Geoscience Reference
In-Depth Information
9.1.5 a
gent
aS
a
P
roxy
for
a
h
uMan
d
eciSion
M
aker
Agents are animated entities that behave in a particular way, that is, they make decisions about their
properties and the properties of other agents.
Spatial agents
are located in space and can make deci-
sions regarding their location. Agents exist in the world of other animated and non-animated entities
and interact with them (Benenson and Torrens, 2004). Decision making is critical for identifying
agents among all spatial entities. In this sense, land parcels are not agents even though agents such
as planners, landowners or developers will decide upon their changes and use.
The notions of agent and AB modelling have come from computer science. A classic paper by
Franklin and Graesser (1996) specifies an
autonomous agent
as (1) a system situated within and part
of an environment that (2) senses that environment and acts on it, over time, (3) in pursuit of its own
agenda, so as (4) to affect what it senses in the future. Franklin and Graesser (1996) further define
specific aspects of the behaviour of agents as outlined in Table 9.1.
A geographic modeller may remain sceptical with regard to the notions of Franklin and
Graesser (1996) and similar definitions found in the literature (Hayes-Roth, 1995; Maes, 1995a,b;
Russell and Norvig, 1995; Wooldridge and Jennings, 1995; Ferber, 1999). The reason is that none
of the characteristics of an agent as listed in Table 9.1 can be formally established and tested.
Apart from mobility, they are all a matter of agreement. However, such a non-operational view
of agents as derived from computer science does not pose any significant problem when the
real-world geographic phenomena are investigated. The GC view of reality is shaped by the
real-world phenomena moulded by human decision makers, who make decisions regarding their
own properties, their relations with other agents and their locations, now and in the future. That
is why geographic agents possess most of the properties listed in Table 9.1. They are reactive,
autonomous, communicative and goal oriented since we see the importance of these properties
for representing human social phenomena. Human geographic agents can also learn and even
have
personality
, in case we consider these properties important for the socio-geographic system
under investigation.
The ability to behave and make decisions is a critical property of agents. The majority of geo-
graphic features cannot be associated with agents because they only passively change their proper-
ties with respect to external influences. The examples of geographic features that are not agents
are numerous - infrastructure elements such as street segments, foundations of buildings and land
parcels, or the units of the coverage employed for representing continuous phenomena such as the
triangles of a TIN representing a land surface. Following the terminology of Benenson and Torrens
(2004), I refer to these non-agent features as
objects
. Thus, geographic systems consist of animated,
decision-making agents and non-animated, passive objects.
AB modelling has entered GC as an ancestor of cellular automata (CA). Let us follow this
line too.
TABLE 9.1
Basic Properties of Agent Behaviour according to Franklin
and Graesser (1996)
An Agent Is...
Meaning
Reactive
Responds in a timely fashion to changes in the environment
Mobile
Able to transport itself from one location to another
Autonomous
Exercises control over its own actions
Communicative
Communicates with other agents
Goal oriented
Does not simply act in response to the environment
Learning
Changes its behaviour based on its previous experience
Character
Believable
personality
and emotional state
