La Literatura como Transformación

A classification of group mirrors from related work is shown in Figure 4.1. A parallel coordi- nates visualization (Inselberg and Dimsdale, 1991) was used, with a slight alteration, which was inspired by the parallel sets visualization (Kosara et al., 2006). A parallel coordinates visualization is one of the standard methods in information visualizations to depict data in a multidimensional space. It consists of parallel axes and polylines with vertices on the axes. One line has one vertex on each axis. However, in the classification of group mirrors, one data point (a line in the parallel coordinates visualization) can have several vertices on one axis. Therefore, a variant based on the parallel sets visualization was used. This is a visual- ization for categorical data that depicts absolute frequencies (visualized in form of ribbons) on parallel axes. The ribbons can split up, a feature which I used for the visualization of the design space of group mirrors. This means that each line can split up and have several vertices on one axis.

I will explain this visualization in our use case and will give a concrete example from our data set. The different characteristics of the design space are displayed on the x-axis. The most important systems from related work are color-coded and listed at the top of the visualization. It can be read from top to bottom. To explain the visualization, I will briefly describe two examples. The MEETING MEDIATOR (Kim et al., 2008) is color-coded in red. The first

axis is the characteristic of the space with the two manifestations co-located and remote. The MEETING MEDIATOR is designed for both scenarios, therefore, the red line splits up

and reconnects in the next category, time, as the system is supposed to serve as a real- timesupport. The second example is the SECOND MESSENGER(DiMicco and Hollenbach,

2006), which exists in several versions. For the aspect of time, one version (the TIMELINE)

is aimed at providing feedback as a replay, all others for real-time. Therefore, the line splits up, and the name of the version is written on top of the line.

It has to be noted, that this design space as well as the classification is not definite. For some of the categories, a continuum seems to be a more fitting choice than discrete points. Nevertheless, in order to give an overview of the design space, a simplified concretization was carried out. For each characteristic, a number of concrete manifestations were chosen,

Fan Histogr_am Fan, B ouncing Balls Timeline Space Time Task Type of Information Type of Visualization Level of Aggregtion Anonymity of Receiver Feedback Creation Cardinality Anonymity of Provider Placement and Privacy Feedback Modality Amount of Guidance co-located remote real-time replay information sharing

structured meeting brainstorming open-ended

quantitative qualitative

subtle overt

individual aggregated

system human

1 to many many to 1 many to many

anonymous identifiable

anonymous identifiable

private semi-public public

visual auditory tactile

mirroring metacognitive guidance

Second Messenger Sturm et al. (2005) Conversation Clock

Reflect

Meeting Mediator _{Streng et al. (2009)}

Conversation V

otes

Brandon et al. (201

1)

Schiavo et al. (2014)

debate _{information sharing meeting brainstorming}

quantitative qualitative

subtle overt

individual aggregated

system human

1 to 1 1 to manymany to 1many to many

anonymous identifiable

anonymous identifiable

private semi-public public (table)

visual auditory tactile

mirroring metacognitive guidance

debate

co-located remote

replay real-time

public (wall)

4 Design Space

for instance, co-located and remote for the characteristic of space. For task, the scenarios were chosen that actually were used in at least one of the studies that were carried out with the systems. The data points for cardinality and anonymity of provider are included under the aspect human, as these only matter in cases in which humans provide the feedback. In the following, the categorization undertaken in Figure 4.1 is described in more detail by discussing each characteristic, starting at the top. Explanations of some of the decisions for classifying the systems are outlined. A detailed description with exemplary images of the classified systems can be found in Section 2.2.2.

As described above, in regard to the aspect ofspace, two possibilities exist: co-located and remote. In this thesis, the focus lies on group mirrors for co-located collaboration. There also exist group mirrors for remote collaboration (often also called “awareness systems”). These are not depicted in this classification. However, one system, the MEETINGMEDIATOR(Kim

et al., 2008), is explicitly designed for co-located as well as remote collaboration and it is therefore categorized under both characteristics.

Switching to the next category, time, one can see that most of the group mirrors for co- located collaboration are designed for real-time deployment. Only one system is designed for the use as a replay. As depicted in the parallel coordinated diagram, this system is the TIMELINEversion of the SECONDMESSENGER(DiMicco and Hollenbach, 2006).

For the categorytask, four different manifestations were chosen, namely the ones that were used in the user studies conducted with the systems. These are very structured tasks such as debates or (scripted) argumentation (summarized under “debate” in Figure 4.1), information sharing tasks, meetings or brainstorming sessions. In Figure 4.1, these tasks are ordered from more structured to more open ended.

Thetype of information, quantitative or qualitative, is ambiguous for two systems that can be classified under both categories. The CONVERSATION VOTES (Bergstrom and Kara-

halios, 2007b) system visualizes speaking times, a quantitative measurement, as well as voting, a qualitative aspect. The system of Brandon et al. (2011) combines quantitative feedback in form of connecting lines between speakers with qualitative feedback in form of agreement of group members.

Relying on Schiavo et al. (2014), a continuum from subtler to more overt was chosen for classifying the type of visualization. Only one version of a system from Schiavo et al. (2014) was categorized as overt, since it provided feedback in textual form.

Thelevel of aggregationis individual for most systems, as the individual participants can be distinguished. Only the FANversion of the SECONDMESSENGER(DiMicco and Hollen-

bach, 2006) shows an aggregated visualization that makes a differentiation of group members impossible.

Thefeedback creationcan either be executed by a system or it can be performed by a hu- man. Systems that use a Wizard of Oz approach (see e.g., Kelley (1983)) in their study are classified under the category “system”, as the Wizard of Oz technique presents a tool in a

simulation of what may be possible in the future even without human intervention. Two sys- tems actually involve human interaction in their design, the CONVERSATION VOTESsystem

(Bergstrom and Karahalios, 2007b), in which group members cast votes on the discussion, and the system by Brandon et al. (2011) in which participants can indicate agreement or disagreement.

In the case that humans provide the feedback, thecardinalityof the feedback provider comes into play. The cardinality can either be one to one, one to many, many to one or many to many. In the considered related work, only group mirrors realizing many to many feedback were implemented.

Another aspect is theanonymity of the feedback provider. He or she can either be anony- mous or identifiable to the other group members. From the two systems that allow feedback from group members, one enables anonymous voting (CONVERSATION VOTES). In the

other system (Brandon et al., 2011), others are able to identify who agrees to whom.

In both cases, when the feedback comes from a system or from a human, theanonymity of the feedback receiver can be determined. From the inspected systems, only the BOUNC- ING BALLS version of the SECOND MESSENGER enables the feedback receivers to stay

anonymous while still representing each group member as an individual representation. The FANis also classified under “anonymous”, as in an aggregated visualization it is naturally

impossible to identify individual persons.

The characteristics of placement and privacyare again more of a continuum. The MEET- INGMEDIATOR(Kim et al., 2008) and the tool by Schiavo et al. (2014) use private displays,

however the other group members are aware of the information that is displayed on the pri- vate devices. The system by Sturm et al. (2006) displays feedback in front of the group members on a table, so that others can glance at feedback directed to the other participants. In terms of thefeedback modality, all of the group mirrors give the feedback visually. The last category, amount of guidance, requires interpretation to some degree. Tools that provide an indication about an optimal state were classified as metacognitive, such as the HISTOGRAM version that indicates if group members are “over”- or “underparticipators”.

In case of the MEETING MEDIATOR one could argue that it is more a mirroring than a

metacognitive tool, as the visualization does not provide an explicit notification about the optimal state. However, from another point of view, the optimal state is visually apparent as the chosen visualization implies that the circle should be aligned in the center of the screen. The same applies to the tool of Streng et al. (2009) in which the flourishing of the trees and the weather can be - borrowed from everyday language - “good” or “bad”. The only system that was categorized as a tool providing guidance was the overt version of the system by Schiavo et al. (2014) that provides explicit hints on how to adapt the behavior.

4 Design Space