24

Design for Contribution

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Design for Contribution

Design for ContributionTheoretical and Practical Approaches

to Motivate Contribution in Online Communities

Roman Eul

July 21, 2008

Abstract

Under-contribution poses a main problem for most online communi-

ties. Aim of this seminar paper is to outline design recommendations

based on current �ndings about motivating contribution, which requires a

thorough understanding of the determining factors of contribution. This

paper seeks to approach the issue of under-contribution based on a sound

theoretical background. Thus, based on a discussion on some theoretical

frameworks for increasing contribution, this paper focuses on reviewing re-

search articles that apply conclusions of those theories in social software

environments.

1

Page 2: Design for Contribution

Contents

1 Introduction 3

2 The Nature of a Community 4

2.1 A Structural Model for Communities of Practice . . . . . . 4

3 Increasing Participation - Theoretical Frameworks 7

3.1 Gerneral Design Principles . . . . . . . . . . . . . . . . . . . 7

3.2 Kollock's Framework . . . . . . . . . . . . . . . . . . . . . . 9

3.2.1 Maslow's Hierarchcy of Needs . . . . . . . . . . . . . 10

3.2.2 Motivations for Contribution - The Framework . . . 10

3.2.3 Implications . . . . . . . . . . . . . . . . . . . . . . . 12

3.3 Bishop's Framework . . . . . . . . . . . . . . . . . . . . . . 12

3.3.1 The Ecological Cognition Framework . . . . . . . . . 12

3.3.2 Implications . . . . . . . . . . . . . . . . . . . . . . . 15

4 Increasing Participation - Practical Approaches 16

4.1 Motivating Contributions Through Framing Uniqueness and

Bene�t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.1.1 Experimental Setting . . . . . . . . . . . . . . . . . . 16

4.1.2 Expected Outcome . . . . . . . . . . . . . . . . . . . 17

4.1.3 Data Analysis . . . . . . . . . . . . . . . . . . . . . . 17

4.1.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . 18

4.2 Motivating Participation by Displaying the Value of Con-

tribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4.2.1 Experimental Setting . . . . . . . . . . . . . . . . . . 19

4.2.2 Expected Outcome . . . . . . . . . . . . . . . . . . . 20

4.2.3 Results and Discussion . . . . . . . . . . . . . . . . . 20

5 Conclusion 22

2

Page 3: Design for Contribution

1 Introduction

With the increasing popularity of social internet communities and social soft-ware, information and communication technology (ICT) is changing the way wecommunicate, consume, interact and entertain. In the course of that develop-ment, traditional social forums have been moving their activities to the internetto utilize on the nearly unlimited amount of social capital provided by ICT.Since the �rst steps of Usenet news sharing programs in the early 1980s onlinecommunities have evolved to mass phenomena, that are used by millions of usersall around the globe to share information, music, videos, pictures, and jokes,discuss mutual interests, play games, and conduct business. Various forms ofonline media, such as wikis, blogs, chat rooms, internet forums, or electronicmailing lists enable users to do so. Wikipedia, Facebook, and immeasurable ofmore or less popular examples, illustrate the vibrancy of online communities.

However, despite their popularity, large numbers of online communities failand disappear from the screen shortly after launch due to under-contributionor even non-participation. The following examples illustrate that issue: In [1],Butler found, that 50% of social, hobby, and work mailing lists lied idle overa 122-day period. Moreover, under-contribution is also an issue in successfulonline communities. Ortega [2] found that merely ten percent of all Wikipediaauthors (i.e. not users) contribute over 90% of all content throughout any of theten most-used languages. Bearing in mind that users face the lowest imaginableentrance barriers for writing or editing articles, the vast gap between active andpassive users stresses the need for scrutiny. In the same way, the P2P sharingnetwork Gnutella su�ers heavily from free-riding and relies on those ten percentof users who share 87% of all content [3].

Even this small range of examples, which is extendable at will though, al-legorises under-contribution as the core issue to all community based socialsoftware, whether accessible to everyone via internet or exclusively available toselected group members via intranet. All these facts and circumstances urge thequestion whether community based software is inevitably condemned to under-contribution, if not failure, or whether it is possible to challenge design featuresof community sites to sustainably improve participation from larger fractionsand generate ongoing contribution. Concurrently, a lot of research is done toinvestigate this core question and thus this paper aims to review recent �ndingsto conclude their potentials and limitations.

Hence, this paper starts o� with a chapter about typical features of socialcommunities with the intention to enlighten why and how they exist. There-upon, an introduction to theoretical frameworks of motivation explains the ratio-nale behind human action in communities, and paves the way for the subsequentreview of current research experiments that aim to improve contribution.

3

Page 4: Design for Contribution

2 The Nature of a Community

Since the de�nition of a community used by scientists varies depending on theirresearch background, we have to get a clear understanding of this concept beforewe can turn towards answering the central question of under-contribution in thediverse types of communities. Traditionally, the term community refers to �aninteracting population of various kinds of individuals in a common location�[4]. However, this de�nition requires speci�cation to be applicable to modernforms of communities, as for instance innovations in information and telecom-munication technology made a common physical location obsolete and directinteraction is no longer characteristic to such a community. Hence, researcherscame up with concepts like community of practice, community of interest, com-munity of action, and so forth, that allow identifying speci�c communities withmore accuracy. All together, these concepts enhance the term of community.

In the context of this paper a community refers to a community of practice(CoP) which Wenger in [5] describes as a �group of people who share a concern,a set of problems, or a passion about a topic, and who deepen their knowledgeand expertise in this area by interacting on an ongoing basis�. Additionally, werestrict the CoP to the �eld of software or social software, respectively. Thisde�nition allows encompassing all kind of structures that enables people to shareand acquire knowledge. Even though this de�nition of a CoP narrows down thefocus of the common de�nition, it still leaves space for variation, which ranges forinstance from classical knowledge management systems (groupware) to modernforms of online communities.

Within the de�nition of the community of practice, Wenger (p. 24-27) enu-merates distinguishing and characterizing factors as the following:

• Small or big : Describes whether a CoP consists of only a few members orlarge amounts of people.

• Long-lived of short-lived : Describes the variation of a CoP's life span.

• Collocated or distributed : Describes the physical distance between themembers of a CoP.

• Homogeneous or heterogeneous: Describes the di�erences of the members'cultural background.

• Inside and across boundaries: Describes the possible interaction with ac-tors outside the community.

• Spontaneous or intentional : Describes whether the CoP emerged or wasplaned from scratch.

• Unrecognized or institutional : Describes the level of formal organisationalstructures.

2.1 A Structural Model for Communities of Practice

Despite the diversity of CoPs that are permitted by the above-mentioned factors,all CoPs share a basic structure. Once all the collective elements of this structureare identi�ed, they constitute a common model to all CoPs. Wenger (p. 27-47)outlines three structural elements:

4

Page 5: Design for Contribution

1. DomainThe domain of a community creates the common scope of operation anda sense of identity. It should be well de�ned in order to clearly displaythe communities' purpose and state the value of it to the contributingmembers and other stakeholders. Creating the domain also implies to setits boundaries. Consequently, the domain motivates the members to con-tribute and participate as it enables them to decide what or what not toshare and in which way to do so. Without a domain, a community wouldbe just a group of friends. A shared domain, though, creates the sense ofresponsibility that gives life to a community of practice. Whether a CoP israther formalised about its domain depends on its member's understand-ing and the level potential embedding in o�cial company structures.For instance, a member a software development team for internal businessprocesses knows exactly what he has to contribute to their internal Wikifor technical documentations as he can see the implication of his work toother group members. That contribution could be anything from raisinga question of common interest, posting a solution to a speci�c problem orwriting technical documentation to new product developments. Addition-ally, it is important to notice that a domain is not a �xed set of problems,but that it evolves with the community and the world around. For ex-ample, the upcoming trend of using asynchronous JavaScript and XML(AJAX) in web pages , might have caused vivid exchanges of opinionsamong the 1und1's developers and consequently led to new challenges intheir community.

2. CommunityThe community creates the social framework of learning by fostering in-teractions and relationships based on mutual respect and trust. A strongcommunity can be the breeding ground for a willingness to share ideas,ask di�cult questions, and listen carefully. Moreover, it can create a mix-ture of intimacy and openness to inquiry which is a particularly importantelement in mixed groups with di�erent cultural backgrounds.For instance, one could think of a collaboration project between a univer-sity research group and a private company. The two sides would have mostprobably di�erences in methodology, terminology, mission, and so forth.Additionally, geographic distance would make working and personal re-lationships di�cult, which could, among others, result to redundancy ofe�orts and ine�ective transfer of project work from one side to the other.Therefore, the community concept aims to build trust and relationshipbetween the members in order to overcome the potential issues. Practicalmeasures of a structural facilitated community could be periodical face-to-face meetings, organised telephone conferences, and so forth.Although the membership in a community may be self-selected or assigned,the success of a community depends highly on the individual commitmentof every member, which cannot be enforced via top-down instructions.Instead, participation (unlike membership) is voluntary and depends oneach member's willingness to participate. Of course, participation mayvery well be encouraged, as shown in section 4.

3. PracticeWhile the domain of a community sets the topic the community focuses

5

Page 6: Design for Contribution

on, the practice is the speci�c knowledge the community develops, shares,and maintains. In that sense, practice represents the body of knowledgeof the community, that kind of knowledge everyone is expected to master.Hence, it encompasses all frameworks, ideas, information, tools, languages,stories, and documents that community members share. Each elementmay include both tacit and explicit aspects of the community's knowledge.In case of 1und1's software developer community, the wiki for technicaldocumentation is such a tool of practice. Within that tool, everythingshared is part of the community's practice as well. Here, elements ofpractice range from rather concrete and explicit objects like procedureguides, best practices, and learned lessons, to less tangible and tacit thingslike a code of behaviour or discussion culture.

These three elements constitute the basic model of a community and reinforcethe de�nition of a community of practice: Not every community is necessar-ily a community of practice, while not every practice inevitably brings out acommunity. Despite its theoretical approach to understanding the nature of acommunity, this model allows concrete implications.

Firstly, it allows distinguishing between communities and other social struc-tures, like informal networks, business departments, operational teams, or projectteams. Among others, the determination can be based on the speci�c structure'spurpose, composition of members, its boundaries, or what holds them together.However, in this paper we simply use this model as a common understandingof the term community and accept that there are various ways of organizingcollaboration besides communities. Secondly, we can derive general design prin-cipals for successfully creating a community, which will be discussed in detail inthe next section.

6

Page 7: Design for Contribution

3 Increasing Participation - Theoretical Frame-

works

Having introduced a clear understanding of a community, it is the next step'stask to extract the factors and principles that lead to a vivid and thrivingcommunity. Since the early days of groupware applications, which providedcommunities with the infrastructure to overcome physical distances, many au-thors conducted research in the �eld of communities and particularly they wereconcerned with the question of how to increase participation.

As communities are as diverse as their domain, their members and theirpractice, it is not surprising that researchers of several areas of expertise sup-plied their ideas about increasing contribution. Hence, this section will presentframeworks stemming from di�erent perspectives. Firstly, the seven principlesfor cultivating a community enumerated by Wenger et al [5] show up an ap-proach based observation, while subsequently the frameworks of Kollock [6] andBishop [7] originating from social science are introduced.

3.1 Gerneral Design Principles

In [5] Etienne Wenger and his co-authors Richard McDermott and WilliamSnyder suggest seven principles that tackle the issue of under-contribution atthe design phase of a community. They base their insight on their experiencethey gained during their work with communities in business environments andresearch projects (p. ix - xii, 51). In that sense these principles derived fromtheir observations �embody (their) understanding of how elements of designwork together�.

1. Design for evolutionAs communities start usually in rather small sized dimensions and growdynamically over time, its design should be able to go along with the com-munity's evolution and adapt where and when appropriate. The more newmembers a community generates the more new ideas they bring in and thefocus of the community may shift. Appointing community coordinator orproblem-solving meetings are concrete measures to facilitate evolution.Wikipedia for instance employs such measures with Jimmy Wales beingthe so-called �benevolent dictator� and elected community members form-ing the �arbitration committee� [8].

2. Open a dialogue between inside and outside perspectivesThis principle is based on the observation, that only people inside thecommunity may have the full understanding of its structure, knowledgeand current problems, but an outsider is more likely to be unbiased incommunity matters. Consequently, fostering a dialogue between peoplefrom the inside of one community and members of other communities oroutsiders enables them to see new opportunities or recognise redundancies.For instance, bringing together an open source development communitywith business experts may help the developers meeting the demands ofpotential users more e�ciently.

3. Invite di�erent levels of participationDue to the fact that people participate in communities for di�erent rea-

7

Page 8: Design for Contribution

Figure 1: Degrees of Community Participation (Wenger: 57)

sons, it is unrealistic to encourage every member to participate equally.Commonly, there are three main levels of community participation. Mem-ber of the �rst level are called the core group or the heart of a community.People in that group actively participate in discussions, debates, commu-nity forums, initiate new projects and identify topics for the communityto address. This group is rather small, though, it consists of merely 10 to15 percent of the community. The next 15 to 20 percent of members whoparticipate actively but without the intensity and regularity of the coregroup, form the so-called active group. The largest proportion of the com-munity forms the peripheral group. In some communities theses membersare labelled as lurkers or free-riders, due to their observing and passivebehaviour. Finally, outside these three groups are people (outsiders), thatare not part of the community but that have interests in the communityThere is no rule that assigns members to one of the speci�c groups. Bound-aries between the three groups do not exist as such, they are imaginaryand rather arbitrarily chosen. Members may temporarily or permanentlychange groups.

The community pioneer Amy Jo Kim suggests in [14] a similar structureof communities, which is often referred to in the context of online commu-nities. According to her so-called membership life cycleevery (potential)community member starts as an visitor, or lurker. Form there, the mem-ber can go along in the cycle, which stages are as follows:Peripheral (i.e. Lurker) � Inbound (i.e. Novice) � Insider (i.e. Regular) �Boundary (i.e. Leader) � Outbound (i.e. Elder).

4. Develop both public and private community spacesMostly, collaboration in a community takes place in public places. Thatis both, face-to-face interaction in the form of meetings or conferences as

8

Page 9: Design for Contribution

well as all kind of public interaction based on electronic tools, like thecommunity's online discussion board. Public interaction is important asthe exchange of new ideas, issues, or techniques is accelerated and madeaccessible to everyone. However, the public nature might prevent somemembers from participating, since some may fear their name is takenin vain. Hence, the community design must allow communication andcollaboration in public as well as private spaces.Discussion boards for instance feature also private messaging beside theirscore function public postings. At 1und1 every worker's private page o�ersa range of direct communication, such as email, instant messenger, orphone number, to get into contact privately.

5. Focus on valueBecause participation is voluntary in most communities, it is essential for athriving community to display the value it has generated. Since the valueof a community is not apparent to every member by its merely knowingits structure (domain, community, practice), people needs to be remindedon a regular basis to keep participation on a high level.

6. Combine familiarity and excitementCommunities should o�er the comfort environment of a home, but alsohave enough interesting events to keep innovation on new ideas high undthe community attractive to potential new members. While in the begin-ning the excitement is naturally high, the danger of becoming too familiarand less exciting rises normally with ongoing maturation of the commu-nity. Hence the community needs to deploy countermeasures actively.For instance, in order to foster divergent thinking, external experts andthe community are brought together to discuss current issues. Initiatinga discussion about publishing Windows APIs between an open source de-velopment community and a Microsoft representative will provide plentyof excitement and participation.

7. Create a rhythm for the communityCommunities need to have a rhythm. Regular meetings, conferences, dead-lines, objectives, goal, etc. set the pace of a community. When the paceis too fast, people feel overwhelmed and participation declines. When thepace is too slow, people lack of challenge, get bored and participationdeclines. Thus, a community needs to �nd its appropriate pace over time.

3.2 Kollock's Framework

While Wenger et al. utilize their practical knowledge to address the question ofsuccessfully designing a community, researchers like Peter Kollock and JonathanBishop take a scienti�c approach. In [6], Kollock analyses online cooperationfrom an economic point of view. In order to assess the di�erence betweenpublic goods and gifts in online cooperation and the traditional conception, heextracts the motivating factors that are driving cooperation in a community.As the underlying motivation theory he employs a content-related need basedtheory. Hence, we brie�y introduce the main concept of Maslow's hierarchicalneed theory, before going into the details of Kollock's framework.

9

Page 10: Design for Contribution

3.2.1 Maslow's Hierarchcy of Needs

Figure 2: Maslow's Hierarchy of Needs

In [9] Maslow's proposes a theory of psychology that aims to explain humanbehaviour based on needs every human wants to ful�l. These needs are hier-archical, meaning a human has to satisfy his lowest need �rst before movingon satisfy the higher ones. Therefore, Maslow's hierarchy is often depicted asa pyramid (see �gure 2). The four lower layers of the pyramid are groupedtogether as the physiological needs or de�ciency needs, while the top layer isreferred to as psychological need. De�ciency needs must be met �rst. Once thelower needs are met, an individual's needs grow on the next layer, which he thenseeks to satisfy.

3.2.2 Motivations for Contribution - The Framework

In [6] Kollock outlines three motivational factors for contributing to online com-munities, which are enhanced by Smith in [10]. Thereby, he assumes that mo-tivation rests on self-interest (i.e. satisfying needs) and thus does not take anypure form of altruism into consideration. Accordingly, his model suggests thatany action is motivated by satisfying personal needs and hence every factor ofthe enhanced model corresponds more or less clearly with one level of Maslow'shierarchy of needs.

1. Anticipated reciprocityOne possibility is that an actor is motivated to contribute valuable in-formation to a group in the expectation that he will receive useful infor-mation in return. Since the actor does not expect the return immediatelythe exchange of information is asynchronous and hence this motivation forcontribution is called anticipated reciprocity. The whole system of asyn-chronous information exchange can be regarded as a network-wide creditaccounting system, in which accounts do not need to be kept continuallyin exact balance. The single actor and the community in whole mutu-ally bene�t from that system. The community is better o� as its bodyof knowledge expands, because its members share information freely anddo not retain it until their account is balanced. To the member himself

10

Page 11: Design for Contribution

the loose accounting system works as a kind of insurance. When he is inneed, he can draw upon the communities resources, without the need toimmediately repay each person.Since this system works as a kind of insurance, it helps to satisfy a user'sneed for safety.

2. Sense of communityIn [10], Smith enhances Kollock's framework with the second motivation,i.e. the sense of community. It states that people, in general, are socialbeings and it is motivating to many people to receive direct responsesto their action. Most online communities enable this by allowing peopleto reply directly to contributions. Fairly often, this motivational factoris supported by communities through, among others, the possibility tocomment directly to other's contributions or value that contribution.

3. Increased recognitionThe third possible motivation is the in�uence of contribution on an actor'sreputation. Providing high quality information or insight knowledge, ando�ering help to others can all lead to increase one's prestige within thecommunity. According to Maslow's hierarchy of needs increased recogni-

tion satis�es a person's need for esteem. Thus, it is likely that contribu-tion's increase if they directly work to improve the contributers recogni-tion.Many online communities have developed features that display its mem-ber's level of contribution. With that information users can easily becompared and ranked according to their status. Very often the level ofcontribution comes along with a system that assigns users into di�erentgroups. At Amazon for instance, the most active users become part of theTOP500, TOP 100 or even TOP10 group, which boldly displays a user'srecognition. That makes comparison a lot easier and thus creates an ad-ditional incentive to contribute more. Similar features can be found atvirtually all community based website like Ebay, Youtube, Myspace, etc.Another example illustrates the importance of recognition. When hackersare involved in illegal activities they need to protect their identity by usingpseudonyms. Even though the repeated use of pseudonyms increases thethreat of being traced by the authorities, hackers are reluctant to changetheir nicknames, as the recognition cumulates with that name.

4. Sense of e�cacyThe fourth possible motivation is that a person contributes valuable in-formation to a group because he can see the e�ect the contribution has onhis environment. If that sense of e�cacy is motivating someone to act,it is likely that contribution increases to the extent that the e�ect of thatwork is recognisable. In hierarchy of needs the sense of e�cacy satis�esthe need of self-actualisation. Unlike the need of esteem, an actor's recog-nition may very well increase through an action motivated by the sense ofe�cacy, but that is not the actual motivation.A contribution to Wikipedia is a good example illustrating motivationbased on the sense of e�cacy. Anonymously, new articles can be createdor existing articles can be edited. However, the results of the work takee�ect immediately.

11

Page 12: Design for Contribution

Figure 3: The Ecological Cognition Framework ([6])

3.2.3 Implications

According to Kollock's framework and the underlying need-based theory, anactor's motivation to contribute depends on the need he pursues to satisfy.Two immediate implications can be derived from that: Firstly, members of acommunity should not be addressed equally in order to increase participation.This implication is identical to Wenger's third design principal (see page 7)but here explained with the di�erent needs members want to satisfy. Secondly,this model states that people want to participate when their contribution helpsto satisfy needs. Hence, a community must be (re-) designed in the way thatis makes clear to the users that their contribution would actually satisfy theirneeds. Section 4 describes some studies that practically approach the issue ofunder-contribution based on that implication.

Just like Kollock a number of authors reason their frameworks to improvecontribution on need-based theories. That is, however, not the only way ofreasoning, as the following section demonstrates.

3.3 Bishop's Framework

In contrary to Kollock, Jonathan Bishop, another virtual community pioneer,does not believe that models employing need-driven theories are comprehensiveenough to explain the collaboration in online communities. Hence he developeda new framework from scratch, which he based on desire being the driving forcebehind motivation.

3.3.1 The Ecological Cognition Framework

Bishop proposes a 3-level framework for understanding why members ofonline communities do or do not participate (Figure 3). On the �rst level he

12

Page 13: Design for Contribution

describes what drives individuals to carry out actions. On the second level heoutlines the factors that determine whether or not such actions are actuallycarried out, and on level 3 he identi�es the means by which they are executed.This section introduces the ecological cognition framework and its three levelsin particular, subsequently followed by a section about the implications of thatmodel.

Level 1 of the model consists of an actor's potential desires. These are indetail:

• Social : Includes the desire to become part of the community throughsocializing and communication with other actors.

• Order : Includes the desire to take and retain control of situations as wellas the desire to arrange and sort things around oneself.

• Existential : Includes the desire for food, water and shelter.

• Vengeance: Includes the desire to respond appropriately to unpleasantattacks.

• Creative: Includes the desire to live out one's personal creativity.

Level 2 of the framework entails an actor's cognitions, which are in particular hisgoals, plans, values, beliefs, and interests. The ecological cognition frameworksuggests that an actor will seek to achieve consistency with his cognition whencarrying out an action. The cognition in detail is de�ned as follows:

• Goals are short-term or long-term objectives, ideas, or targets an actorstries to achieve.

• Plans are conceived strategies, which are stored in an actor's memory,describing how to act when a desire is experienced.

• Beliefs are an actor's perception of what is true or false.

• Values are clearly de�ned principles that represent an actor's degree ofimportance.

• Interests are connections with something or somebody that the actor at-tempts to maintain.

Level 3 of the model encompasses an actor's means to interpret and interactwith his environment. These are the following abilities:

• Olfactory : sense of smell

• Auditory : sense of hearing and acting through speaking

• Haptic: sense of touch and interaction through touch

• Gustatory : sense of taste

• Visual : sense of sight and imagine visual images

13

Page 14: Design for Contribution

Finally, an actor's environment is made up of other actors, artefacts, structuresand all other things surrounding him.

On the basis of this model, Bishop provides for each level a principle ex-plaining the interdependency between the di�erent levels and �nally answeringthe question why people participate in online communities.

1. Principle - an actor is driven to act by his desires

With his �rst principle Bishop opposes the thought that people performactions in order to satisfy needs, as he considers need-based theories as notappropriate for online communities. Instead, actors are driven to act bytheir desires. In the context of this model �desires are thoughts or requestsfor action that an actor experiences�, but �not responses to emotions, asthey are not su�ciently connected with feelings�. This understanding alsoexcludes the interpretation of desire as something that can be satis�ed inthe way how Maslow describes needs. That means, desires must be actedout and not satis�ed.In the perspective of this model, a member of the active or core group(Figure 3) is referred to as an elder. Replying regularly to other memberspostings would be a typical pattern of an elder's behaviour. In doing so,he acts out several desires. By conceiving solutions to other membersproblems, an elder acts out his desire for creativity, and by posting hissolution and comments he acts out his social desires. The e�orts of manyusers to ensure that postings or whole discussions do not go o�-topic, canbe interpreted as acting out the desire for order, while '�aming' or negativecomments may be explained by the desire for vengeance.However, accepting an actor's desire as the driving force to act, doesnot explain why a member of the peripheral group does not participatedespite his desire for creativity, being social, etc. Whether or not anaction is performed after all determines the second principle derived fromthe framework.

2. Principle - an actor's desire to act is limited by his goals, plans, values,

beliefs and interests

According to the second principle an actor will take their existing plans,goals, values, beliefs, and interests into account before acting based ontheirs desires. The ecological cognition framework suggests that once anactor has experienced a desire he conceives a plan to act out that desire.However, if he then �nd this plan inconsistent with his cognition (existingplans, values, etc), he will refrain from acting out his desire according tothat plan.If for instance a lurker (i.e. a member of the peripheral group) desires tobe social, he develops a plan to communicate with other actors in orderto help them with their current issues. Even so, in case he believes hiscontribution will not help the other members, he will not act out his planafter all.As contrary to need-based theories (e.g. Maslow), desires are not arrangedin a hierarchical structure. That means, that there is no preference amongthe desires on level one, so only level two determines which desire currentlyantecedes. If an actor want to communicate (social desire) and eat (exis-tential desire) at the same time, he will act according to what he believes

14

Page 15: Design for Contribution

is more important at the moment.Thus, principle 1 and 2 explain why an actor decides to take action: Hewants to act out his desire, in order to do so he needs a plan consonantwith his cognition. Remains the question how he is going to put his planinto action.

3. Principle - an actor will act based on how he perceives his environment

As a result to a desire to do something, an actor develops a plan consonantwith his cognition. The next step is now to interact with his environment.In order to do so, an actor can use all means provided by level three.The ecological cognition framework suggests that the perception of hisenvironment, which consist of among other things, artefacts, structures,and actors, determines how an actor will perform his action.For instance, an elder wants to contribute a solution to an other membersproblem because he can act out his desire for creativity and being social.However, his environment determines how he is going to carry out hisplan. If the structure is an online discussion board, he will most probablypost the solution or send a private message. If the other member is anovice he might elaborate on details or add visuals such as screenshots.

3.3.2 Implications

The ecological cognition framework argues that the experience of a desire mo-tivates an actor to carry out action, such as contributing a solution to anotheractor's problem (level 1). Then he evaluates whether taking this action is con-sistent with his cognitions (level 2). If yes, he carries about the action accordingto his abilities (level 3) and in�uenced by his environment.Remains to evaluate in how far one can draw conclusions from that model. Asso far, this model has only been treated in theory, �ndings from the ecologicalcognition framework were not yet applied in experiments. Hence, this is still anopen question an will be discussed in further detail in section 5.

15

Page 16: Design for Contribution

4 Increasing Participation - Practical Approaches

4.1 Motivating Contributions Through Framing Unique-

ness and Bene�t

According to the theory of Kollock's framework people contribute valuable infor-mation to a community due to di�erent motivation. In [3] Beenen et al. try toassess the propositions posited by theories of social science like those of Kollock.Therefore, they chose a online community in order to validate whether the the-oretical propositions are supported in an experiment under realistic conditions.In particular they examined the salience of uniqueness, the salience of bene�tand the combination of both. By displaying the uniqueness of a contributionthe researchers want to call on the user's sense of e�cacy. And by framingthe bene�t a contribution has to others and to the user himself, respectively,they remind the user on their sense of community and the reciprocity of theirparticipation.

4.1.1 Experimental Setting

The experiment was realised a community website called MovieLens [11]. Movie-Lens is a web-based movie recommendation community, where members can ratemovies, write movie reviews and receive recommendations for movies. Based onthe user-submitted movie ratings, MovieLens generates personalized recommen-dations to that user with movies he will like or dislike. The technique they useto recommend movies is called collaborative �ltering, which works by matchingtogether users with similar opinions about movies. Each member of the commu-nity has a so-called neighbourhood consisting of other users with a similar tasteof movies. Ratings from these neighbours are used to generate the personalizedmovie recommendations. Despite having more than 80.000 users and millionsof ratings, more than 20% of the movies listed in the system have so few rat-ings that the collaborative �ltering algorithm cannot make accurate predictionsabout them [12].

The researchers rolled out a movie rating campaign among those users, whichwere active in the last six month before the experiment and which make upsome 10% of all users. To recruit users to participate in that campaign, anindividualized email was send to all the active members. The content of themessages varied depending on the need the users should be reminded on. Thenthe contribution behaviour was tracked for the one week following the invitation.

The email contained a standard invitation text and two variables, which werecalled uniqueness and bene�t. The uniqueness variable could be set to a messagestating the importance of that speci�c user's contribution due to his uniquetaste. While the bene�t variable could be set to a message either emphasisingthe bene�t of the user's contribution to the community (�Rating more movieshelps the MovieLens community! ...�) or the bene�t his ratings have to himself(�Rating more movies helps you! ...�). By the means of those variables thesubjects were divided into groups that where reminded on either uniqueness,others-bene�t, self-bene�t or both uniqueness and bene�t. A separate controlgroup received only an invitation email without any of the two variables set.

16

Page 17: Design for Contribution

4.1.2 Expected Outcome

Based on social theory they constructed the following hypotheses:

1. Hypothesis: MovieLens users will rate more movies when the uniqueness

of their contributions is made salient.

Kollock's model posits that people' sense of e�cacy is one motivation forpeople to contribute valuable information. That means, if a user can seethe e�ect of his contribution to his environment his sense of e�cacy willmotivate him to act. Hence, if the users think their contribution is unique,they should be more motivated to participate, because their contributionsare more likely to have a recognizable outcome.

2. Hypothesis: MovieLens users will rate more movies when the personal

bene�t they receive from doing so is made salient.

Kollock also suggests that reciprocity is an important motivator. Accord-ingly, users will rate more movies if they can see that they get somethingin return for their contribution, i.e. they bene�t. MovieLens' collaborative�ltering system uses its user's ratings to predict how much a subscriberwill like a movie. The more movies a user rate the more precise are themovie recommendations made by MovieLens. Hence, users will rate more,if the personal bene�t is made salient

3. Hypothesis: MovieLens users will rate more movies when the bene�t they

provide to the community is made salient.

The community as a whole bene�ts from rating movies as well, becausethe amount of movies and the accuracy of recommendations increases.However, members may not be aware of the importance of their contribu-tion to others. But since the sense of community is one of the motivatorsfor participation proposed by Kollock, making explicit the bene�t thatthe community receives from the user's contribution should increase theirratings.

4. Hypothesis: MovieLens users will rate more movies when the perception of

both unique contribution and bene�ts to the community are made salient

than when only unique contribution or bene�ts are made salient.

Since both the salience of a user's uniqueness and the salience of bene�tlead to increased contribution, Beenen et al. conclude that a combinationof both variables brings about additional participation among the respec-tive users compared to those users to whom only one variable was madesalient.

4.1.3 Data Analysis

During the observation period, considerably more users logged on to MovieLensand rated movies. Among the invited users, the average rating rate jumpedfrom the six month average of 5.4 movies per week to some 20 movies per week.As this numbers also apply to the members of the control group, merely sendingthe email led to a signi�cant increase in ratings. Regarding the hypotheses, thebreak down of the data supplies ambiguous results (Table 1).

Hypothesis 1 stated that making the uniqueness clear would increase themembers' motivation to rate more movies. This is supported by the results

17

Page 18: Design for Contribution

No Bene�t Bene�t to self Bene�t to othersNon-unique 25.26 13.19 13.22Unique 31.53 19.28 9.48Total 28.28 16.1 11.33

Table 1: Mean Number of Movie Ratings (adapted from [3])

of the experiment. On average members of the unique group rated 18% moremovies than those who received the non-unique message.

Hypothesis 2 stated that emphasising personal bene�ts would increase themember's motivation to rate more movies. Surprisingly, mentioning the indi-vidual bene�t depressed the user's participation. As the rating rate deviatessigni�cantly from the mean value of both the unique and non-unique group, the�ndings of this study discon�rm Hypothesis 2.

Just as Hypotheses 2, the assumption made in Hypothesis 3 that the salienceof bene�ts to the community would increase participation was discon�rmed inthis experiment. People who received either the self-bene�t or the other-bene�tinformation rated less movies than those who did not receive that information.This requires further discussion as it contradicts the assumptions based of theunderlying theory (see below in 4.1.4).

Bearing in mind that uniqueness was con�rmed as a motivating factor andbene�t, whether to oneself or others, was discon�rmed, the result of the com-bined group is as is was expected: Hypothesis 4 can not be supported by sta-tistical evidence, i.e. it can neither be con�rmed nor discon�rmed.

Table 2 summarises the study's results:

Hypothesis Results

H1: Unique group rates more than non-unique group + (supported)H2: Bene�t-to-self group rates more than no-bene�t group - (discon�rmed)H3: Bene�t-to-others group rates more than no-bene�t group - (discon�rmed)H4: Comined Group o (not supported)

Table 2: Results Summary (apapted from [3])

4.1.4 Discussion

Firstly, the most obvious result of the experiment should be stressed: By simplyreminding the community members via email the level of participation increasedconsiderably. The more interesting result is, though, that the content of the mes-sage made a di�erence. Partly, social science can help to explain the di�erence,especially when it comes to the salience of uniqueness. Even so, in the case ofbene�t, the results are inconsistent with the theory based predictions.

This being the case, either the underlying theory is inappropriate for onlinecommunities or the implementation was �awed. The authors assume that themembers of the active user segment are anyway highly committed and alreadyaware of the bene�ts of their work. Sending these people a message highlightingonly one side of the bene�t (self- or group-bene�t) might have undermined theirprevious perception of their work and thus narrowed down their focus. Other

18

Page 19: Design for Contribution

possible explanations stated that the emails were either poorly phrased or toolong to read carefully and thoughtfully.

However, all possible explanations for the deviation between the predictedand actual results cannot be underscored with any data from the original ex-periment and thus remain speculative. In the following section, another teamof researchers is presented, that approached the particular question of bene�tby applying di�erent means.

4.2 Motivating Participation by Displaying the Value of

Contribution

In [13] Rashid et al. continue the research presented by Beenen et al (seeabove in section 4.1). Although Beenen et al. were able to produce signi�cantchanges in user behaviour, the results were mixed and deviated partly fromthe hypotheses. One result of their study was that the mere email informationmotivated already most members to increase participation. On the other hand,the authors suspected the phrasing of the email having a major in�uence onthe dissatisfying results. In order to avoid these issues, Rashid et al. changedthe interface of MovieLens directly. In the new design the user could see thebene�t of his rating right away. Moreover, the amount of bene�ciary groupswas enhanced compared to the previous experiment.

4.2.1 Experimental Setting

The major di�erence to the original project was the new way the bene�t of acontribution was displayed. The new design displayed up to three smilies nextto each movie indicating the value of rating that speci�c movie. For instance,the rating of a rarely rated movie was more valuable than the rating of the latestblock-buster. Rashid et al. assessed the bene�t of a contribution by the value(smilies) assigned to a movie. Moreover, all participants were assigned to groupsof a similar taste and the value of a movie varied depending on the bene�ciary.For instance, the value of a rarely rated comedy movie is more valuable to amember of the 'comedy' group than to a member of the 'crime' group. Insteadof merely di�erentiating between bene�t to oneself and to others, Rashid et al.took the rating behaviour between the bene�ciary groups into account.

Four main groups were designed accompanied by a �fth group, the controlgroup. Each group consisted of 32 members and each member was presented alist of 150 movies. The groups were:

Self In this group, the bene�ciaries of the ratings were the membersthemselves. Hence, the smilies indicated how much the rating of agiven movie would bene�t the member himself.

Similar In this group, the bene�ciaries of the ratings were the members ofthe same group, i.e. those members who have a similar taste. Hence,the smilies indicated how much the rating of a given movie wouldbene�t members of a particular genre.

Dissimilar Similar to the Similar Group, expect that the rating person did notlike the genres that the bene�ciaries like.

19

Page 20: Design for Contribution

All The bene�ciaries were all MovieLens members.

Control In that group, the smilies were not related to any bene�ciary group.They simply indicated the likelihood that a subscriber has seen themovie in the past.

4.2.2 Expected Outcome

The authors posited four hypotheses:

1. Hypothesis: Value > No value.In general, users will more likely rate movies with more smilies rather thanthose with fewer smilies, when smilies indicate value to someone. In thecontrol group the amount of smilies should not in�uence the likelihood ofrating a movie.

2. Hypothesis: Self > Others ( All, Similar Group, Dissimilar Group)Smilies should have a greater in�uence on the likelihood of rating,if theyindicate value to the rating member himself. This assuption is based oneconomic utility theories, which state that people are sel�sh and ratherhelp themselves.

3. Hypothesis: Subgroup (Similar Group, Dissimilar Group) > AllPeople feel greater concern towards subgroups, as they rather have tan-gible reference to similar an dissimilar people than to other, anonymouspeople. Therefore, smilies should have greater in�uence on the likelihoodof rating a movie, if the smilies indicate value to a particular subgrouprather than to all community members.

4. Hypothesis: Similar Group > Dissimilar GroupIf the smilies indicate value to people similar to the rating person ratherthan to dissimilar people, smilies should have a greater in�uence on thelikelihood of rating a movie.

4.2.3 Results and Discussion

Table 3 summerizes the results of the data analysis. H1, H3 and H4 werecon�rmed by the altered experiment. Each hypothesis yielded a slightly, butyet signi�cantly, greater likelihood of the assumed values.

Hypothesis Result

H1: Value > No value + (con�rmend)H2: Self > Others - (discon�rmed)H3: Subgroup > All + (con�rmend)H4: Similar Group > Dissimilar Group + (con�rmend)

Table 3: Testing of the four hypotheses (adpoted from [13])

In contrary to the original experiment, Rashid et al. were successful todemonstrate a connection between bene�t and increased participation. Partic-ularily H1 con�rmed that displaying bene�t to the user in a subtle manner may

20

Page 21: Design for Contribution

help to increase the level of contribution and does not depress participation assuggests earlier.1

1H2: According to a survey among the participants, it was rather demotivating to most

users, when the system suggested a 'value to oneself' that did not match the personal assess-

ment of that user.

21

Page 22: Design for Contribution

5 Conclusion

Since the advent of the usenet and groupware application, virtual communitieshave made all their way into the centre of public life. Today, a life without onlinecommunities would be unconceivable for most of the people in our society. Yet,despite the overwhelming success, the issues seem to remain the same. Since theearly forms of online communities, under-contribution poses the main problemto that form of collaboration.

At the same time, researchers started to pay attention to the issues relatedto communities. Wenger's seven principles are a good example for that. Hisexperience of nearly two decades accumulates in seven principles, which yieldsome kind of minimum requirements for a functioning community.

With the increasing popularity of online communities the research in that�eld intensi�ed. Authors like Kollock and Bishop supply frameworks explainingparticipation in communities with theories of social science. Thus, also theresearch about virtual communities has come a long way and has made all itsway into the centre or focus of research attention.

Moreover, designing communities in the regard of optimizing contributionhas become a structured approach. Instead of merely understanding the designof communities as a constant error-and-trial process, researchers use the �ndingyielded from theory in practical experiments in order to validate them. The twoexperiments that were presented in this paper illustrate this kind of systematicapproach. Finally, community designers employ con�rmed insights in order toimprove participation in their community sustainably.

However, despite the results of the intensi�ed research in the area of com-munities, under-contribution is still a major problem and it does not look likeresearchers are close to an encompassing solution. Moreover, those assumptionsthat could be con�rmed in experiments were only tested for a short period oftime. The question whether these results can still be supported after long-termstudies, has yet to be answered.

22

Page 23: Design for Contribution

List of Figures

1 Degrees of Community Participation (Wenger: 57) . . . . . . . . 82 Maslow's Hierarchy of Needs . . . . . . . . . . . . . . . . . . . . 103 The Ecological Cognition Framework ([6]) . . . . . . . . . . . . . 12

List of Tables

1 Mean Number of Movie Ratings (adapted from [3]) . . . . . . . . 182 Results Summary (apapted from [3]) . . . . . . . . . . . . . . . . 183 Testing of the four hypotheses (adpoted from [13]) . . . . . . . . 20

23

Page 24: Design for Contribution

References

[1] Butler, B.: When is a group not a group: An empirical examination of

metaphors for online social structure. In: Social and Decision Sciences,1999, Carnegie Mellon University: Pittsburgh, PA.

[2] Ortega, F.; Gonzalez-Barahona, J. M.; Robles, G.: On The Inequality of

Contributions to Wikipedia. In: Proceedings of the 41st Hawaii Interna-tional Conference on System Sciences, 2008, p. 6.

[3] Beenen, G.; Ling, K.; Wang, X.; Chang, K.; Frankowski, D.; Resnick,P.; Kraut, R.E.: Using Social Psychology to Motivate Contributions to

Online Communities. In: Proceedings of ACM CSCW 2004 Conference onComputer Supported Cooperative Work, 2004.

[4] Marriam-Webster Online Dictionary: De�nition of community.www.marriam-webser.com, accessed: 13.06.2008.

[5] Wenger, E.; McDermott, R.; Snyder, W.M.: Cultivating Communities of

Practice. Harvard Business School Press, 2002, Bosten, MA.

[6] Kollock, P.: The economics of online cooperation: Gifts and public good

in cyberspace. In: Communities in Cyberspace, Kollock, P. and Smith, M.(eds), Routledge Chapman & Hall, London, 1999.

[7] Bishop, J: Increasing participation in online communities: A framework for

human-computer interaction. In: Computers in Human Behavior 23, 2007,p. 1881-1893.

[8] Wikipedia: Wikipedia. http://de.wikipedia.org/wiki/Wikipedia#Rezeption,accessed: 14.06.2008

[9] Maslow, A. H.: Motivation und Persönlichkeit. Rowohlt, 2002.

[10] Smith, M.: Voices from the WELL: The Logic of the Virtual Commons.UCLA Department of Sociology.

[11] MovieLens.org. http://www.movielens.org.

[12] MovieLens.org: The MovieLens Tour. http://movielens.umn.edu/html/tour/index.html,accessed 15.06.2008

[13] Rashid, A. M.; Ling, K.; Rassone, R. D.; Resnick, P.; Kraut, R. E.; Riedl,J.: Motivating Participation by Displaying the Value of Contribution. In:Proceedings of the CHI, 2006, Montreal.

[14] Kim, A.J.: Community Building on the Web: Secret Strategies for Success-

ful Online Communities. Addison Wesley, 2000, London.

24