IEEE Technology and Society Magazine - December 2015 - 80

Table 3. Parameters and Importance Percentage
for all attributes
Attributes

Level

Importance

Transparency

low

28.46%

-1.56***

Medium

-0.61***

High

Costs

Range 0 to
60€

56.84%

-0.05***

Reliability

8.78%

-0.41***

audience

CPA

-0.48***

Only you
Publicly
available

5.92%

0.33***
0

No choice

-2.39***

159

likelihood Ratio

867.03***

-2 lOG likelihood

4192.30

also included the actual probability of all choice tasks
included in this experiment based on the obtained
utilities and this formula. This shows that the both control tasks still have a clear preference for one option
over the other. Furthermore, for choice tasks 3, 6, 9,
10, and 11 the actual probability remains close to the
predicted probability. However, for choice tasks 2, 4,
and 12 the actually preferred option is the opposite of
the predicted one, and for choice tasks 1, 5, 7, and 8 a
clear preference is observed, whereas it was predicted
that either option would be preferred equally. In order
to understand these differences, between the predicted
and actual probability, we will need to look at the importance of each attribute.
The importance percentage gives an indication of
how important the attribute was for decision making.
The importance percentage is obtained by dividing
the utility range of the levels of a certain attribute by
the sum of utility ranges for all attributes. This results
in a proportion of how much impact the attribute
could have on the total utility of a given option. In
Table 3 we provide the importance percentages for all
attributes. The attribute audience has an importance

of 5.92%, reliability has 8.78%, transparency has
28.46%, and cost has 56.84%. This shows that the
cost of a Cloud service was an especially important
factor in respondents' decisions. Indeed, if we compare the actual and predicted probabilities in Table 2,
we see that all differences are always in favor of the
cheaper option.

Market Simulation and Willingness to Pay
Having established the individual utilities for all levels and the fact that cost was the most important
attribute for respondents' decision making, we will
now explore respondents' willingness to pay for
transparency tools in more detail through a market
simulation. The practice of converting differences
between attribute levels to a monetary scale is potentially misleading. The value of product enhancements
can be better assessed through competitive market
simulations. In order to do so, we will compare two
cloud services. First, we consider a vanilla cloud service - i.e., a cloud service that provides low transparency, where reliability is provided by the cloud
provider itself, and where transparency information
is available for only you. This vanilla cloud service is
compared with a premium cloud service that provides the optimal transparency tools - i.e., a cloud
service that provides high transparency, where reliability is provided by an independent authority,
and where the transparency information is publicly
available. We also take into account the possibility that respondents will not want to use either of
these two services.
We start the simulation with the assumption that
both the vanilla and premium cloud services are free.
We use the same formula described previously to
obtain the preference distribution, only now with the
utility of the vanilla (U = -1.64) and premium service (U
= 0), and also the utility for no choice (U = -2.39). This
shows us that there is a 77.8% probability that respondents would prefer the premium service, only 15.1%
probability that they would prefer the vanilla option,
and 7.1% probability that respondents would not chose
either one.
In order to obtain a similar market share, (as a result
of an equal utility), the premium version should cost
32.8 euros per year. As a result, both the vanilla and
premium service would have a probability of 40.5% of
being selected by respondents. This would leave 19.1%
probability that respondents would choose neither. This
shows that as the utility of the services decreases, the
probability that respondents choose neither increases.
It can be determined that if the utility of a service falls
beneath -2.39, respondents become more likely to go
for the no choice option.

IEEE Technology and Society Magazine

∕

December 2015

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2015