38
Joachim Funke
the effects of different presentation modes (active vs. passive exploration;
numerical vs. graphical display) have also been found to be influential.
Still lacking is a theory of how people build up a mental representation of
a system and how they derive interventions from that model. Such a theory
should also explain under which circumstances failures in a system could be
detected and how, for example, operators can cope with dangerous situations.
Also, the model should incorporate human error which occurs even in cases
where perfect knowledge is available (e.g., with overtrained pilots).
Klahr and Dunbar (1988) recently developed an integrated model for
scientific reasoning that seems to be applicable to our scenarios. They argued,
“that scientific reasoning can be conceptualized as a search through two
problem spaces: an hypothesis space and an experiment space” (1988, p. 7).
These spaces result from the task of a (naive or well-trained) scientist:
“The successful scientist, like the successful explorer, must master two
related skills: knowing where to look and understanding what is seen. The
first skill—experimental design—involves the design of experimental and
observational procedures. The second skill—hypothesis formation—involves
the formation and evaluation of theory.” (Klahr & Dunbar, 1988, p. 2).
This assumption seems applicable to the research topic of this paper
insofar as the experimental situation Klahr and Dunbar were concerned
with—exploration of a hitherto unknown object—is basically identical to the
situation of exploring and controlling an unknown linear equation system.
Furthermore, they conceptualized the process of knowledge acquisition in
terms of hypotheses that are developed either more inductively or more
deductively. With respect to interindividual differences one could assume
that some of the subjects follow more closely the hypothesis-oriented ap-
proach (and do some sort of model testing), whereas others proceed more
data oriented (and try to inductively form a model).
Besides the great similarity in the experimental procedure and in the
theoretical frame of reference, however, there are some differences. The main
difference can be seen in the way subjects’ knowledge is measured. Klahr
and Dunbar primarily used verbal data (see Bainbridge, 1979, as well as
Ericsson & Simon, 1980, for a critical comment on verbal data in this
context), whereas in our procedure different approaches are taken to diagnose
the structural knowledge a subject acquires. This difference is partly due to
our “object” of exploration: Subjects explicitly have to anticipate the next
states of the system, they have to write down their hypotheses about struc-
tural relationships, and they have to control the system as well as possible.
Brehmer (1989) conceptualizes process control in terms of “dynamic
decision tasks”—in contrast to static or sequential decision tasks—with the
following four characteristics: “(a) a series of decisions are required; (b) these
decisions are interdependent; (c) the decision problem changes, both auton-
omously and as a consequence of the decision maker’s action; and (d) the
decisions are made in real time” (Brehmer, 1989, p. 144). With exception of