![[Table of Contents]](/icons/up.gif){kind=icon}
Glaser's research partner, Leona Schauble, explains some of the structural differences among the three subject areas and their effects on students discovery efforts: "In Smithtown, students make qualitative inferences about the relations among variables. The structure of the Smithtown microworld is one of covariation--if you change one variable, regular changes may occur in some of the others. Moreover, students have strong expectations about what those relations will be, since they have a lot of experience with buying and selling. Some of these expectations are consonant with the way economic theory describes them, and some are not. The ones that are not are very hard to change because they are so firmly rooted in everyday experience." In Voltaville, on the other hand, the form of reasoning is quantitative rather than qualitative. People have had less direct experience with electricity, and their experimentation is therefore less likely to be driven by strong conceptions--either correct or incorrect. The structure of this microworld is a set of underlying rules in the form of mathematical formulae, and the learner's task is to discover these principles by seeking mathematical relations among the variables that they explore. The third computer laboratory, Refract, has a mixed structure, sharing features of Voltaville's rule-discovery structure and Smithtown's covariation structure.
In order to learn what kinds of experimentation strategies students would bring to bear in these three different domains, researchers asked twelve college undergraduates to work in all three laboratories over a period of several weeks. In each case, Refract was the last microworld the students explored. The purpose of the study was to track students cumulative learning over all three of the laboratories and to determine which experimentation strategies were general across the three domains and which were particular to one or more domains. Tests administered before and after students' sessions with each laboratory provided a measure of learning. Researchers also made close observations of students' experimentation activities in the laboratories, including their generation and interpretation of evidence and the ways in which they organized and recorded the information they gathered.
Each computer laboratory presented students with a different number of variables and parameters and, as mentioned above, different structural relations among them. The "problem space," or the total number of unique experiments that could be run in each was different. Voltaville supported the smallest problem space, Smithtown the largest. Though Smithtown has only one major variable--price--it has eight parameters, including income level, population, interest rates, and even weather. A minimum of 50 experiments is required to discover all of Smithtown's laws. In contrast, it takes a minimum of six experiments to discover the laws that apply in Voltaville and about 20 in Refract. Therefore, the process of discovering relationships and regularities in Smithtown is more complex and difficult than in either of the other two laboratories, not only because of its larger problem space but also because it is very easy to confuse parameters with variables.
In general, Glaser and Schauble found that successful students were sensitive to the different structures and complexities of the laboratories and varied their experimentation activities accordingly. Because it is necessary to generate three price values at two or more levels of a parameter to discover many of the principles in Smithtown, students changed parameters more often in that microworld than in the others. In contrast, they changed variables more often in Smithtown and Refract than in Voltaville, because laws in those microworlds more often take the form of covariations among variables and outcomes.
Given that successful students tuned their experimentation to the structure of the domain they were working on, the obvious question is what this specificity implies for learning general experimentation skills in science. Rather than a pattern of a set of invariably successful experimentation skills applied across all three domains, the findings revealed that students did not tend to apply the same activities and strategies across all three microworlds. Rather, they applied similar strategies to similar tasks, regardless of which laboratory the tasks appeared in. As Schauble notes, even "a group of students who all have identical scientific reasoning skills may vary considerably in how and when they apply them." The researchers therefore looked closely at the experimentation activities and strategies of the participating students. They found that different students were successful as a function of the different skills required in the various discovery situations, but they also found that all students learned more as they progressed through each of the laboratories in turn. This finding indicates that somehow the strategies being practiced in Smithtown and Voltaville were helping students in their work with Refract. Thus, students were to varying degrees "learning how to learn," and the significant question was to identify the ways in which the better learners developed this ability within the laboratories.
What Glaser and Schauble found, in examining the learning of students across all three microworlds, was that those who learned the most about the processes involved in successful scientific experimentation tended to engage in several activities that were different in kind or in quality than those of the less successful learners. In identifying and reporting these activities, the researchers compared the efforts of "Allen," a student who learned efficiently and well, with those of "Joe," whose learning was less successful. Glaser and Schauble analyzed the two students learning in terms of their search for evidence, their persistence, focus, use of disconfirming feedback, ability to make good use of prior knowledge, and ability to apply relevant heuristics and tools of analysis.
Glaser and Schauble contrasted Allen's behavior to Joe's in regard to all six of these criteria. In his searches for evidence concerning the laws of a given microworld, for example, Allen typically arrived at a tentative hypothesis after only one or two experiments. His further searches were therefore "hypothesis driven"--they represented efforts to confirm his clear expectations about the rules and relationships that characterized the subject area. Joe's experiments were "data driven"--he generated quantities of information without distinguishing between data that could suggest a hypothesis and further data that could support or disconfirm it. He did not seem to design experiments in a patterned or systematic way, and he conducted several that were redundant or served no clear purpose. He generated fewer hypotheses than Allen, and seemed to regard those he did generate as conclusions rather than as statements that needed to be tested against further evidence.
Not surprisingly, Allen also understood better than Joe did when to persist in a line of investigation and when to abandon it as unfruitful, and he pursued his investigations systematically, completing the work related to one law before proceeding to the next. Joe tended to jump around from one line of inquiry to another within a topic, and as a result he made discoveries more slowly and in a more disconnected way.
These qualities of persistence also relate to learners' abilities to focus their attention and to take disconfirming evidence into account in their investigations. For example, Allen was not only better able than Joe to identify and persist in fruitful as opposed to extraneous experiments but was also more likely to think about the microworld problems between sessions and to reorient his attention effectively when he resumed his explorations. He also was more sensitive than Joe to indications that he was pursuing an erroneous hypothesis. When he received computer feedback that disconfirmed his ideas, he was more likely than Joe was to wrestle with his error until he corrected it. Joe, on the other hand, often misread or misinterpreted disconfirming feedback and persisted in unproductive experimentation.
Similarly, Joe was less willing than Allen to let go of mistaken prior knowledge or expectations. In Smithtown, for example, he was unable to see the difference between causal and noncausal factors in an observed relationship between price and demand because he regarded his prior expectation of a causal relationship as "proof" rather than as a hypothesis. Allen, however, worked deliberately to resolve discrepancies between his background knowledge and the new information he obtained from running experiments in the laboratories.
Finally, Allen's more generally mindful and attentive approach also made him more likely than Joe to think analytically about problems in the microworlds, to perceive shifts in the patterns formed by interactions between variables and/or parameters, and to apply mathematical or graphical procedures appropriately.
The general skills Glaser and Schauble identify are, like the skills of self-explanation identified by Chi and the text-processing skills described and modeled by Beck and McKeown, largely self-regulatory. In order to learn how to learn, students apparently must become increasingly sensitive to the differences and similarities among learning tasks in different disciplines. They must also adapt their existing skills accordingly, developing an ability to determine which of their skills are appropriate to which tasks, how to apply them to those tasks, and how to develop further skills as they are needed.
The question for instruction is how to help students develop these self-regulating abilities, and at least part of the answer lies in providing rich opportunities to practice with scientific problem solving in situations that students can manipulate directly. The microworlds are only one kind of environment for such practice, but, as Schauble explains, their hands-on features are particularly effective, especially at demonstrating the true nature of scientific exploration. "There is no point in expecting children to read the scientific method section that starts every science textbook and then expect that they will know how to think about the results of experiments and the meaningfulness of evidence and how to interpret data." Schauble adds that most elementary and intermediate science texts do not teach or discuss thinking skills. In fact, she says, "I think they go overboard in the other direction. . . . They present strings of disembodied facts; there's little coherent explanation anywhere. . . . Strings of declarative facts appear to be the norm."
These failings of texts are clearly the basis for many strands of NRCSL work by Perfetti, Beck, McKeown, and Chi. Of Chi's work, in particular, Schauble observes, "It is so complementary to ours that it is almost like the other side of a penny. . . . She zooms in on smaller episodes of reasoning, how people generate self-explanations when they are reading. We zoom out and say, how does experimentation happen. She zooms in and says, let me compare the structure of knowledge in this biology topic and this physics topic. We zoom out and say, when students are studying in this microeconomics laboratory, what strategies do they use that are different from the strategies that they use in this electrical circuits laboratory. Our concerns are very similar, but we go about them a little differently."
-###-
[Intelligent Tutoring System Enriches Microworlds]