1994) and Pieterset al. (1995). Instead of classifying responses into three labels, thisapproach, which is based on network analysis (Scott, 1991), pro-duces a hierarchical structure by comparing the number of timeseach element is mentioned as the means versus the end. The ap-proach uses out-degrees and in-degrees in order to estimateabstractness of each element. Out-degrees of a particular elementrefer to the number of times the element serves as the source ororigin (means) of linkages with other elements (i.e., the row sumof the element in an implication matrix), whereas in-degrees ofthe element indicate the number of times the element serves asthe object or end of linkages with others (i.e., the column sum ofthe element in an implication matrix) (Pieters et al., 1995).Abstractness of an element is the ratio of in-degrees over in-degreesplus out-degrees of the element, and ranges from 0 to l (Pieterset al., 1995). Elements with high abstractness scores are regardedmainly as ends, while ones with low abstractness scores arethought of primarily as means. Based on the alternative approach,we created an implication matrix (see Table 4). Additionally, in or-der for informative analysis, this study calculated centrality of eachelement, which represents the degree to which the element has acentral role in the structure (Knoke & Burt, 1982). Centrality is cal-

Fig. 1. Laddering procedure.

Table 3Topics (super-codes) and sub-codes.

Topics Codes

T1. Addiction (removed) Addiction/hookedT2. Creating Creating virtual objects/decorating own avatar or property/creativityT3. Educational tool (removed) Educational toolT4. Escapism Escaping from reality/free from disabilityT5. Social relations Social interaction/meeting people/chatting/social events/dating

Finding friendly usersHaving time with family or offline friendsHelping peopleWorldwide availability/interacting with diverse people

T6. Exploring Exploring/traveling/adventuring/walking aroundLots of places/large scale of the virtual world

T7. Financial Economic freedom/auction systemFinancial/running a business/making money

T8. Knowledge acquisition Acquiring new ideasDoing researchImproving practical skills (programming)

T9. Positive administration (removed) Free membershipGood administration of There.comTutorial & class for training how to develop virtual objects

T10. Amusement Fun/enjoymentGaming (paintball, spades)Movies/musicRelaxing/reducing stressSomething to do/goofing off/killing timeBuggies

T11. Technical features 3D environment/virtual reality/graphicHuman-like avatars’ behaviorSimple operation/easiness of navigationTeleportThe mapVoice/IM/Email (communication tools in There)

222 Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225

culated by dividing the ratio of in-degree plus out-degree of a par-ticular element by the sum of all active cells in the implication ma-trix (the sum = 167 in the current study).

The next step was to generate a hierarchical map according tothe information in the implication matrix. In this stage, the impor-tant point was determining what linkages were included in a hier-archical goal map. Because inclusion of all linkages could decreasea map’s usefulness and informativeness, we did not embrace alllinkages and decided to employ a cutoff level (Reynolds & Gutman,1988). Following Bagozzi and Dabholkar’s (1994) method, we builtTable 5 to choose a cutoff level and finally selected a cutoff of four,indicating that the included relations are counted at least fourtimes. This cutoff level represented 30.0% of the active cells and64.1% of the active linkages, which corresponds to a measure ofvariance (Gengler & Reynolds, 1995). According to the cutoff, T1(Addiction), T3 (Positive administration), and T9 (Educational tool)

were excluded because they had no linkage to satisfy the cutoffcriterion.

The hierarchical goal map in Fig. 2 offers a graphical summaryof the means-end structure pertinent to using a SVW. In themap, the topics are placed relative to their abstractness scores.Accordingly, the more abstract a topic, the higher it is located inthe map.

5. Discussion

A means-end chain analysis played an effective role in clarifyingSVW users’ goal structure. The analysis is summarized in the hier-archical goal map, which provides a quick and rich understandingof SVW users’ goal structure. The findings show that SVW usergoals are usually overlapped with VC members’ and GVW players’

Table 4Implication matrix.

Abstractness Centrality Topics T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 Out-degrees

0.000 0.018 T1. Addiction 3 30.442 0.257 T2. Creating 3 4 4 5 1 5 2 240.800 0.060 T3. Educational tool 2 20.731 0.156 T4. Escapism 1 2 4 70.532 0.461 T5. Social relations 2 1 7 1 2 1 21 1 360.158 0.114 T6. Exploring 2 4 8 2 160.250 0.096 T7. Financial 7 2 1 2 120.722 0.108 T8. Knowledge acquisition 1 1 2 1 50.167 0.072 T9. Positive administration 3 2 3 1 1 100.725 0.413 T10. Amusement 1 2 12 1 1 2 190.195 0.246 T11. Technical features 3 1 3 18 1 2 5 33In-degrees 0 19 8 19 41 3 4 13 2 50 8 167

Out-degree: the number of times the element serves as the source or origin (means) of linkages with other elements.In-degree: the number of times the element serves as the object or end of linkages with others.Abstractness: (in-degrees)/(in-degrees + out-degrees).Centrality: (in-degree + out-degree)/the sum of all active cells.

Table 5Statistics for determining a cutoff level.

Cutofflevel

Number ofactive cells intheimplicationmatrix

Percentage ofactive cells ator above thecutoff level (%)

Number ofactivelinkages in theimplicationmatrix

Percentage ofactive linkagesat or above thecutoff level (%)

1 50 100.0 167 100.02 33 66.0 150 89.83 19 38.0 122 73.14 15 30.0 107 64.15 11 22.0 91 54.56 8 16.0 76 45.5

Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225 223

goals; yet they also have unique goals. Compared to VC members,SVW user goals embrace VC members’ goals and further include

Fig. 2. Hierarchical goal map f

distinctive goals, such as Creation, Financial, and Exploring. On theother hand, SVW users have differential goals, or Creation, Finan-cial, and Knowledge acquisition, from those of GVW players, anddo not mention achievement, which is one of GVW players’ goals.Accordingly, a SVW can be regarded as a novel cyberspace that al-lows users to engage in activities that are different from traditionalVCs and GVWs. The specific findings are summarized below:

� In terms of the topics’ centrality, Social relations (0.413) andAmusement (0.416) are predominant, and Creating (0.257) andTechnical features (0.246) follow.

� Amusement, Escapism, and Knowledge acquisition are the mostabstract topics (i.e., high-level purposes), whereas Financial,Technical feature, and Exploring are the least abstract ones. Creat-ing and Social relations play an intermediary role in the hierar-chical structure (i.e., intermediary purposes).

or a social virtual worlds.

224 Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225

� In examining linkages, we have seven findings. First, the mainchain is Technical feature ? Social relations ? Amusement. Thischain is 23.3% of all relations in the implication matrix.

� Second, Exploring is a source for Social relations and Amusement.� Third, Creating supports the highest abstraction out of three pur-

poses including Amusement, Escapism, and Knowledge acquisition.� Fourth, Social relations functions as a means for Escapism.� Fifth, Escapism leads Amusement.� Sixth, Amusement has a strong reciprocal feedback loop with

Social relations. Both can be a means or an end for each other;in other words, the both are deeply interwoven.

� Finally, Creating has a reciprocal feedback loop with Financial.

The results show that Social relations and Amusement are maingoals for using SVWs from the viewpoint of abstractness and cen-trality, indicating that SVW users have goals that are similar to VCmembers and GVW players. Social relations correspond to VC mem-bers’ major goal, and thus we can conclude that a SVW is anotherchannel for expanding social relations, as are other social cyber-spaces. Amusement is also mentioned as one of VC members’ maingoals. In the context of GVWs, Amusement is not explicitly stated asa goal (Yee, 2006); it can be considered an intrinsic goal in playingGVWs.

The most predominant relation is the link of Technical fea-ture ? Social relations ? Amusement. This result indicates that var-ious technical features (e.g., voice chatting, avatar interaction) helpusers expend social relations, which subsequently lead to amuse-ment. Ultimately, considering the strong reciprocal feedback rela-tion of Social relations and Amusement, users tend to aim atenjoyable social relations supported by technical features in SVWs.Both Social relations and Amusement are also supported by anothertopic of Exploring, which represents users’ traveling to 3D spatialplaces where users come together for dancing or social events.Thus, Exploring is directly relevant to social communications. Inaddition, traveling to fantastic places (e.g., exotic heaven) or real-like places (e.g., virtual New Orleans) may provide visitors withan experience that produce pleasure.

Escapism, which implies that users try to get out of their routineor constrained real-life environments (Hirschman, 1983), is one ofthe most abstract goals and corresponds to one of GVW players’goals. Escapism has a much higher position in the hierarchical goalmap as the end of two intermediary goals (Social relations and Cre-ating). Some users achieve the Escapism goal by making social rela-tions in SVWs. To users with limited social activities (e.g., adisabled person, a housewife caring for four children), social inter-action in SVWs can be a way to overcome their constraints in thereal world, which eventually leads to positive feelings oramusement.

Creating is a goal that distinguishes VC members’ goals fromGVW players’ goals. Despite the fact that not all SVWs enable theirusers to create virtual objects and even sell them, many SVWs doso. This behavioral autonomy is a unique characteristic of SVWsand is regarded as a central goal (3rd highest centrality). In partic-ular, Creating plays an important role of an intermediary goal thatsupports all three highest purposes in the hierarchical structure. Bycreating virtual objects, users have fun and improve their practicalskills, such as computer programming. Also, by designing theiravatars’ appearances, which may be significantly different fromtheir real appearances, or by creating imaginary virtual objects,they escape from constraints in real life.

Another point is that Creating has a reciprocal feedback loopwith the Financial goal. Besides making virtual objects to satisfytheir creative needs and imagination, users produce them partiallywith the intent to sell them in SVWs. Whether their intention is togain self-contentment or do business, their creatures become fun-damental contents of SVWs. VC research has stressed the impor-

tance of member-generated contents and further regarded themas a vital factor in VCs’ success (Filipczak, 1998). However, accord-ing to prior studies, only 10% of members of one popular peer-to-peer sharing VC produced 87% of all contents (Adar & Huberman,2000), and 4% of members in an open-source development VC pro-duced 88% of new codes and 66% of code fixes (Mockus, Fielding, &Andersen, 2002). Minor devotees produce almost all contents inpre-existing VCs, whereas SVW users consider the behavior forproducing contents, or Creating, a central goal. Therefore, mem-ber-generated contents are more prominent in SVWs than in tradi-tional VCs. Furthermore, transactional systems (e.g., virtualcurrency, virtual market for trading virtual objects) play some sortof incentive for users’ creating behavior. SVWs’ transactional sys-tems can encourage users to create contents spontaneously with-out incentive systems supported by SVW providers. Thus, thecombination of Creating and Financial has a strong self-sustainingfunction for SVWs. This finding provides current and future SVWproviders with managerial insights: advancing a creating tool, rein-forcing transactional systems, and guaranteeing securetransactions.

Knowledge acquisition is one of four of VC members’ main goals(i.e., information exchange). Knowledge acquisition is, however, not acentral goal to SVW users (the second lowest centrality), though itis one of the most abstract goals. Compared to Social relations andAmusement, Knowledge acquisition seems to be considered a minorgoal in SVWs.

6. Contributions and limitations

This study’s contributions are threefold: conceptualizing SVWs,empirical investigation of a timely topic, and finding a new ap-proach towards analyzing user goals for social cyberspaces. First,although SVWs have their own distinctive characteristics that aredifferent from traditional MMOGs (or GVWs), many studies perti-nent to SVWs have considered them to belong to one broad singlecategory together with GVWs, or VWs. This study clarifies the no-tion of SVWs through differentiating them from GVWs, and wehope that our initial conceptualization of SVWs stimulates discus-sion about what SVWs are, and how different these new cyberspac-es are from others. Second, while SVWs have dramaticallyattracted the attention of business mangers, educational practitio-ners, and researchers, there is little empirical research on SVWusers. The study offers preliminary knowledge about SVW usersby empirical investigation of user goals. Lastly, the study intro-duces a means-end chain analysis which provides a richer under-standing of SVW user goals. In prior studies that examinedindividuals’ goals for using social cyberspaces, those goals are sep-arately listed without any explanation of relations among them. Ameans-end chain analysis used in this study presented the hierar-chical goal map, which consists of separate goals and their rela-tions, and thus offers a better explanation of SVW user goals.

This study has the following three limitations. First, the study’ssamples may be biased in that the study surveyed users during ashort term, and chose a convenient sampling method. The otherpotential bias with our data is that because the samples voluntarilyresponded the survey our results may be based on highly-moti-vated users’ responses. Accordingly, the study has a limitation infully generalizing the findings. Second, the study deals with onlyone type of SVW which supports 3D interfaces and endows usersto create virtual object and sell them. Thus, the findings of thestudy should be re-examined on other SVWs, which have differentenvironments, in future research. Furthermore, in order for furtherunderstanding of SVWs user goals, future research needs to com-pare them to user goals for other various social cyberspaces, suchas conventional text-based VCs, social networking services, or

Y. Jung, H. Kang / Computers in Human Behavior 26 (2010) 218–225 225

weblogs. Finally, this study did not control variance caused by cul-ture, which may function as a crucial variable in fully understand-ing SVW users’ goals, in a sense that There is a worldwidecyberspace and the users come from various countries. Thus, cul-tural consideration can be a prominent aspect in research onSVW users’ behaviors.

References

Adar, E., & Huberman, B. A. (2000). Free riding on gnutella. First Monday, 5.Available: <http://www.hpl.hp.com/research/idl/papers/gnutella/gnutella.pdf>.

Bagozzi, R. P., & Dabholkar, P. A. (1994). Consumer recycling goals and their effecton decisions to recycle: A means-end chain analysis. Psychology & Marketing, 11,313–340.

Bartle, R. (2003). Designing virtual worlds. Indianapolis: New Riders Press.Botschen, G., & Hemetsberger, A. (1998). Diagnosing means-end structures to

determine the degree of potential marketing program standardization. Journalof Business Research, 42, 151–159.

Botschen, G., Thelen, E. M., & Pieters, R. (1999). Using means-end structures forbenefit segmentation. European Journal of Marketing, 33, 38–58.

Bressler, S. E., & Grantham, C. E. Sr., (2000). Communities of commerce. Buildinginternet business communities. New York: McGraw-Hill.

Capozza, D., Falvo, R., Robusto, E., & Orlando, A. (2003). Beliefs about internet:Methods of elicitation and measurement. Papers on Social Representations, 12,1.1–1.14.

Castronova, E. (2005). Synthetic worlds: The business and culture of online games.Chicago, IL: The University of Chicago Press.

Dreyfus, H. L. (2008). Faking it. California, March/April, 51–54.Filipczak, B. (1998). Trainers on the net: A community of colleagues. Training, 35, 70–76.Fleiss, J. L. (1981). Statistical methods for rates and proportions. New York: John Wiley & Sons.Gengler, C. E., & Reynolds, T. J. (1995). Consumer understanding and advertising

strategy: Analysis and strategic translation of laddering data. Journal ofAdvertising Research, July/August, 19–33.

Gutman, J. (1982). A means-end model based on consumer categorizationprocesses. Journal of Marketing, 46, 60–72.

Hagel, J., & Armstrong, A. G. (1997). Net gain: Expanding markets through virtualcommunities. Boston, MA: Harvard Business School Press.

Hirschman, E. C. (1983). Predictors of self-projection, fantasy fulfillment, andescapism. The Journal of Social Psychology, 120, 63–76.

Hobson, N. (2007). Should businesses get a Second Life. Knowledge ManagementReview, 10, 5.

Holmstr}om, H., & Jakobsson, M. (2001). Using models in virtual world design. InProceedings of the hawaii international conference on system sciences–41 (HICSS-41). Big Island, HI.

Jakobsson, M. (2006). Virtual worlds and social interaction design. Unpublisheddoctoral dissertation, Umeå University.

Juul, J. (2005). Half-real: Video games between real rules and factional worlds.Cambridge, MA: MIT Press.

Klenosky, D. B. (2002). The ‘‘pull” of tourism destinations: A means-endinvestigation. Journal of Travel Research, 40, 385–395.

Knoke, D., & Burt, R. S. (1982). Prominence. In R. S. Burt & M. J. Minor (Eds.), Appliednetwork analysis (pp. 195–222). Beverly Hills, CA: Sage Publications.

Kruglanski, A. W., Shah, J. Y., Fishbach, A., Friedman, M. R., Chun, W. Y., & Sleeth-Keppler, D. (2002). A theory of goal systems. Advances in Experimental SocialPsychology, 34, 331–379.

Linden Lab (2009). Linden lab goes shopping, buys virtual goods marketplaces tointegrate web shopping with Second Life. Available: <http://lindenlab.com/pressroom/releases/01_20_09>.

Locke, E. A., & Latham, G. P. (1990). A theory of goal setting and task performance.Englewood Cliffs, NJ: Prentice Hall.

MMOGchart.com (2008). An analysis of MMOG subscription growth. Available:<http://www.mmogchart.com/charts>.

Mockus, A., Fielding, R. T., & Andersen, H. (2002). Two case studies of open sourcesoftware development: Apache and Mozilla. ACM Transactions on SoftwareEngineering and Methodology, 11, 309–346.

Newell, A., & Simon, H. (1972). Human problem solving. Englewood Cliffs, NJ:Prentice-Hall.

Olson, J. C., & Reynolds, T. J. (1983). Understanding consumers’ cognitive structures:Implications for marketing strategy. In L. Percy & A. G. Woodside (Eds.),Advertising and consumer psychology (pp. 51–57). Lexington, MA: LexingtonBooks.

Peffers, K., Gengler, C. E., & Tuunanen, T. (2003). Extending critical success factorsmethodology to facilitate broadly participative information systems planning.Journal of Management Information Systems, 20, 51–85.

Pervin, L. A. (1989). Goal concepts in personality and social psychology. Hillsdale, NJ:Lawrence Erlbaum.

Pieters, R., Baumgartner, H., & Allen, D. (1995). A means-end chain approach toconsumer goal structures. International Journal of Research in Marketing, 12,227–244.

Reynolds, T. J., & Gutman, J. (1988). Laddering theory, method, analysis, andinterpretation. Journal of Advertising Research, 28, 11–31.

Reynolds, T. J., & Rochon, J. P. (2001). Consumer segmentation based on cognitiveorientations: The ChemLawn case. In T. J. Reynolds & J. C. Olson (Eds.),Understanding consumer decision making? The means-end approach to marketingand advertising strategy (pp. 283–298). Mahwah, NJ: Lawrence Erlbaum.

Reynolds, T. J., & Whitlark, D. B. (1995). Applying ladder data to communicationsstrategy and advertising practice. Journal of Advertising Research, 35, 9–17.

Rheingold, H. (1993). The virtual community: Homesteading on the electronic frontier.Cambridge, MA: MIT Press.

Ridings, C. M., & Gefen, D. (2004). Virtual community attraction: Why people hangout online. Journal of Computer-Mediated Communication, 10 [Article 4].

Rifkin, A. (1985). Evidence for a basic level in event taxonomies. Memory andCognition, 13, 538–556.

Scott, J. (1991). Social network analysis: A handbook. London: Sage Publications.SLOG (2008). Second Life Statistics: 22-Feb-2008. Accessible: <http://

secondslog.blogspot.com/2008/02/second-life-statistics-22-feb-2008.html>.van Rekom, J., van Riel, C. B. M., & Wierenga, B. (2006). A methodology for

assessing organizational core values. Journal of Management Studies, 43,175–202.

Voss, R., Gruberb, T., & Szmigin, I. (2007). Service quality in highereducation: The role of student expectations. Journal of Business Research,60, 949–959.

Walker, B. A., & Olson, J. C. (1991). Means-end chains: Connecting products withself. Journal of Business Research, 22, 111–118.

Warburton, S. (2009). Second Life in higher education: Assessing the potential forand the barriers to deploying virtual worlds in learning and teaching. BritishJournal of Educational Technology, 40, 414–426.

Wellman, B. (1996). For a social network analysis of computer networks: Asociological perspective on collaborative work and virtual community.Presented at the ACM SIGCPR/SIGMIS conference on computer personnelresearch, Denver, CO.

Yee, N. (2006). Motivations of play in online games. CyberPsychology & Behavior, 9,772–775.