Review by: Reichert, Raimond (2005-03-15)

The discovery learning environment Smithtown is a well-known example of an intelligent tutoring system. It pursues two goals: First, to enhance students’ scientific inquiry skills, and second, to let students discover some basic principles of microeconomics. In this often-cited article, the authors report on the results of two large-scale evaluations of Smithtown.

Smithtown covers the microeconomic concepts of demand and supply, market equilibrium, surplus and shortage, and some others. To learn those concepts, students can manipulate variables, e. g. decrease the price of coffee or increase the cost of labor, and see what happens to the simulated markets, numerically and in the form of graph plots. Students can formulate hypotheses in a formal, yet semi-natural language form to test their understanding of the underlying market forces, e. g. “as price increases, demand decreases.” The goal is for students to discover the laws of demand and supply by experimenting with the influence of different variables on the simulated markets.

The results of the first experiment showed that students using Smithtown for about 5 hours learned as much about microeconomics as did students receiving 11 hours of classroom instruction. The authors also detail differences that distinguished more and less successful learners, such as more successful learners testing the generalization of the hypotheses in multiple markets, designing more complex experiments, and manipulating variables more systematically. As a curious aside, it is interesting to note that the less successful group of students had taken considerably more science courses since high school than the more successful group.

The second study investigated the relationship between general intelligence and learning outcome. The results showed that the intelligence test measure accounted for less than 1% of the unique variance in predicting the number of concepts learned. In contrast, 38% of the unique variance was attributable to hypothesis-driven behaviors (i. e. the effective way of scientific inquiry: generating hypotheses and then testing them by manipulating single variables). This is interesting as learning scientific inquiry behaviors could be more trainable than general intelligence.

This seminal article is an excellent must-read for those involved with computer-based discovery learning, and has influenced much research since its publication. The results are important with regards to discovery learning in general, and they also show that computer-based discovery environments can be very effective. It is a pity, but there seems to be no current implementation of Smithtown available on the web.

For designers of discovery world environments, the following article, also reviewed here, investigates which design measures affect learning in such environments: de Jong, Tom; van Joolingen, Wouter R. (1998). Scientific Discovery Learning with Computer Simulations of Conceptual Domains. Review of Educational Research, Vol. 68, pp. 179–201.