Page 264 -

12 Participatory Approaches 263

participants’ reactions to this object and some behavioural patterns observed that
can provide new information for the modellers. This connection is activated by
the participants working through specific events and focus on the use of the tool.
Control is still on the side of modellers, who frame the interactions. The format of
information is largely formalised from model to stakeholders. It is not formalised
from stakeholders to model.

12.3.2 Knowledge Engineering: Between Artificial Intelligence
and Social Psychology

Knowledge engineering focuses on a specific time of the interaction between
stakeholders and a simulation model in the design stage: the process of translat-
ing tacit knowledge into conceptual or sometimes computational models. Many
knowledge elicitation techniques are useful in transforming written or oral text
into pieces of simulation models. The purpose of these techniques is to separate
the contributions made directly to the model from the design of the model itself.
Knowledge engineering aims to provide interfaces for this gap.
To deal with this interface, techniques have been developed, grounded in artificial
intelligence, (social) psychology and cognitive science. Behavioural patterns in
social simulation models are often borrowed in simplified versions from these fields
(Moss et al. 2000; Pahl-Wostl and Hare 2004). This cross-pollination of disciplines
can be potentially fruitful for model design. As an example, Abel and colleagues
have built upon the concept of a mental model. They assume that individuals have
representations of their world which may be formalised in causal rules. Working
in the Australian bush, they have designed specific individual interview protocols
and analysis frameworks to elicit these mental models (Abel et al. 1998). In this
case, interaction with the model occurs through the interviewer who in this case
was also the modeller. There was no collective interaction. Researchers dealing
with the interviews and the corresponding model design clearly guide the process.
The format of information is speech (in the form of a transcribed text), which is
transformed into a modelling language in this elicitation process.
Building upon Abel’s work, Becu has further minimised the involvement of the
modeller, still using individual interviews. This has led him to collaborate with an
anthropologist and to use ethnographic data as a benchmark. Individual interviews,
with the interviewee in the environment suitable to the purpose of the interview,
led him to identify objects and relations among these objects. These constitute the
initial basis for an exercise, labelled as playable stories: stakeholders, in his case
farmers from Northern Thailand, are asked to choose the key elements to describe
their world from their own viewpoints (with the possibility of adding new elements),
then to draw relations among them and to tell a story with this support (Becu 2006;
Becu et al. 2006). In this case, interaction between stakeholders and the simulation
model is still on an individual basis. The format of conveyed information is finally