Exploring Mutual Engagement in Creative Collaborations

1

Exploring Mutual Engagement in Creative Collaborations.Nick Bryan-Kinns

.IMC Group.Department of Computer Science, Queen Mary,.University of London, Mile End, London. UK.

[email protected]

.Patrick G. T. Healey and Joe Leach. IMC Group

.Department of Computer Science, Queen Mary,.University of London, Mile End, London. UK.

[email protected]

ABSTRACTGroup creativity is a fundamental form of human activity.In this paper we explore what constitutes mutually engaginginteraction between people – interaction in which creativesparks fly and we lose ourselves in the joint action. In thispaper we present the results of an experiment to comparethe effect representation of identity in a user interface, andtask instruction, has on mutual engagement between remoteparticipants. Surprisingly the results indicate that providingno cues to identity increased mutual engagement betweenparticipants. We also discuss the appropriateness ofquantitative, qualitative, and self-report data for identifyingpoints of mutual engagement.

Author KeywordsCollaboration, creativity, mutual engagement, design,evaluation.

ACM Classification KeywordsH.5.3 Group and Organization Interfaces; H.5.2 UserInterfaces: Theory and methods.

INTRODUCTIONWe are fundamentally social, creative creatures, yet ourunderstanding of the technology we develop hasconventionally focused on individual work orientedactivities. Recently there has been a move away from thisposition with the acknowledgement of the importance ofcollaboration cf. [19][5] and enjoyment [1]. In order tomove further along this path we need to examine the formsof interaction evident in creative collaborations and start tounderstand what it would mean to design for this richbehavior. We propose that a key feature of creativecollaborations are the points at which participants aremutually engaged with each other - the points at whichpeople spark together, lose themselves in their joint action,and arrive together at a point of co-action ‘where you are

when you don't know where you are’ [18]. In suchsituations participants are involved in activities which arefocused and directed, yet sociable and enjoyable. The keyquestion we explore is how mutual engagement can besupported when collaborators are not in the same space. Inthese situations we need to consider how to bring peopletogether through technology in situations without the richcues that normally support mutual engagement.

In the rest of this paper we outline what mutual engagementis and how it relates to other views of creativecollaborations. We then discuss designing for mutualengagement and detail a study of the effect a user interfacefeature has on mutual engagement. We discuss the resultsof this experiment and consider the kinds of data we mightcollect to identify points of mutual engagement. Finally weconclude the paper and point to future research directions.

MUTUAL ENGAGEMENTEngagement with technological artifacts can becharacterized as a point at which participants are able tochange and appreciate changes in the form being producedthrough the artifact [7]. We see similar phenomena inaccounts of flow [6] - optimal experiences in which‘attention can be freely invested to achieve a person's goals’which results in a merging of action and awareness andconsequent lack of self awareness and distortion of sense oftime. Moreover, in order to achieve flow, there needs to bea balance between the skill of the person and the challengeof the activity.

In mutual engagement participants are engaged with boththe product at hand and with others in the collaboration,which we could characterize as group flow cf. [15].Similarly, Miell and MacDonald [12] propose that mutuallyengaged states are indicated by the ‘presence of reasoneddialogue, the exploration of the ideas of more than oneperson and the attempt to integrate these’. Their work onanalyzing the interaction between pairs of children involvedin creative musical composition focused on the effectfriendship had on mutual engagement. They focused onexamining the forms of musical interaction and the contentof chat between friends and non-friends, and found thatpairs of friends engaged with each other more in joint musiccomposition tasks. In contrast, Sawyer’s work focused onethnomethodological exploration of larger groups and thegroup flow evident in their interaction. He studied groups in

a variety of settings including theatre and musicalimprovisation. In these group situations it is not only theinteraction within the group that is of interest, but also theirinteraction with the audience and the wider sociologicalcontext that shapes their behavior. Primarily his workfocused on descriptions of the interaction in order toexplicate what it means to experience group flow.

Our focus is on identifying and manipulating key attributesof artifacts which have an effect on participants’ ability tomutually engage with each remotely other throughtechnology. Our position is that mutual engagement isessential to high quality collaboration which constitutesinteraction beyond routine tasks or transactions. Forexample, collaborative design, improvisation,brainstorming, gossip and gaming all involve high levels ofmutual engagement.

Identifying Mutual EngagementPoints of mutual engagement are inherently difficult toidentify and measure as the act of reflecting on mutual-engagement undermines some of the characteristic qualitiesof the experience such as spontaneity. For current purposes,we propose that the key characteristic of mutualengagement is that it involves engagement with both theproducts of an activity and with the others who arecontributing to those products. So, it is a form of creativeactivity where we are: a) involved in self-expression andyet b) we co-create the product with other people.

In order to identify points of mutual engagement incollaboration we propose examining the forms ofinteraction that take place between participants. Thisreduces the reliance on subjective, introspective assessmentof participants’ feeling state as is often used to identifypoints of flow cf. [6]. Furthermore, we follow Monk andWatt’s critique of the appropriateness of task analysis [19]for collaborative work in which they argue for the objectiveanalysis of communication rather than subjectiveintrospective analysis of tasks and roles. Similarly, ourapproach allows us to objectively identify points of mutualengagement without subjective analytic modeling of theactivities per se.

In terms of joint production we propose that the followingfeatures of interaction indicate points of mutualengagement:

• Proximal interaction. In environments with a spatialmetaphor we propose that participants will work closer toeach other when they are mutually engaged.

• Mutual modification. Increased mutual engagement isindicated by increased modification of each others’contributions to the joint production.

• Contribution to the joint production. Increasedcontribution to the joint production, as opposed tocontribution to individual productions or discussion oflogistical issues such as technological problems, (cf.

articulation of the task [9]) indicates increased mutualengagement.

• Attunement to others’ contributions. As participantsengage with each other they will start to repeat, ortransform each others’ contributions. In previous work[3] we suggested identifying points of attunementbetween participants as indicators of mutually engagingcollaboration. These are points at which participantsrespond not only to major changes, or contributions, butalso to smaller moment-to-moment changes ininteraction. This framework has strong parallels to thealignment of words and gestures in everydayconversation which indicates a level of interaction andengagement with each other where joint meaning andunderstanding is being created [17]. However, byfocusing on attunement of action, we do not limitourselves to verbal or gestural interaction. Identifyingthese points of attunement relies of meta-level analysis ofpatterns of behaviour which is outside the scope of thispaper. In this paper we use this idea to define three levelsof mutual engagement indicated by attunement:

• Acknowledgement – participants show that they areaware of the contribution of another – a basic level ofengagement.

• Mirroring – participants mirror, or reflect, others’contributions indicating a medium level of mutualengagement as they are able to (re)produce others’contributions.

• Transformation – participants transform others’contributions, demonstrating an understanding of thecontent and intent of others’ contributions – a high levelof mutual engagement.

DESIGNIjsselsteijn et al. [11] distinguish between the goal-orientedand social/ emotional oriented purpose of communication.In their view CSCW informs design of support for theformer, and research on connectedness informs design forthe latter. We see mutually engaging activities as residing atthe boundary between goal and emotional orientedcommunication. There is very little work on designing forthis form of interaction, yet, as discussed in the previoussection, we do have some understanding of what mutuallyengaging collaborations might be like.

In a previous study [2] of the use of the remote group musictool WebDrum II [4], and from features of humancommunication in musical interaction, we identified threedesign features from the CSCW literature which webelieved would contribute to mutually engagingcollaborations. These are briefly outlined below in order toillustrate the kinds of design features we see as important:

• Mutual awareness of actions. In normal conversation weare aware of who is contributing what by virtue of our co-location. We proposed that awareness mechanisms cf.[10] are important to the emergence of mutual

3

engagement. For example, highlighting new contributionsto the joint product, and representing authorship - whomade contributions. Such cues give participants a senseof self and otherness in their interaction. We use the termmutual awareness to distinguish it from workspaceawareness [10] as we are interested in awarenessmechanisms which focus on creative interaction ratherthan workplace management.

• Shared and consistent representations. In conversationwe share the same aural space, and to some extent thesame visual space. This consistency needs to be retainedin collaborative systems, for instance by ensuring that allviews on the joint product are the same. Similarly,Robertson [14] identified the shared nature ofrepresentations as a key aspect of shared activities in ashared physical space. We would expect thatcollaborators would find it easier to co-locate theircontributions, and mirror others’ contributions, with ashared and consistent representation.

• Mutual modifiability. From studies of tool use it wasclear to us that being able to modify each others’contributions is important for mutual engagement.Mutual modifiability implies an egalitarian approach torole assignment within the tool rather than explicitlyenforcing role mechanisms – in such an approachparticipants co-ordinate their activity in a subtle anddynamic manner [8].

USER STUDYIn order to objectively evaluate the effect user interfacefeatures have on mutual engagement we need to examinethe interaction that takes place in a creative collaboration.To this end we undertook a study of the use of thecollaborative music tool – Daisyphone [2]. Musiccollaboration provides us with a basic and underexploredform of interaction with which to explore the features ofmutual engagement. Indeed, it is the paradigmatic exampleof a form of human interaction in which engagement,innovation, and ensemble co-ordination are key objects ofthe interaction, and which involves ‘performers whosechoices are continuously modifying and transforming eachother’ [16]. Previous studies of the use of Daisyphone[2][3] concentrated on qualitative assessment of patterns ofinteraction between participants. In this study we aimed toexplore the utility of different forms of data collection inassessing levels of mutual engagement. We collectedquantitative measurements of interaction, qualitativeassessment of patterns of interaction, and subjectiveparticipant responses to their experience of thecollaboration. These are then used to compare the effectuser interface features had on mutual engagement, and tounderstand how best to assess mutual engagement.

DaisyphoneDaisyphone is a remote collaborative music environment inwhich up to 10 remote participants can create and edit ashort shared loop of music semi-synchronously – typically

updates take under one second to be shared. This providessupport for a form of remote group music interaction whilstrequiring little network bandwidth. As with other remotegroup improvisation tools such as WebDrum [4],Daisyphone works by remote clients sharing indications ofmusical contributions via a central server through theinternet and providing a shared and consistentrepresentation of musical loops being constructed.

There is no restricted ownership of contributions orinstruments in Daisyphone – people can edit each others’notes and play the same instruments. Indeed, Daisyphonehas no explicit notion of public or private contributions. Itis worth noting that in previous studies [2] we identifiedexamples of players using different parts of the loop as theirown space, and only contributing to the ‘joint’ compositionover time.

The Daisyphone user interface is illustrated in figure 1.Notes are lower in pitch towards the edge of the circle. Asthe grey arm rotates clockwise, the notes underneath areplayed, so each of the spokes represents notes played at thesame time. Hues of notes indicate who contributed them(this can be systematically manipulated), and intensity ofcolor represents the volume of the note (note that colour hasbeen removed from all illustrations in this paper). Differentshapes represent different instruments including piano(circle), and rhythm (diamond). Volume levels andinstruments are selected from the four central spokes.

Experimental DesignIn previous qualitative studies [2][3] identification ofownership of contribution was repeatedly identified as animportant design feature. Similarly, work such as Gutwinand Greenberg [10] stress the importance of identity incollaboration with the ‘who’ category forming a key part of

Figure 1. The Daisyphone user interface

their framework for understanding workspace awareness. Interms of our design features, this forms a core part ofproviding mutual awareness of action – participants need tobe aware of both who is contributing, and what they arecontributing.

We also found that the structure of activities (fromundirected play to use as part of assessed coursework) withDaisyphone changed participants interaction with eachother. We see this as similar to the key concept of‘collaboration readiness’ [13] – the willingness andincentive to collaborate. Therefore, for this study wedecided to focus on the effect identification of contributionhad on mutual engagement as well as the effect ofmotivation for the collaboration. This leads to twohypotheses:

• From our design principles we hypothesized that: H1:mutual engagement would be greater where people haveexplicit cues to attribution of actions.

• Given that mutual engagement is in part a product of themotivation of participants we hypothesized that: H2:mutual engagement would be greater where people wereencouraged to collaborate.

Two independent variables were manipulated:

• A within-subjects factor of Instruction (Collaborationvs. Skill). In the Collaboration condition participantswere instructed to work as collaboratively as possible; inthe Skill condition they were instructed to attempt todemonstrate the most skill in using the interface

• A between-subjects factor of attributed Identity (Cue vs.No cue). In the Cue condition participants’ contributionswere distinguished by hue, whereas in the No cuecondition all contributions were based on the same hue.

In order to objectively identify points of mutualengagement we measured three dependent variables derivedfrom the indicators of points of mutual engagement outlinedat the start of this paper:

• Proximal interaction measure: closeness of participants’contributions to others’ contributions.

• Mutual Modification measure: number of deletions ofparticipants’ own notes, and other participants’ notes.

• Contribution to joint production measure: number ofnotes contributed.

In addition we examined logs of the interaction to identifypoints of attunement to others’ contributions exemplifiedby repetition and transformation of others’ patterns.We restricted communication between participants tomusical contributions. This allowed us to explore howmutual engagement is sustained without recourse toverbalization. Indeed, the structure of social musicalinteraction is similar to that of conversation in that it istypically co-present and multimodal; the contemporaneousnature of music interaction provides a novel aspect ofinteraction to explore. Moreover, given Miell and

MacDonald’s [12] emphasis on the ‘presence of reasoneddialogue’ as an indicator of mutual engagement we wantedto explore what form this musical dialogue might take whenno verbal communication channels were available.Moreover, ongoing public use of Daisyphone on an un-moderated website without text chat facilities for the lastfour years without suggests that mutual engagement waspossible purely through the musical interaction.Logs of interaction were kept for qualitative analysis of thepatterns of collaboration that emerged in the experiment.

ParticipantsAdvertisements were posted at the authors’ college campus,and 24 students were recruited to participate in the study(10 male, 14 female, average age 21). We offered £10 perparticipant as an incentive to take part in the study. Wespecifically did not recruit any Computer Science studentsas they may have taken courses on interface design andcollaboration in the department. Moreover, they would haveextensive experience of using computers which may havebiased the results. All participants were novice users withno previous experience of using Daisyphone. In order tofocus on the engagement between participants rather thaninitial social encounters participants were recruited as pairsof friends. Indeed, Miell and MacDonald [12] posit thatpairs of friends would be more likely to become mutuallyengaged than non-friends, so selecting pairs of friendsincreases our chance of observing the behaviour we areinterested in. All pairs had known each other for at least sixmonths in a social or study capacity, and some hadpreviously worked together on student projects. Whilstusing pairs of participants meant that a participant wouldalways be aware of the origin of contributions (ascontributions were always either their own or the otherparticipant’s), over time the large number of contributionsin a typical Daisyphone session makes it difficult forparticipants to be sure who contributed what if there is noexplicit representation of authorship. All but 1 participantregarded themselves as ‘beginner’ musicians, with 1participant rating themselves as ‘intermediate’.

ProcedureThe study took place in our lab which consists of twointerconnected rooms. One participant sat in each roomwith a tablet PC running Daisyphone. They woreheadphones to hear the sound produced by Daisyphone andcould not directly see or hear the other participant – theonly communication between participants was throughDaisyphone. One facilitator moved between the rooms andhad access to another PC on which they could view theprogress of the participants. A video camera on a tripodrecorded a view of one of the participants.

Participants were briefed that they were trialing theDaisyphone software in order to see how well it works indifferent situations. Additionally they were informed thatthe software records all actions using made using theinterface, and that this data would be used for research

5

purposes. They were then asked to sign consent forms forsubsequent use of software logs, audio and videorecordings, and questionnaire responses. They were alsoasked to complete a demographic questionnaire. All datawas held anonymously

Their tasks were to work together with a partner to jointlyremotely compose two ringtones for a 12 year old boy.They worked for up to 20 minutes on each composition –from previous studies we found that 20 minutes wastypically the maximum time people would spend creatingone short loop (6 secs) in Daisyphone. Participants weretold that they could stop the task at any point and wouldstill be given their cash incentive. Their task motivation wasmanipulated between compositions; they were eitherinstructed to work as collaboratively as possible, or todemonstrate the most skill in using the interface. Order oftask instruction was randomly changed. They were also toldthat their interaction would be judged for collaboration/skill(depending on the instruction) and that the best examplesform the study would win cash prizes.

After each joint remote composition had taken place, apost-task interview composed of open questions and ratingsof the participants’ interaction and their perception of theirexperience was completed by the pairs of users. The ratingsscale was “lots / sometimes / medium / not much / not atall”, and related to the questions in table 1. These questionswere aimed at identifying whether participants understoodthe nature of their collaboration. We also asked participantsto rate each question in relation to their partner e.g. the firstquestion was rephrased as “Did your partner build on your

notes, patterns or sequences by adding to them.” Thesequestions were aimed at identifying whether participantshad an understanding of their partner’s role in theinteraction.

We also asked a range of open-ended questions about howparticipants interacted with their partners including:

• What effect did working with your partner have on eachringtone?

• Did the other person make edits you didn't agree with? [ifso please explain]

• Did you modify your partner's notes, patterns, orsequences (i.e. reposition notes rather than removethem)?

• Were there any problems integrating both your ideas? [ifso please explain]

A further interview was completed at the end of the sessionin which participants compared the two tasks, and fed backin an open ended way on their experience.

ResultsAll participants undertook the tasks for the full 20 minutes.In this section we detail the results for the three dependentvariables, results from the questionnaires, and finallyqualitative observations on the interaction.

Frequency of ContributionThe effects of Instruction and Identity on frequency of newnotes being added was evaluated in a 2*2 Chi2 test. Thisproved reliable: Chi2

(1)=330, p =0.00.

As Figure 2 shows, both Instruction and Identity affectfrequency of contribution. Participant's who had explicitcues to the origin of contributions showed no effect ofinstruction and also produced approximately a third asmany notes as those who did not. Where there were no cuesto Identity, instructions that emphasized collaboration leadto more frequent addition of notes.

Did you build on your partner's notes, patterns orsequences by adding to them?

Did you repeat your partner's notes, patterns orsequences?

Did you repeat in a modified way your partner's notes,patterns or sequences?

Did you place your notes close to your partner's notes,patterns or sequences?

Did you modify your partner's notes, patterns, orsequences (i.e. reposition notes rather than removethem)?

Did you delete your partner's notes, patterns, orsequences, accidentally or intentionally?

Were there times when you were only removingnotes?

Did you start ideas by creating new notes, patternsand sequences?

Did you respond to your partner by taking their notes,patterns, or sequences as a starting point?

Table 1. Questions asked after tasks. Figure 2. Frequency of contributions

Frequency of DeletesAs Figures 2, 3 and 4 show, the overall frequency withwhich participant's deleted their own (self-deletion, figure3) or others' contributions (other-deletion, figure 4)correlates with the overall frequency of contributed notes.This suggests that participants edited a similar proportion ofcontributions regardless of the experimental condition theywere in, or the amount they produced. However, Self andother deletions were not affected in the same way by themanipulations of Instruction and Identity. As Figures 5 and6 illustrate, if we allow for the changes in overall frequencyof contribution, the proportions of self-deletions arerelatively constant across conditions whereas the levels ofother-deletion are more sensitive to the experimentalmanipulations. They are more common where theinstruction was to collaborate (Matched Pairst(1,23)=1.95,p(one-tailed)=0.03) and more common wherethere were no explicit cues to identity (Independent samplest(1,23)=-1.83,p(one-tailed)=0.04). Other deletions were thusmost frequent when participants were instructed tocollaborate and had no explicit cues to identity ofcontributions.

ProximityTo provide a measure of the extent to which participantswere directly collaborating on their compositions wescored, for each note added by a participant, how manynotes the other participant had already placed on the samespoke (i.e., notes that would be played at the same time asthe added note).

The scores for each participant were averaged and enteredin a mixed design analysis of variance with Instruction as awithin-subjects factor and Identity as a between subjectsfactor. This showed a significant main effect of Instruction(F(1,22)=4.99, p=0.04), no effect of Identity (F(1,22)= 3.36,p=0.08) and a significant Instruction * Identity interaction(F(1,22)= 6.57, p=0.01).

If subjects were instructed to collaborate they worked moreclosely than if they were instructed to be skillful. Thiseffect of instruction was enhanced if there were no cues toidentity. Overall, participants worked more on the samesub-components of tunes when the instruction was tocollaborate and this was enhanced where cues to identitywere not present.

Figure 4. Frequency of other deletion

Figure 5. Mean deletions against instructionFigure 3. Frequency of self deletion

Figure 6. Mean deletions against identity

7

Interview resultsOf those who expressed a preference, fifteen participantsrated their second composition as their favorite whilst fiverated their first as favorite. There was no correlationbetween order of instruction and favored composition.

We identified a range of comments about the experience ofusing Daisyphone. However, there were no correlationsbetween the comments and the conditions participants used.Generally the comments related to the confusionexperienced by participants which they explained as a lackof control, and a clash of ideas as outlined below.

Lack of control – participants often commented that theyfelt a loss of control e.g. “yes, my input was great and thenmy partner took away most of it” and “yes, I had my ownmusic, she destroyed”. This frustration was also expressedin terms of a lack of co-ordination e.g. “sometimes it wasnot coordinated at all”, and in response to the questionabout what effect their partner had on the ringtone “it wouldchange the way I had planned in my head I would have toconstantly adapt”.

Clash of ideas– participants complained that they and theirpartner had different visions of what the ringtone would bee.g. “yes, I wanted a melody, my partner wanted barrymanilow”. This was often related to a lack ofcommunication support e.g. “yes, we could notcommunicate which made it difficult to collaborate” and“yes, because I don't know the exactly idea of her”.

However, collaboration was also seen as a positiveexperience – participants commented that although it wasdifficult to collaborate, the overall effect was more thancould have been achieved as an individual e.g. “I got moreideas that I would have had alone. On the other hand, itturned out quite chaotic because we were not working to thesame pattern” and “it is better [in collaboration]. My partnerhas many good ideas I have not think about, and my ideasalso different from his. So put them together sounds better”,and “interesting, more fun, more idea”. This lends supportto the assertion that participants were mutually engagingeven though they had no verbal communication channels.

Analysis of the interactionWe reviewed participants’ interaction using our logplayback tool. This provided us with an overview of theinteraction as it unfolded over time, and allowed us toreplay and review previous interaction in real time.

Our first observation was that participants were much morefocused on the task of creating music than in previousstudies where we found that participants tended to drawgeometric shapes such as lines and curves. This is probablyexplained by the fact that we offered participants financialincentives to take part in the study, and because they wereset a very specific task. It may also have been due to theincreased training time we gave participants. If so, this mayindicate that Daisyphone provides an intuitive way of

collaborating through music whose learning time is in theorder of tens or minutes.

We identified many points of attunement (our fourthindicator of mutual engagement) between participantsillustrated below.

Acknowledgement – participants show that they are awareof the contribution of another. In the case of Daisyphoneparticipants acknowledge each others’ contributions whenthey purposefully complement each others’ contributionse.g. by creating a chord. Figure 7 illustrates an example ofacknowledgement of contributions. In this case one playerhas created a pattern of curves which the other has fittedtheir notes to e.g. the diamond contributions are aligned tothe other participant’s circle contributions. We distinguishthis from mirroring as the notes are adjacent to each otheras opposed to the pattern being mirrored elsewhere in theDaisyphone.

Mirroring – participants mirror, or reflect, others’contributions thus demonstrating that they themselves areable to produce it. In Daisyphone we could identify thiswhen musical patterns are repeated verbatim around theDaisyphone. Figure 8a and 8b illustrate two kinds ofmirroring highlighted in black shapes. In 8a a participanthas mirrored the 3 note motif of the other participant in

Figure 8a. Mirroring | Figure 8b: Mirroring

Figure 7. Acknowledgement

different places around the Daisyphone – first the top set ofthree circles is mirrored below it by three circles of theother participant, then the three squares at the bottom aremirrored to the right by three diamonds of the otherparticipant. In 8b a participant (light grey) has mirrored themusical motif of the other (dark grey).Trans format ion – participants transform others’contributions, indicating a high level of mutualengagement. In Daisyphone this would be indicated byrepetition of musical patterns with some modifications, ordirect manipulation of others’ patterns. Figures 9a and 9billustrate two forms of transformation of others’contributions. In figure 9a the dark grey participant hasedited and extended the light grey’s straight linecontributions. In figure 9b the dark grey participant hascopied and modified the musical pattern of the light greyparticipant around the Daisyphone.Finally, as mentioned previously, the interaction betweenparticipants in this study appeared to be more focused onthe task in hand. Indeed, unlike previous studies, we onlysaw one example of writing using the notes of Daisyphoneeven though it was the only means of communication.Figure 10 illustrates the only example of textualcommunication in the studies which, it is worth noting, isnot related to the task itself.

DISCUSSIONIn this section we discuss the results or our study. First wediscuss the effects of user interface features on mutualengagement. Then we consider the experience ofcollaboration through Daisyphone, discuss theappropriateness of data capture methods, and propose futuredesign directions.

Effects of user interface featuresOverall, the results of the experiment show that participantswere more willing to contribute when there were no explicitcues to the identity of contributions’ authors. This does notsupport our hypothesis H1. They produced almost threetimes as many notes when there were no cues than whenthere were cues to identity. This is interesting in the contextof our previous findings that participants devoteconsiderable effort to establishing their identities incollaboration. In pervious versions of Daisyphone thatallowed graphical annotation we found that participantswould often take the effort to write their names around theedge of the screen [2]. This writing is the same color as theparticipant’s notes and so provides a form of identificationof contributions – participants essentially make it clearwhich color they are. Figure 11 illustrates the name writingwe have seen in previous studies of Daisyphone. Moreover,most communication tools such as on-line text chat systemsprefix every contribution with the name of the contributor,or some visual representation such as an icon.

The results also suggest that mutual engagement isenhanced when people are explicitly encouraged tocollaborate. This supports our hypothesis H2. However, thisinteracts with the availability of cues to the origin ofcontributions. Instructions to collaborate lead to greateroverall levels of activity, including editing of each other'scontributions and direct collaboration on components of thecomposition. These effects were enhanced when theinterface provided no explicit cues to identity. Thishighlights the importance of well structured experimentswhere participants are motivated to collaborate.

Figure 10. Written communication – “I love you”

Figure 9a. Transformation | Figure 9b: Transformation

Figure 11. Name writing in previous use of Daisyphone

9

It seems plausible that the increased inclination tocollaborate when cues to identity are removed may be dueto a reduced sensitivity to ‘ownership’ of particular parts ofthe composition i.e. deletion of someone else's notes is amore socially marked activity when they are labeled asbelonging to the other participant. However, thisexplanation does not account for the previous accounts ofparticipants’ self-motivated assertion of identity in theshared space through writing their own names. The increasein contributions when cues to identity are removed mayalso relate to the kind of activity participants were involvedin. It may be that as cues to identity move the interfacemore towards a work oriented interface which does not fitwith the playful nature of the interaction in the study.

An alternative explanation is that the apparent increase inmutual engagement is essentially just due to greaterconfusion arising from lack of cues to identity. From thispoint of view people delete more of the other participant'ssimply because they are unclear about who has done whatand contribute more overall because they effectively giveup on trying to collaborate. However, this explanationdoesn't really fit the data. The effects of the manipulationsare different for Self and Other deletions. Whereas Selfdeletions remain a more or less fixed proportion of totalactivity in all conditions, Other deletions are selectivelyaffected. This sensitivity to whose contributions are deletedshows that the participants are not simply confused.Perhaps more importantly, participants’ use of space showsa selective response in terms of their propensity to work onthe same elements of their composition. The fact that this isnot affected by identity cues per se but that the instructionsinteract with identity suggests this is strategic and not theresult of confusion. Moreover, reports of confusionoccurred in each condition and did not correlate with anyparticular condition.

Through analysis of the interaction we identified a richerset of examples of attunement than in previous studies.Previously we had primarily seen participants drawinggeometric lines and curves around the Daisyphone withattunement indicated by other participants mirroring andtransforming these shapes. This difference may be due tothe more focused structure of the activity, or may have beena product of the restricted communication channels (so allattunement occurred through the music).

Daisyphone as a collaborative experienceWhilst it was clear that participants experienced a lack ofcontrol and often felt that their ideas clashed with theirpartners’, they still responded positively to the collaborationitself. And, moreover, it is worth noting that no participantsabandoned the tasks before the end of the allotted time eventhough they were explicitly told they could, though thismay be because participants were offered financialincentives to take part and felt that they ought to work forthe whole time allotted.

The expressed preference for the second constructedringtone probably indicates that participants became moreskilled at using Daisyphone (the instruction condition wasrandomized across participants). We take this as anindication of the short learning time associated withDaisyphone which indicates that it has potential to beintuitive for a range of users.

The increase in indicators of mutual engagement (proximalinteraction, mutual modification, and joint contribution)when participants were instructed to collaborate suggeststhat these indicators do capture mutually engaginginteraction as opposed to engagement with the interface ormusical product itself.

Appropriateness of data collectionIt is interesting to note that the interviews provided verylittle substantive qualitative data on participants’experience. Only the rating of favorite ringtone providedany conclusive qualitative data. The lack of any conclusivefeedback from participants illustrates the problems withusing self-reporting evaluation of mutual engagement. Suchapproaches may be of use for understanding what mutualengagement is, and the kinds of situations in which itoccurs (cf. questionnaires used to identify flow activities[6]), but they do not help us in finer grained analysis of theeffect of user interface features on mutual engagement.Retrospective protocols may be appropriate, but extremelytime consuming, for this level of analysis. This reinforcesour position that for systematic understanding of designingfor mutual engagement we need to be able to objectivelycompare interaction between participants rather than relyingon subject self-reporting.

Future ResearchIn terms of future design, we could identify points ofattunement in the interaction and develop interfaces whichadapt to the participants. For instance, in Daisyphone wemight propose that as evidence of transformation increaseswe increase the distinction in color of the participants’contributions. This would reduce the sensitivity toownership of parts of the contribution for novices, and yetas they felt more comfortable with collaboration, theexplicit cues to ownership (and so, to some extent prowess)increases. Of course, it would be the case that such adaptiveinterfaces remove explicit control of representation of selffrom the participants which may be counter productive –participants may start to feel out of control if their colorchanges and becomes more marked as they become moreexperienced. This also has parallels to work on flow [6]where it is important to balance between users’ skills andthe challenges of the activities they are involved in.

In order to further explore our ideas about mutualengagement we are developing a new remote collaborativemusic tool called Daisyfield. The underlying structure ofremotely shared loops will be retained, but our new designis informed by the findings of this study. In particular, we

will provide greater flexibility of positioning ofcontributions so that we can examine people’s placementstrategies. This will be achieved by providing scope formultiple loops (Daisies) in one space which can be movedby participants. Moreover, these Daisies can be joined withother Daisies to create combined tunes e.g. by giving bothloops the same rhythm or musical scale. In this way wehope to provide direct support for mirroring andtransformation of each others’ contributions.

Using such an approach we can examine the effect not onlyof identity on the interaction, but also various positioningstrategies at different levels of granularity (within the loop,between loops, and the merging of loops). As with the studyreported in this paper we plan to systematically manipulatea range of interface features including identity, symmetry ofawareness of action, and mutual modifiability.

Studies of Daisyfield will also be carried out over longerperiods of time so that we can better understand how formsof interaction change over time and how this relates tomutual engagement.

SUMMARYIn this paper we reported on the experimental study of theeffect representation of identity and task instruction had onmeasures of mutual engagement. The result that providingno cues to identity increased mutual engagement wasunexpected. We explained this in terms of a sense of‘ownership’ and social awkwardness with deleting others’contributions.

Our experiment highlighted the need for objective measuresof mutual engagement such as number of contributions, andamount of deletion of others’ contributions. Furthermore,we showed how collaboration could be supported withoutrecourse to verbal communication which we believe isimportant for the design of radical new paradigms forcreative collaboration. Future work will continue to explorethe role of representation of attribution of contribution indifferent situations. We will also explore how to design andevaluate engaging and mutually engaging interfaces.

ACKNOWLEDGMENTSSupported by EPSRC Grant ‘Engaging Collaborations’(GR/S81414/01).

REFERENCES1. Blythe, M. A., Monk. A. F., Overbeeke, K., and Wright,

P. C. Funology From Usability to Enjoyment. KluwerAcademic Publishers, London, UK. 2003.

2. Bryan-Kinns, N. Daisyphone: The Design and Impact ofa Novel Environment for Remote Group MusicImprovisation. In Proc. of DIS 2004, ACM Press(2004), 135-144.

3. Bryan-Kinns, N. Mobile group music improvisation. InProc. of Engagability and Design 2004, Birmingham,UK. (2004).

4. Burk, P. Jammin' on the Web - a new Client/ServerArchitecture for Multi-User Musical Performance.Presented at ICMC 2000. (2000).

5. Carroll, J. M., Rosson, M. B., and Zhou, J. CollectiveEfficacy as a Measure of Community. In Proc. of CHI2005, ACM Press (2005), 1-10.

6. Csikszentmihalyi, M. Flow: The Psychology of OptimalExperience. Harper Collins. 1991.

7. Douglas, Y. and Hargadon, A. The Pleasure Principle:Immersion, Engagement, Flow. In Proc. of Hypertext2000, ACM Press (2000), 153-160.

8. Dourish, P., and Bellotti, V. Awareness andCoordination in Shared Workspaces. In Proceedings ofACM Conference on Computer-Supported CooperativeWork (CSCW'92), 107-114. (1992).

9. Gerson, E. M., and Star, S. L. Analyzing Due Process inthe Workplace. ACM Transactions on OfficeInformation Systems, Vol. 4, No. 3, 257-270. (1986).

10. Gutwin, C., and Greenberg, S. A DescriptiveFramework of Workspace Awareness for Real-TimeGroupware. Computer Supported Cooperative Work,Vol. 11, 411-446. (2002).

11. Ijsselsteijn, W., van Baren, J., Romero, N., andMarkopoulos, P. The Unbearable Lightness of BeingThere: Contrasting Approaches to PresenceEngineering. In Proceedings of SPIE, Vol. 5150. (2003).

12. Miell, D. and MacDonald, R. Children's creativecollaborations: The importance of friendship whenworking together on a musical composition. SocialDevelopment 36 (2000), 348-369.

13. Olson, G. M., and Olson, J. S. Distance Matters. HumanComputer Interaction (in press). (2006).

14. Robertson, T. Cooperative Work and Lived Cognition:A Taxonomy of Embodied Actions. In Proceedings ofECSCW, 205-220. (1997).

15. Sawyer, K. Group creativity: Music, theater,collaboration. Lawrence Erlbaum Associates (LEA),NJ, USA. 2003.

16. Smith, H. and Dean, R. Improvisation, hypermedia, andthe arts since 1945. Harwood Academic Publishers,Amsterdam B. V. 1997.

17. Tabensky, A. Gesture and Speech Rephrasings inConversation. Gesture. Vol. 1, No. 2, 213-235. (2001).

18. Tufnell, M., and Crickmay, C. Body Space Image.Dance Books, London, UK. 1990.

19. Watts, L. A. and Monk, A. F. Reasoning about tasks,activities and technology to support collaboration.Ergonomics 41, 11 (1998), 1583-1606.