|
Article Excerpt
The Convince Me computer environment supports critical thinking by allowing users to create and evaluate computer-based representations of arguments. This study investigates theoretical and design considerations pertinent to using Convince Me as an educational tool to support reasoning about public policy issues. Among computer environments designed to support argumentation, Convince Me is unique in that it computes a measure of an argument's coherence and presents this information to users as feedback. This measure is based on the ECHO computational model, a connectionist implementation of the Theory of Explanatory Coherence. The study seeks to better understand this coherent argumentation measure by comparing it to other measures, including a measure of the stability of one's views and the number of statements in an argument. Ten 17-year-old students and one scientist used Convince Me to create arguments about policies designed to ameliorate global warming; they also participated in pre- and post-intervention surveys and interviews. Positive correlations were found among the coherent argumentation measure, measures of stability, and number of statements in an argument, and these findings raise considerations for designing educational activities with Convince Me. A debriefing interview's results illustrate further considerations, including the role of the user's stance towards the software.
**********
The process of argumentation, including evaluating evidence related to differing assertions, is central to critical thinking (Kuhn, 1991, 1992). This study investigates reasoning about public policy issues using Convince Me, a computer environment that supports argumentation (Diehl, 2001; Diehl, Ranney, & Shank, 2001; Schank, Ranney, & Hoadley, 1999; Schank, Ranney, Hoadley, Diehl, & Neff, 1994; Siegel & Ranney, in press; Weidner, Ranney, & Steinbach, 1998). Convince Me is one of a variety of computer environments that support users in creating and evaluating arguments. Such environments include Belvedere (Suthers, Connelly, Lesgold, Paolucci, Toth, Toth, & Weiner, 2001; Suthers & Weiner, 1995; Suthers, Weiner, Connelly, & Paolucci, 1995), CSILE (Scardamalia & Bereiter, 1991), and Sensemaker (Bell, 1997, 1998; Bell & Linn, 2000). The Convince Me argument construction software is unique in that it computes a value that is derived from comparing a user's argument and belief evaluations to a model of coherent reasoning. These "Model's Fit" values represent a construct central to Convince Me and are taken as a kind of desideratum of good reasoning. They are derived using a connectionist computer model called ECHO, which is a computational implementation of the Theory of Explanatory Coherence developed by Thagard (1989). The present study seeks to contextualize the environment's Model's Fit measure by comparing it to other measures, including measures of the stability of participants' views. For further experimental information, participants were asked in a debriefing interview to discuss how well they thought their Convince Me arguments reflected their thinking. The study uses policies proposed to ameliorate global warming as a context, since global warming raises highly important public policy questions for citizens.
THEORETICAL BACKGROUND
Previous research investigating argumentation in the context of societal issues has included topics including crime, education, and unemployment (Kuhn, 1991, 1992), as well as problem solving in international relations (Voss, Lawrence, & Engel, 1991). Belvedere, a computer environment that supports argumentation, was studied initially as a tool for 12- to 15-year-olds in areas related to science and public policy issues (Suthers, Weiner, Connelly, & Paolucci, 1995) and more recently as a tool to teach college students problem-solving in economics (Cho & Jonassen, 2000). Toulmin's (1958) model of argument, delineating components of arguments including data, warrants, backings, and conclusions, has been influential in much of this work.
Designers of computer environments to support argumentation must make assumptions about what constitutes a good argument and assumptions about how the process of constructing such arguments may best be supported. For example, the design of the initial implementation of Belvedere involved the creation of graphical icons for various argument components inspired by categories of Toulmin's (1958) model of argument. Although it has some similarities to other computer environments to support argumentation, the Convince Me computer environment is unique in its strong ties to a philosophical and theoretical tradition that better arguments and better thinking involve more coherent explanations. Given that multiple theoretical choices are possible when designing a computer environment to support argumentation (e.g., using Toulmin's approach or using explanatory coherence), this study is oriented toward understanding more about the advantages and limitations of Convince Me's design.
Perspectives valuing coherence as emblematic of good reasoning are seen in the literatures of political science, public opinion research, and cognitive science. In political science, an influential paper by Converse (1964) posited that holding a coherent political ideology is a sign of greater sophistication. In Converse's view, an ideology is coherent if a person's position on one issue predicts his or her position on another issue.
Similarly, in the area of public opinion research, Yankelovich used coherence as an indicator of a better opinion. He proposed two formal criteria for the quality of opinion about policy issues: consistency (whether the opinion was consistent with one's other beliefs) and volatility (whether the opinion was firmly held or changes) (Yankelovich, 1991, p. 24). Yankelovich, Skelly, and White (1981) developed a measure designed to gauge, relatively quickly, the latter of these criteria, volatility (or stability). The measure, which Yankelovich colorfully named the "Mushiness Index," consisted of a set of four questions. Three questions related to hypothesized sources of stability: whether a person had a personal stake in an issue, had more information about an issue, or has had discussions with others about the issue. A fourth question simply asked respondents to estimate the likelihood that they will change their minds. Yankelovich et al. (1981) found that the index was relatively good at predicting whether persons would change their positions about a policy issue. The present study uses Yankelovich's Mushiness Index, but renames it the "Stability Index" for greater clarity, since higher values on the index correspond to higher stability.
In the area of cognitive science, techniques employing connectionist models also provide a way to gauge the coherence of a set of propositions. Thagard (1989) developed a Theory of Explanatory Coherence, embodied in ECHO, a computational model of reasoning (Ranney & Thagard, 1988). The Theory of Explanatory Coherence is based on a set of principles with psychological interpretations such as: (a) all other things being equal, a statement is more believable if it is supported by evidence, and less believable if it is contradicted by strong alternatives; (b) a statement is more believable if it is supported by plausible beliefs and less believable if it is supported by implausible beliefs (Ranney & Schank, 1998). ECHO has been used to model scientific controversies in many areas, including the history of science (Thagard, 1989) and physics reasoning (Ranney & Thagard, 1988). In these studies, coherent argumentation according to ECHO was employed to evaluate reasoning. Further, Ranney and Schank (1998) utilize ECHO to draw parallels between scientific and social reasoning.
Convince Me allows users to create and evaluate representations of their own thinking. Among computer environments that support the creation of arguments, Convince Me is unique as it provides feedback based on the ECHO computational model. This feedback serves as a prompt for users to reassess their arguments. Figure 1 shows the interface of the Convince Me program.
The software provides a way for a user to enter a set of statements and categorize them as either hypotheses or evidence (Figure 1, upper left quadrant). The user then creates a set of links among the statements, indicating which statements explain or contradict others. The program displays these relationships graphically (Figure 1, upper right quadrant). In addition, the user enters a set of Believability ratings indicating how much he or she believes each individual statement (using a scale where "1" is low and "9" is high). The program does not understand the meaning of the statements, but when the ECHO model is run, it generates activation values that may be thought...
|
