Page 168 - Kaleidoscope Academic Conference Proceedings 2021

2021 ITU Kaleidoscope Academic Conference

Management may use a visualization method to put network 5. DECISION-MAKING IN A MIXED-REALITY,
data closer to decision-making and facilitate the NETWORK DATA VISUALIZATION
interpretation of analytics applied to network data. ENVIRONMENT

The context of the problem presented here refers to the We argue that displaying the data on an immersive graphical
assumption that a particular business organization has access network representation enriches not only the interaction but
to a network data set, which may be its own or a third party’s. the quality of the analysis, with potential for better decisions.
Let us consider a specific scenario where such a business
organization finds it of outmost importance to its commercial Experiments were aimed at both individuals and teams. In
activities to identify the most important members of the this paper we report on experiments performed with
network. individuals exclusively. Experiments consisted of explaining
the purpose of the experiment to the invited lab subject, in
We approach the question of who is most important in a the first place, and then continued with a demonstration of
network by exclusively using information about network the devices and platforms over which the experiment would
connectivity, that is, the bilateral relationships between unfold.
nodes or members of the network. We appeal to known
measures of network centrality, more specifically Lab utilization of a conventional 2D flat screen for data
quantitative descriptors associated with a node, such as visualization does not raise any concern regarding any
degree, betweenness, closeness and eigenvector. These are effects on the subject’s wellbeing. On the other hand, getting
defined as follows: the subject accustomed to wearing an HMD in the short time
needed for experimenting turned out to be more challenging.
Degree centrality tells us about how much a node can Also, since negative reactions to wearing such devices
directly influence other nodes. Quantitatively the degree of a cannot be anticipated, our lab protocol contemplated
node in a network is the number of its neighbors, that is, the provisions for affected subjects to stop at any time and take
number of nodes that share a link with it. a rest, or to completely cancel the session.
Closeness centrality gives an estimate of the power of a node The experimental stage of our work aimed to elicit answers
to spread information efficiently through the network. This on three related and increasingly more complex aspects. In
is a measure of indirect influence, unlike the degree. the first place, we wanted to investigate which environment,
Closeness is measured as the average inverse distance of the an immersive XR environment or a 2D flat screen
node to all other nodes. When a node´s closeness is high its visualization facility, allowed a decision maker to respond
distances to all other nodes are typically shortest. faster to the questions. Next, we wanted to know which
environment would enable more effective decision-making.
Betweenness centrality tells us about the influence a node Last, our research interest focused on whether engagement
has over the flow of information in the network. A node’s with one environment led a decision maker to make better
betweenness is measured by the fraction of shortest paths decisions than those made while engaging with the other
between any two nodes of the network that contain said node. environment.
This measurement is akin to estimating the amount of control
a node may exert on the flow of information within the The first issue refers to the time a decision maker took to
network. answer each question, finishing a task or making a selection
while engaging with each visualization tool.
Eigenvector centrality tells us about the influence of
a node in a network as a measure of the influence of its The second issue is supported on the subject’s reaction to the
neighbors as well. A high eigenvector value means that the environment usability and the easiness with which tasks
node is connected to many nodes who themselves have high were performed.
eigenvector values.
Finally, for the third issue observation and analysis of the
On its own each measurement can be used to rank the nodes responses to standard and more demanding questions
of a network. Factoring out the fact that ties may appear, the allowed us to judge their quality and deal with the quality of
rank is unambiguous, and the most important node can be decision-making.
determined. However, considered together, even if only two
of them, the measurements may not determine It needs to be said too that investigating whether, in the
unambiguously what the most important node is. This is context of network data visualization, group decision-
because not all rankings produced return the same order. It is making occurs faster and more effectively as well as better
certainly observed that nodes with a low degree tend to rank decisions are achieved in one environment rather than the
low on the other measurements rankings as well as high other is one of our major research objectives. This endeavor,
degree nodes tend to rank high on the other rankings. But as though, is not part of this paper.
we narrow down our search for highly connected nodes, we
may encounter difficulties in deciding which node to choose
as being the most important.

– 106 –