Page 169 - Kaleidoscope Academic Conference Proceedings 2021
P. 169
Connecting physical and virtual worlds
5.1 Experiment design Every participant was read a story which asked them to
impersonate a mid-manager in charge of advising a
The task at hand was the determination of the most important marketing campaign. From this point on this section will
members of a network, given that network information was refer to any experiment participant as a “decision maker”. A
available in two distinct ways. First, network members, decision maker in the context of our experimental work
represented by nodes, had been classified as either makes decision on her/his own. Group decision-making was
“influencers” or “regular members”; besides, each node also not part yet of the experimental phase reported here.
contained demographic information. Second, network
relationships were represented by links. The network data set
was made up by a node data set and a link data set.
The network data set was used as the object to be displayed
and visualized by the subjects on both, the immersive
platform, Aroaro, as well as the 2D desktop network data
1
visualization tool, Gephi . Both Aroaro and Gephi were fed
the same network data set. Figure 2 shows a picture of the
network visualized by a decision maker using an HMD
running Aroaro client. Figure 3 is a screenshot of a desktop
computer screen that runs Gephi with the same network data
set displayed in Figure 2.
Figure 3 – Sample network visualized in Gephi
Demographic information for each node was available and
would be displayed on a label by simply activating it from As repetition of the questions for a single decision maker in
the user´s menu. The values of centrality measurements, both environments was to be avoided, the question sets for
discussed in Section 4.2, for the node could also be seen on the two platforms were not identical. Indeed, questions in
the label. A subject could then use the information on the one set were of the same kind corresponding to questions on
label to answer the questions. Subjects were informed and the other set, with respective phrasings being different. Three
trained on manipulating the visualization tools to gain a questions were deemed Low Cognitive-Effort (LCE)
different visual perspective when they considered it questions. These questions were cognitively uncomplicated
necessary as, for instance, in their judgement a node´s label and helped us record response times both in Aroaro and
information would have fallen short of being enough for Gephi for each participant and their level of understanding
answering the question.
of the basic postulate of a centrality measurement. An LCE
question did not entail a cognitive effort beyond what was
already required from participants in terms of understanding
how a graph is a graphical representation of the information
about connectivity contained in a network.
The other questions, deemed High Cognitive-Effort (HCE)
questions, were cognitively more complex, referring to
choices over the network nodes (understood as members of
the fictitious social network coded in the network data set).
These questions required participants to engage in
visualization of the corresponding data set to make decisions
based on multiple attributes and measurements of its
component nodes.
Figure 2 – Sample network in Aroaro’s VR space
The decision maker needed to make twelve decisions (split
To study what environment, either a VR-based environment into two groups of six) to support the fictitious marketing
or a 2D desktop-based visualization facility, would allow campaign by using both platforms. To remit any order effect,
subjects to make faster, more effective and/or higher-quality the first group of decision makers used Aroaro first, and the
decisions, a within-subject study was conducted with 15 other group used Gephi first. Decision makers were shown
participants. Participants were randomly divided into two the way each tool interprets and displays a network data set,
groups. The experiments were conducted on individuals. taught the meaning of degree centrality, closeness centrality
and betweenness centrality, and assisted with preliminary
tests to recognize nodes.
1 https://gephi.org/ Gephi claims to be a “leading visualization and
exploration software for all kinds of graphs and networks.”
– 107 –