Information Technology Reference
In-Depth Information
1. Foundations of Action Theories
1.1 Purpose
Before we discuss the purpose of action theories, let us try to provide a
suitable and compact denition of the subject in question:
An action theory consists of a formal language that allows adequate spec-
ications of action domains and scenarios, and it tells us precisely what
conclusions can be drawn from these specications.
Of course this informal denition cannot be appreciated without further clar-
ication of the crucial terms used therein. To begin with, by \action domain"
we mean any aspect of the world worth formalizing in which the execution
of actions plays a central role. This is the case, for instance, if one wants
to model an agent that interacts with its environment, i.e., the part of the
world it is able to aect. Most importantly, an
autonomous
agent needs pre-
cise knowledge as to the eects of its actions in order to act purpose-oriented
and so to achieve pre-determined goals. The latter requires to draw the right
conclusions from this knowledge in view of particular situations, in which the
agent has acquired partial knowledge about the current state of the environ-
ment and has a certain goal in mind.
By \action scenario" we mean exactly these particular situations: Given is
some information as to the current, the past and/or even a future or counter-
factual state of the world. The task then is to appropriately interpret these
observations so that the right conclusions can be drawn. Action theories
provide this. They include a formal entailment operation that determines the
set of conclusions which a scenario within an action domain allows.
Both a general action domain and a particular scenario are specied using
the formal language underlying an action theory. This language determines
the expressiveness of the theory. There exist action theories, for instance,
that support the specication of actions with non-deterministic eects (such
as rolling a dice), others don't. Whether or not a certain action theory is
suitable for a particular application depends on the expressiveness required.
If, for instance, it suces to consider discrete state transitions only, then
there is no need to employ an action theory designed for modeling continuous
change. The two aspects discussed in this topic, however, viz. indirect eects
and qualications of actions, are arguably fundamental issues and need to be
