Information Technology Reference
In-Depth Information
Example 5.
We continue Example 4 and consider the desires “
buy cheap food
”and
“
buy fruits from your own country
”. Usually, in resource-bounded situations the first
desire is preferred to the latter one. If the agent becomes wealthy the situation is secure
enough to give “
buy fruits from your own country
” a higher motivation.
Definition 6.
Let
Λ
be a function
Λ
:
Ω
×B→
Γ
that creates a new motivational
structure
Λ
(
M, B
)
for some motive state
M
and beliefs
B
.
Having adjusted the motivational structure the agent now has to decide which desires to
follow. In general, the agent is best off in selecting its maximally motivated desires as
goals, i. e. desires
D
with
γ
such that
μ
is maximal in
γ
.Thisistrueforsitua-
tions where the agent does not currently pursue any intention or the world has changed
drastically and the agent has to reconsider its course of action completely. Usually, the
agent is currently pursuing some intentions and switching to a new desire—because it is
slightly more motivated as the current ones—does normally not make sense. The agent
might start alternating between desires and switching to a new desire likely has further
ramifications as it can be incompatible with currently pursued intentions which would
need to be dropped. It is crucial to carefully deliberate on this decision as dropping
an intention that has been pursued for some time may imply a considerable waste of
resources. At this point we do not go into more details on the selection of desires but
assume there is some mechanism to either select zero or more desires from the current
motivational structure for pursuit. Figure 1 gives a rough overview on the agent model
we developed in this section. In the figure, solid lines indicate action flow and dashed
lines indicate information flow.
(
D, μ
)
∈
Agent
B
weight
adjustment
belief
revision
motivation
adjustment
desire
selection
deliberation
and planning
γ
M
Fig. 1.
A simple model of a motivated agent
4
Adjusting Motivation
In this section we elaborate on how the motivation adjustment function
Λ
of our agent
model (see Definition 6) can be realized. In particular we present functions to combine
