Abstract
Decision aiding in transportation and logistics usually concerns a set of alternatives (solutions, options, actions, etc.) evaluated
from multiple points of view considered relevant by a Decision Maker (DM). The aim is to select a subset of best alternatives or
to rank alternatives from the best to the worst. As the points of view, called criteria, are usually in conflict, the only objective
information that stems from the decision problem formulation is a dominance relation in the set of alternatives (alternative a
dominates alternative b if a is at least as good as b on all considered criteria). While dominance relation permits to eliminate
many irrelevant (i.e. dominated) alternatives, it does not compare completely all of them, resulting in a situation where many
alternatives remain incomparable. This situation may be addressed by taking into account preferences of a DM. In this talk, we
will focus on decision aiding methods based on intelligent construction of DM’s preferences. Comparing to traditional decision
aiding methods, intelligent decision aiding does not require from the DM a difficult to elicit preference information, like
exhaustive pairwise comparisons, criteria weights or trade-offs, but it constructs the DM’s preference model from decision
examples. As model building from decision examples is typical for Artificial Intelligence, we call this approach intelligent
decision aiding. In case of choice and ranking, decision examples provided by a DM have the form of pairwise comparison of
selected alternatives. A preference model should be able to reconstruct the provided pairwise comparisons. In general, the model
construction follows logical induction. In case of real function models, this induction translates into ordinal regression. We show
construction techniques for three kinds of preference models: a set of value (utility) functions, a set of outranking relations, and a
set of “if…, then…” monotonic decision rules. An important feature of all these techniques is identification of all instances of the
preference model that are compatible with (i.e. reconstruct) the input preference information – this permits to draw robust
conclusions regarding DM’s preferences when any of these models is applied on the whole set of considered alternatives. Finally,
we show how these construction techniques can be applied to real world problems from the area of transportation, where
alternatives are evaluated on subsets of criteria structured into technical, functional and strategic levels.