Information Technology Reference
In-Depth Information
technical progress in the provision of better tools for developing and analysing
ABSS, there are still rather dicult open methodological issues with regards to
the life-cycle development of these models. As a consequence, to date there is no
de facto
standard to evaluate and validate how an ABSS is developed, including:
-
(a) how available qualitative evidence can systematically be used in ABSS;
-
and (b) how the modeller decide the implementation of qualitative evidence.
Thus provided there are enough resources and time, modellers could effectively
experiment with different implementation methods in search of the most suc-
cessful for carrying out the task of correctly representing a social phenomenon
in an ABSS. However this is rather unrealistic, resource- and time-wise, so it is
likely that ad-hoc, particular techniques will be used by the modeller. And that
is probably the one which the researcher has most experience with. Generally, a
different coding technique is tested only -if at all- when the model is replicated.
These aspects are important in the modelling life cycle of ABSS because chal-
lenges recur within the processes of data collection, analysis, and particularly
during the process of deciding the roles of both qualitative and quantitative data
in the simulation experiment [Polhill et al., 2010] [Geller et al., 2010].
The next page contains an explanation of the Evidence-Driven Approach
to Modelling (henceforth EDAM) which has been integrated with the adapted
EDTM that is introduced in the next section. Considering the process illustrated
in Figure 1, in case the ABSS consistently diverge from what has been observed
in reality about the social phenomena, it is likely that something in the model has
either been misrepresented, implemented incorrectly or that parameters were set
unrealistically [Lucas, 2011]. Both within the academic ABSS community and
practitioners beyond it, the common understanding is that models should at
least generate results that are plausible in light of the existing (qualitative and
quantitative) evidence [Moss and Edmonds, 2005] [Edmonds et al., 2013].
Furthemore, it is recommended that modellers test technicalities of the ABSS
well before reaching the 5th step, described in the next page. A full validation
of obtained results is only possible once the model has been scrutinised and
deemed plausible, both by academics and stakeholders. From this milestone on-
wards modellers can attempt to mediate the development of knowledge about
phenomena -not the model itself - via the interpretation of simulation results.
The lifecycle for developing ABSS models is described in the following steps:
1. The target system is the social phenomena itself, from which evidence is col-
lected and analysed. This might require, for instance: administration of ques-
tionnaires, survey socio-economic circumstances or setup of an automated
strategy for collecting social data
3
from the social phenomenon participants.
2. With analysed evidence, modellers proceed to discuss the plausibility of ob-
servations and assumptions with stakeholders. Here there is a potential loop
as researchers and domain experts must reach an understanding of what has
been analysed and whether hypotheses are based on realistic assumptions.
3
Larger datasets have greater chances of providing richer, more useful findings.
