Page 267 -

266 O. Barreteau et al.

the conceptual model through a series of diagrams and a set of logical sentences.
The final interaction diagram and the attached logical sentences are then translated
by the modeller in computer code. It is in this type of process that a deep interaction
can occur between participants and the model. This interaction conveys information
on the model content, which is attached to the representations and knowledge of
each participant.

12.3.4 From Statistical Modelling

Bayesian belief networks have been developed to include in the computation
of probabilities, their dependence on the occurrence of any event. They can be
useful to represent complex systems and increasingly used in participatory settings
because their graphical nature facilitates discussion (Henriksen et al. 2004). A
group of participants can be asked individually or collectively to generate relations
between events and possibly probabilities as well. Henriksen and his colleagues
propose a method in seven stages which alternates between individual and collective
assessment and revision of an existent Bayesian belief network diagram.
This approach is reported to still present some difficulties in encouraging
strong participant involvement due to the mathematical functions behind the
network structure. However, other researchers and practitioners have improved
their communication and facilitation of the technique with their own Bayesian
belief network processes and are receiving positive stakeholder engagement in
the modelling processes (Ticehurst et al. 2005). In the example of Henriksen
and colleagues, the process is controlled by the modeller and includes only a
rather superficial coupling between participants and the model. The translation of
participant-provided information into probabilities is mediated by the modeller and
is rather opaque, as in many participatory modelling approaches.

12.3.5 From the Social Sciences

The association of participatory approaches and social simulation modelling also
originates from disciplines not focusing on the production of tools but on under-
standing social systems. Social psychology, economics, management and policy
sciences have all developed their own interactive protocols to involve stakeholders
in the design and/or use of their models. Sociology is still at the beginning of
this process (Nancarrow 2005). These protocols propose a variety of structures of
experimental settings, from laboratory to in vivo experiments through interactive
platforms (Callon and Muniesa 2006). These three categories vary according to their
openness to the influence given to participants. The in vivo category is beyond the
scope of this paper since it does not involve modelling: the society in which the
experiment is embedded provides its own model (Callon and Muniesa 2006).