2.3 Addressing the complexity and dynamics of problem solving

In recent years, psychological research on problem solving has turned to increasingly complex, authentic problems with a broader scope (Sternberg and Frensch, 1991). It is no longer concerned with well-defined "puzzles" (in the extreme case, reasoning tasks as used in psychometric tests of human intelligence) that can be solved by applying suitable operations. Instead, it addresses the thinking of experts in scientific and professional domains (Reimann and Schult, 1996; Zsambok and Klein, 1997), planning and problem solving in reallife contexts (Funke and Fritz, 1995; Jeck, 1997; Lave, 1988), and the understanding and control of complex ecological, economic and technical systems (Dörner, Kreuzig, Reither, and Stäudel, 1983; Frensch and Funke, 1995). Computer simulation has proved to be an important tool for investigating complex problem-solving performance. In interaction with the computer, the problem solver explores the simulated system, generates and tests (more or less systematically) hypotheses about relationships and regularities, acquires knowledge, and may in this way learn to control the system by purposeful intervention. Systems used in computer-based research include realistic simulations of highly interconnected ecological or economic systems (Dörner, Kreuzig, Reither, and Stäudel, 1983), and systematically constructed, discrete, smaller-scale systems ("finite state automata"; Buchner and Funke, 1993) and virtual experimental environments. Adaptations of such systems in school contexts are described by Leutner (1992). From an educational psychology perspective, such procedures can be understood as environments for discovery learning (Boshuizen, van der Vleuten, Schmidt, and Machiels-Bongaerts, 1997; Leutner, 1992). From the perspective of problem-solving research, these instruments provide a new quality of problem tasks, distinguished by high levels of complexity and, in particular, by a dynamic character. These dynamic tasks have three advantages over static paper-and-pencil tasks:

1. The task demands are enhanced by an active search and by the continuous processing of external information and feedback. Although paper-and-pencil problem-solving tasks may also trigger the application, evaluation and — if necessary — the modification of processing strategies, the interaction with the computer makes such a course of action inevitable.
2. Computer simulations offer much more authentic problem situations than a written test.
3. Not only the results, but also the course of the problem-solving process can be recorded and assessed, i.e., the type, frequency, length and sequence of interventions made by the subjects. This provides process-based indicators of problem-solving strategies.

These three advantages demonstrate the benefits of using computers in the assessment of problem-solving performance. There are, however, serious theoretical and methodological problems when it comes to the measurement of strategies. The definition of such measures, their reliability, the extent to which they are comparable across different simulated systems, and the impact of motivational factors are research questions which are not yet adequately answered.