![](bg3.png)
is in formulation. The e-Brainstorming provides a
structuring frame to express, formalize and
reformulate ideas. There is no moderator, the group
has to be self-moderated.
The questionnaire is scripted with an easy tagged
language, writable in a form, in a mobile phone
application for instance, then sent to a web server.
After generation on a server, a return mail is sent to
the author with the URL of the web questionnaire, to
be distributed to the chosen participants. e-
Brainstorming can be applied in three modes:
without free comments, with free comments, with
free comments and capacity for adding new
questions. Among the features: new questions are
highlighted, participants can check a box “Does not
interest me”, they can access statistics and graphics;
the system uses traces to evaluate the collective
intelligence phenomenon. Data can be exported: it is
possible to exploit it in information processing
systems (semantic web, etc.). Cheap, easy and quick
to implement, it can be used with distributed
participants. This is another example demonstrating
that technology can change spatiotemporal
interactions between individuals (Lancieri et al., 2005).
3.2 Delphi, Real-time Delphi
“The primary strength of Delphi is its ability to
explore, coolly and objectively, issues that requires
judgment” (Gordon, 2003). In the 1950s, on demand
of the US Army, RAND (Research ANd
Development - a thinktank) was in charge of
creating a method to make forecasts from the
opinion of experts about the possibility (etc.) of an
enemy attack during the Cold War. RAND achieved
the Delphi method. A sample of experts on the topic
of the study enters an iterative process of
questionnaires, administered by researchers, which
stops when a pre-defined criterion has been reached.
Iterative process: the questionnaire has to be
prepared, the questions refined. In a first
questionnaire, the experts are asked to provide their
judgment. Data collected is analyzed, synthesized.
The next round begins with the sending of a new
questionnaire. The experts have to be more accurate,
some extreme positions have to be reassessed,
justified (Gordon, 2003). There are multiple rounds
until a defined criterion is reached, then the
administrator stops the questionnaires. Gordon says
that at the end, more often than not, group of experts
move toward a consensus. If not, the reasons are
known. The reasons and arguments for the
consensus as well as for disagreements can be highly
valuable and useful: with this material, it becomes
possible for planners to make judgments. Synthesis
is done to reduce the force of oratory of some people,
by this way every participant is equal; the Delphi
method was designed to encourage a true debate.
Okoli and Pawlowski (2004) have compared a
traditional survey approach with the Delphi method
in order to investigate what would be the critical
success factors for e-commerce in Sub Saharan
Africa. In a traditional survey, a questionnaire is
designed, the participants respond, data collected is
analyzed. In the Delphi method, a questionnaire is
designed, submitted to a group of experts, responses
are analyzed, then the questionnaire is re-designed
and re-submitted and this process is repeated. The
sample of the Delphi method is made of some
experts on the topic (literature recommends 10 to 18
experts). In a traditional survey, a large sample is
preferred, for generalization. Individuals responses
are averaged in a traditional survey while in the
Delphi method, “Studies have consistently shown
that for questions requiring expert judgment, the
average of individual responses is inferior to the
averages produced by group decision processes;
research has explicitly shown that the Delphi method
bears this out”. For a traditional survey, reliability is
an important concern, assured by multiple tests,
while in a Delphi study, the experts have to revise
their responses, the importance degree is different.
In a traditional survey, the construct validity is
assured in the design phase and participants only
respond. In the Delphi method, the experts are asked
to validate the design (researchers' interpretation and
categorization of the variables). Participants are
always anonymous in a traditional survey. In the
Delphi method, respondents are anonymous to each
other but not to researchers, then, if a participant
drops-out, researchers are able to discover the reason
by asking directly. Non-response is an error to be
reduced in traditional survey while in the Delphi
method, experts have been selected and solicited to
give their opinion. In a traditional survey, the quality
depends on the question, design, the follow-up, the
respondents... while in the Delphi method, multiple
iterations provide rich data, furthermore there may
be follow-up interviews: interviewers can come with
open-ended questions to learn more.
In 2004, the Defense Advanced Research Project
Agency (DARPA) asked for the development of a
Delphi-based method for improving the speed of the
Delphi method. A company, Articulate Software,
made the “Real-Time Delphi”. This computerized
method is quick and there can be more participants.
They introduced Artificial Intelligence and Natural
Language during the analysis phase of non-
numerical answers. It works “roundless”: every
ICAART 2012 - International Conference on Agents and Artificial Intelligence
522