Motivating Agents in Unreliable Environments: A Computational Model - Multiagent System Technologies - page 67

by an agent's motivation [5,6]. In what follows, we formalize the notion of motivation
in order to incorporate it into a formal agent architecture. This allows for a more rational
representation of an intelligent agent.
The theory of motivation is concerned with the question of how an agent determines
its desires. Motivation is driven by motives which describe reasons for some specific
behavior and are meant to be as basic as possible, e. g. hunger or love. A classification
of basic motives is provided by Maslow's hierarchy of needs [5] that distinguishes be-
tween five levels of motives. From top to bottom motives in the corresponding levels
are ordered by their importance and motives in higher levels are only active if some
appropriate portion of lower level motives are satisfied . In its bottom level are the most
basic motives, those for physiological needs such as health and food followed by safety
needs (law and order), love and belonging (family, friends, love), followed by esteem
(independence, respect) and finally self-actualization such as individuality. Motives of
different levels have different susceptibilities to deficiency, in general more basic mo-
tives are more prone to deficiency, while top level motives may not be satisfiable. There
are critical claims on the applicability of the hierarchy of needs for the human reasoning
process due to the simplification of a strict hierarchy [9]. Nonetheless, we choose this
conceptual framework to be used within our model as it provides a sufficient abstraction
for the relationships of motives.
Let L with L, L ∈L

means that motives on level L are more basic than motives on level L . In the following
we use

( L , ) be a totally ordered set of motive levels such that L

L M =( {

sa ( pn stands for
physiological needs , sn for safety needs , etc. as listed above), thus using Maslow's hi-
erarchy of needs to represent importance of basic motives. Our framework, however, is
open to other types or quantities of motive levels. We also adapt the notion of deficiency
needs by partitioning the set of motive levels into deficiency and non-deficiency levels.
For

sa , es , lb , sn , pn } , ) with pn sn lb es

L M we define the deficiency levels as d ( L M )= {

. This simplification of
the above mentioned susceptibilities of motives might be generalized to a more granular
or continuos representation. But it will be sufficient to enable the agent to focus on the
deficiency needs in unreliable situations.
For our agents we assume some (finite) set Mot of basic motives. For every motive

pn , sn , lb }

m denote the motive level of m .

The importance of individual motives and motive levels for an agent constitutes its
personality. To measure importance of motives we employ the unit interval as the gen-
eral range for weights 1 . In general, a smaller weight indicates a less important motive
level.

∈ Mot let L( m ) ∈L

Definition 1. A level weight range function wr on L maps a motive level L ∈L onto

a continuous subset of the unit interval, i. e. wr ( L )=[ l L ,u L ] with a lower and upper

bound l L ,u L ∈ [0 , 1] and l L ≤ u L . We abbreviate δ L = u L − l L .

For a motive level L ∈L the value wr ( L )=[ l L ,u L ] indicates that each motive be-

longing to L has at least an importance of l L and at most an importance of u L .

1

Note, that our approach can be generalized to ranges represented by any totally ordered set;
we choose to use the unit interval only for reasons of simplicity of presentation.