Review by: Reichert, Raimond (2005-01-31)

Discovery learning is a highly student-centered and self-directed form of learning. In this seminal and comprehensive literature review, the authors analyze several problems associated with scientific discovery learning and examine how computer simulations can offer instructional support to overcome these problems.

The review starts by examining the problems that learners encounter in scientific discovery learning, and these include the following:

• Generating hypotheses is difficult. For example, learners may not know what a hypothesis should look like. They have difficulty in modifying their hypotheses to afford the data they gather. Or they make inferences based on variables that remain unchanged between two experiments.
• Designing experiments for deciding on the validity of a hypothesis is also a major challenge. For example, learners have a certain tendency to seek information confirming their hypothesis instead of trying to falsify their hypothesis. Or they design experiments which vary too many variables at once so that no conclusion can be drawn.
• Self-regulation of the discovery learning process is a key issue which separates successful learners from unsuccessful learners. Successful discoverers tend to follow a plan going through their experiments, where unsuccessful learners use a more random strategy.

The second part of the review summarizes measures that can be taken in designing computer simulations for scientific discovery learning. These measures can provide scaffolding for the learner and guide him through the discovery process:

• Direct access to the underlying domain knowledge, provided “just-in-time” the moment the learner needs it, seems to have a positive effect on problem solving and on transfer of knowledge.
• Support for hypotheses generation, for example by providing the basic structure of a hypothesis, seems to have positive effects on the performance of learners.
• Support for designing experiments, for example in the form of hints, seems to positively affect the learners’ experimentation abilities, but does not seem to influence the learning outcome.
• Support for regulation the learning process includes measures such as step-by-step model expansion (e. g. expanding the complexity of the model), planning support (e. g. using guiding questions), monitoring support (e. g. show what has already be done in the simulation), and structuring the discovery process (e. g. providing students with a sequence structure such as “set-up, do, reflect”).

The article concludes that, based on the evidence reviewed, two measures are particularly promising: Direct access to domain knowledge, and planning support using questions or assignments. It also notes that what students learn varies from study to study, or rather, the different studies are hard to compare, having tested for many different outcomes. However, it seems that simulations are helpful when the instructional goal is the mastery of the discovery process per se. The article calls for more research into scaffolding for the discovery learning process. Of particular interest in this context is the question of how to prevent cognitive overload and how to design support tools in order to make them unobtrusive.

The article is highly readable and provides a wealth of interesting references to many renowned research projects. It should serve as a guide to those who wish to implement scientific discovery learning, whether with or without the support of computer-based simulations.