Page 103 -

6 What Software Engineering Has to Offer to Agent-Based Social Simulation 99

did not engage with business partners. The team consisted of five core members
who would participate regularly in the focus groups. Over the years we have made
the experience that for our purposes smaller focus groups work best. Whenever
we describe a task, in the following, we also briefly mentioned when and how the
required knowledge was gathered.

Defining the Objectives

The first step within the framework is to define objectives in relation to the aim of
the study. In our case this was done through a combination of a literature review and
focus group discussions. After some iteration we came up with the following:
• Our aim is to study normative comparison in an office environment.
• Our objective is to answer the following questions:
– What are the effects of having the community influence the individual?
– What is the extent of impact (significant or not)?
– Can we optimise it using certain interventions?
• Our hypotheses are:
– Peer pressure leads to greener behaviour.
– Peer pressure has a positive effect on energy saving.
With the objectives defined, we then need to think about how we can test these
objectives. For this we need to consider relevant experimental factors and responses.
Experimental factors are the means by which the modelling objectives are to be
achieved. Responses are the measures used to identify whether the objectives have
been achieved and to identify potential reasons for failure to meet the objectives
(Robinson 2004). In other words, experimental factors are simulation inputs that
need to be set initially to test different scenarios related to the objectives while
responses are simulation outputs that provide insight and show to what level the
objectives have been achieved. In our case the hypotheses are very helpful for
defining an initial set of experimental factors and responses:

• Experimental factors
– Initial population composition (categorised by greenness of behaviour)
– Level of peer pressure (“individual apportionment” vs. “group apportion-
ment”)
• Responses
– Actual population composition (capturing changes in greenness of behaviour)
– Energy consumption (of individuals and at average)
The experimental factors and responses defined at this stage are still very broad
and need to be revisited when more information about the model becomes available.