Information Technology Reference
In-Depth Information
Testing Model Robustness - Variation of Farmers'
Decision-Making in an Agricultural Land-Use Model
Georg Holtz and Marvin Nebel
Institute of Environmental Systems Research, University of Osnabrück, Germany
{gholtz,mnebel}@uos.de
Abstract.
Two empirically grounded agent-based models of agricultural land-
use change are presented and compared with respect to factors driving ground-
water use. The models are identical in most aspects, but one model uses a
utility-approach to represent farmers' decision-making, while the other uses
satisficing. The model-comparison exercise helps to distinguish substantial
from implementation-dependent conclusions drawn from the model(s), and
confirms the importance of model robustness tests.
Keywords:
model comparison, robustness, land-use, utility, satisficing.
1
Introduction
The treatment of uncertainty is a major challenges for agent-based modelling ([1],
[2]). An important procedure to enhance a model's credibility is testing the robustness
of conclusions drawn from the model. Robustness means that there is no significant
change in the results when supposedly insignificant changes are made in the model
structure or specification of particular relationships in it. In many studies a model's
sensitivity against random number sequences, (some) parameter values and
sometimes also accessory (technical) assumptions ([3]) are tested. But often there is
also a certain level of uncertainty related to a model's core assumptions, such as
decision-making of agents. In case the agents represent humans, a wide range of
theories from psychology, social psychology, economics, sociology and other (sub-)
disciplines is available to describe their decision-making process and related
behaviour ([4], [5]). A major difficulty and source of uncertainty hence relates to the
“correct” choice of rules for agent behaviour ([6]). Often, this choice is made based
on researchers' own background and available data, and involves some degree of
arbitrariness.
More specifically, in the context of the case-study to be presented below, many
approaches exist to represent farmers' decision making. Related to farming models [7]
argue that one reason for the poor use of such models as decision support tools is the
poor understanding by researchers of the actual process of decision-making by farmers.
Economic approaches assuming the profit-maximising farmer for long have played an
important role ([8], [9]). Since the 1990s, a considerable range of complementary
factors has been identified in empirical studies, from socio-demographics and the
