Dealing with Dynamic Systems:... 27
by encouraging the subjects to explore the system) and a phase of knowledge
application in which given states of the problem space should be reached by
the subjects as quickly as possible.2 In this last phase, performance measures
should precisely indicate the quality of a subjects’ intervention.
2.1 The DYNAMIS Shellfor Scenarios
Trivially, before you can control a complex system, you must learn how it
works. To study experimentally the acquisition as well as the application of
knowledge we confront our subjects with computer-simulated scenarios. As
a universal tool for constructing these scenarios a computer program called
“DYNAMIS” serves as a shell with which the experimenter can implement
in a simple way different types of simulated systems which all have in
common one formal background. This general frame is a linear equation
system (see e.g. Steyer, 1984) which consists of an arbitrary number of
exogenous (= x) and endogenous (= y) variables according to the following
equation:
yt+1 = Ax yt + B× xt, (!)
where yl+l and y* are vectors representing the state of the у-variables at times t+1 and t; xt is
a vector representing the values chosen by the subject for the х-variables; A, S are matrices
containing the weights for the variables.
A set of measures for describing such systems formally has been suggested
(e.g., Hiibner, 1989). An equation system is constructed according to
theoretical considerations about the presumed influence of certain system
attributes on task complexity (e.g., the effect of Eigendynamik, or the in-
fluence of side effects or effects due to different interdependencies). It is not
intended to simulate a domain of reality adequately because that kind of
simulation places too many constraints on the attributes of the system to be
useful for basic research on problem solving. Consequently, most of the
simulated systems used in our research group have been “artificial.” With
respect to a distinction made by Hays and Singer (1989) one can say that
what we want our systems to possess is not physical fidelity but rather
functional fidelity. As an example see the SINUS system shown in Figure 1.
2 This distinction is somewhat artificial: In most cases knowledge acquisition occurs in order
to reach certain goals. The goals in our settings concern the acquisition of the causal
structure of the system under exploration with the goal of achieving control over that
system (which is related to the application phase).