Empirical Inference of Numerical Information into Causal Strategy Models by Means of Artificial Intelligence - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 296

As it is obvious from the above theorem, the underlying notion of causality follows the logic of logical
empirism which regards a hypothesis as true as long as it cannot be falsified. Therefore it is the task of a
causality proof to rule out non-causal associations according to the above criteria from a given strategy
model consisting of nomothetic cause-and-effect hypotheses. The next section develops an appropriate
approach for the automated proof of causality.

An eMPIr ICAL VALIdATIon APPro ACh for CAuSAL STr ATeg Y Mode LS

As the previous section provides a homogeneous notion of causality it remains to identify an approach
to apply this definition to the causal knowledge of an enterprise in order to distinguish between genu-
ine and spurious cause-and-effect relations. The conceptual basis for the construction of proven causal
models for strategic decision support is therefore represented by the necessary and sufficient properties
of causality as defined above.

Modeling nomologic Cause-and-effect hypotheses

The starting point for this approach is the compilation of causal knowledge in the form of implicit mental
models or corresponding data into an explicit model of nomologic cause-and-effect hypotheses. The
latter are contained in a rudimentary cause-and-effect model which has to be given by strategic decision
makers and represents the first necessary causal property of a priori knowledge.
One possible approach to model causal strategy maps has been proposed by Hillbrand and Karagi-
annis (2002b, p. 53): The meta-model of their modeling framework consists of indicators which can be
of crisp or fuzzy type and two types of associations which connect a pair of indicators:

• Defined influence relations represent causal associations between variables which can be fully

explained by decision makers. Usually the target variable of this type of relation is some kind of
synthetic ratio which has been axiomatically devised by a formula consisting of several input pa-
rameters. A well-established example for influence relations are profitability measures like ROI,
ROCE, etc. As a consequence defined influences can only be modeled between crisp indicators
where the values of the result variable can be fully explained by some algebraic function of input
indicators.

• All remaining potentially causal associations between either crisp or fuzzy variables which do not

fulfill both conditions for a defined influence relation are called undefined. For this type of relation
it is not possible to assume causality a priori as in the preceding case. Rather the decision makers
are forced to rely on nomothetic cause-and-effect hypotheses. However, the empirical knowledge
of past time-series enables managers to scrutinize the causal content of the interjacent potential
cause-and-effect relation.

Consequently it is the focus of this section to provide appropriate methods in order to analyze the
hypothetically defined model base of a strategic DSS with respect to its causal validity. This task of the
proposed approach is to detect so-called α-errors 2 of nomothetic cause-and-effect hypotheses between
variables. Therefore the starting point for the reconstruction of a proven causal model is a rather over-
defined rudimentary model as described above.