314
J. Funke
complex problem solving (for a review see Funke, 1988,1991). For instance,
in a computer-administered microworld called LOHHAUSEN, a subject has
to take the role of an omnipotent mayor of a little town (Ddrner11980,1987).
In other work, a subject plays the role of a manager of a little shop
(TAILORSHOP) or of an engineer in a Third World country (MORO; Putz-
Osterloh & Lemme, 1987). In general, the new approach deals with the
exploration and control of complex and dynamic systems by human
individuals.
According to D6mer, subjects in such situations have to cope with the
following task requirements: (1) they must deal with the complexity of the
situation and with the connectivity of the variables involved since typically not
only a few variables have to be handled with (LOHHAUSEN consists of about
2000variables!); (2) they must deal with the intrαnspαrency or opaqueness of
the situation since typically not all information that is needed is available; (3)
they must deal with dynamic developments of variables which - over time -
change their states autonomously and make it necessary to anticipate trends;
(4) in contrast to simple tasks having only a single goal they must deal with
multiple goals some of which may contradict others (e.g., as a manager: pay
high wages due to the trade-union's request and at the same time maximize
the company's profits).
This paper describes an approach in the area of complex problem solving
developed in Bonn during the three years of the DYNAM1S project. It can be
seen as an attempt to establish an experimentally and systematically oriented
line of research on complex problem solving which should overcome some of
the problems of early research (see the critical aspects mentioned by Eyferth,
Schomann & Widowski, 1986, or by Funke, 1984). During its early phase,
research was less coordinated and less rigorous with respect to traditional
criteria of scientific precision. The main intention was the establishment of
a new research paradigm, the method of computer-simulated scenarios. The
Bonn approach represents a second generation of research activity no longer
under pressure to argue for the existence of certain phenomena, but being able
to introduce the first lines of ordinary research in a more settled phase of
scientific development.
The paper starts in part 1 with an outline of the DYNAMIS research
philosophy, including a description of the dynamic task environment and of
the dependent variables which measure quality of system identification and
control. Part 2 deals with three experiments being in line with the presented
Chapter ljf Microworlds Based on Linear Equation Systems
315
research philosophy. Part 3 summarizes the results and also give some ideas
for further activities.
The DYNAMIS Approach to Complex Problem Solving
Early work in the research domain of complex problem solving suffered from
certain weaknesses, some of them being:
♦ In most cases the definition of a subjects’ solution quality turned out to
be highly arbitrary: how should one, e.g., determine the success of a
town mayor? If one adds up the number of employees in the town, the
energy used and the amount of money in the bank: what kind of
measure would that be? What about its reliability and validity?
♦ Influence of previous knowledge on dealing with a microworld was at
the same time assumed to be of high importance, but has never been
controlled for. Even at the end of a simulated period it was absolutely
unclear if and what subjects would have learned during the session.
♦ Each microworld was a world by its own. Only on a very global level
comparisons to other microworlds were possible. Because of their
idiosyncratic structure, phenomena turned out to oscillate greatly:
sometimes certain effects were observed, sometimes not. Also, due to
the missing replications, it was unclear how stable the results, often
found with small samples, would be.
♦ From the beginning, subjects had to control the microworld without
any opportunity to test certain hypotheses about assumed dependencies
in the system. Due to this procedure, acquisition of knowledge could
not be separated from its application. Also, if some subjects run the risk
Ofhypothesis testing it could happen that according to their intervention
and according to its nonlinear structure the microworld was brought in
such a bad shape that they never escaped such an attractor despite good
problem solving attempts.
The line of research done in our Bonn laboratory therefore established the
following principles: (1) It should always be possible to define the quality of
a solution by comparing it with an optimal solution strategy, (2) The situation
should realize the features of complex problems (complexity, connectivity,
intransparency, Eigendynamik [i.e., autonomous changes without
intervention], and multiple goals) as far as possible. Also, different
microworld simations should be comparable with respect to these criteria.