Although the algorithm we have described can deal with uncertainties
having any number of possible outcomes, we have so far discussed only
examples with two possible outcomes. In fact, two-outcome
uncertainties suffice to describe the majority of problems that we
have considered. Indeed, technically, any situation could be
described in terms of some number of two-outcome uncertainties.
However, it is not hard to think of situations that might naturally be
represented in terms of a source of uncertainty with more than two
outcomes. For example, suppose the planner were interested in getting
hold of a particular object in a situation in which the object were
known to be in one of three places. In such a case, the * start*
pseudo-operator would naturally be represented as having three
uncertain effects (one for each possible location of the object) all
associated with alternative outcomes of a single source of
uncertainty. Cassandra's plan for acquiring the object would then involve
three contingencies, one for each possible location.

Louise Pryor <louisep@aisb.ed.ac.uk>; Last modified: Mon Mar 18 17:11:48 1996