![](bg2.png)
stand-alone versions (Checkbox Survey, 2013).
For scientific surveys, especially surveys involv-
ing sensitive information such as medical details, on-
line services are not always appropriate. This is espe-
cially true when ethics approvals do not allow third-
parties to store the information for a researcher; only
local systems can be used in these cases. Since QuON
is designed primarily for scientific surveys, this sec-
tion will concentrate on survey systems that can be
installed on local infrastructure.
One of the complications with survey software is
that surveys are not always linear. For example, in
many paper surveys there are instructions similar to
“If ’No’, go to Question 7”. Web surveys do not
need to display these instructions; the system can au-
tomatically skip certain questions based on previous
responses. However, in order to support such oper-
ation, it is necessary to specify this branching logic
when the survey is defined. Different systems allow
this information to be entered in different ways.
One way to allow selective display of questions,
supported by (Checkbox Survey, 2013; Digivey Sur-
vey Center, 2012; LimeSurvey, 2013; SurveyMonkey,
2012), is to have a condition for each question, such
that the question is only displayed if the condition is
true. The problem with this approach is that, if the re-
searcher wants to skip a group of questions together,
then the necessary condition must be entered once
for each question. Thus, any change in the condition
logic would need to be reentered in multiple places,
which can be tedious and error-prone.
Another option to allow skipping of questions
is to include, with each question definition, an op-
tional condition that specifies which question to dis-
play next, based on the response given to the cur-
rent question. Such a technique is used by systems
such as (Digivey Survey Center, 2012; SurveyMon-
key, 2012). However, the condition for these systems
can depend only on the value of the current ques-
tion. This makes it impossible, for example, to dis-
play a question that discriminates between respon-
dents, then display some common questions, and then
branch based on the discriminatory question. Other
systems, such as (Checkbox Survey, 2013) do allow
more complex conditions that are based on questions
other than the current one, but the branch logic is still
part of a question, meaning that multiple questions
cannot reference the same conditions. It is possible
to overcome these limitations by displaying any dis-
criminatory question after the common questions, but
this changes the survey’s design. Another option is to
define the common questions or branching logic mul-
tiple times, once for each of the discriminated types,
but any change to the common questions would then
require each copy to be altered.
User management is also important when con-
ducting surveys. Often, it is useful to collect anony-
mous information, where the researcher does not
know the respondent’s identity. Other times, only
authorised users should have access to complete a
particular survey, so a username and password is re-
quired. In other cases, such as when only known re-
spondents should be able to participate but passwords
would place too great a burden on the respondent, it
would be useful to assign an identifier to a respon-
dent without a password. The individual can be iden-
tified for the current, or any future, survey, without be-
ing required to enter authentication information. This
is particularly useful when combined with custom
URLs generated for each respondent that include the
identification information automatically. While cur-
rent systems do support some combination of anony-
mous and authorised access, we have no knowledge
of systems that support identified but unauthenticated
respondents.
Another consideration, for scientific surveys in
particular, is publishing the fact that data has been
collected. While particular studies, based on col-
lected data, can be published in journals or confer-
ences, there is currently no easy way for researchers
to publish information about existing datasets, and
thus it is difficult for others to discover any survey re-
sponses that have already been collected. No known
existing survey software publishes metadata about the
surveys that have been conducted. However, if re-
searchers could publish such metadata to a registry,
then other researchers could dynamically discover the
dataset and may be able to produce further results
without any extra data collection. Any ethics require-
ments could still be enforced, since only information
that the dataset exists, and how to request access to it,
is released with the metadata. However, discovery of
the metadata would allow new researchers to exam-
ine all requirements necessary to gain access to the
dataset, and to determine whether it is better to access
the existing data, or whether new data should be col-
lected.
QuON is a system designed to overcome the above
limitations of existing survey software. It can be in-
stalled on a local system, meaning that no collected
data is held by a third party. It also supports com-
plex branching to allow customised surveys, different
user types to meet researcher requirements, and the
ability to publish survey metadata to allow external
researchers to discover the collected datasets.
WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies
112