Can a software tool aid learners to develop critical thinking - skills of reasoning and critiquing? A theoretical model and an experimental computer program based on it are described. The program assists a user to critique a line of argument or to develop her own reasoning in support of a claim.
The goal of this project is Peitho, a software tool which helps a user to evaluate the assertions of others and to formulate arguments in support of her own claims. This tool can be used in many situations including, for example, the development of a written argument in support of a conjecture. This situation is faced in almost every university program by students who write essays or reports.
For someone in this situation, there are programs available which help with spelling, grammar and style. At an earlier stage, there are 'brainstorming' programs which claim to stimulate the generation of ideas. Peitho supports the intermediate step of structuring the argument from the ideas generated. The result is an outline for the writing process.
Peitho is a productivity tool but it is also a learning tool. A novice can use the software to step through the process of structuring and reviewing an argument. With experience, she should grow independent of its guidance though she may continue to employ it as an organizer.
We often hear laments that our students lack critical thinking skills. Sometimes, the concern is that learners are too slow to question knowledge presented to them; sometimes, the problem is an inability to construct a solid argument. To develop a program which supports the learning of critical thinking skills, we must start with a definition of critical thinking.
Classical rhetoric, the art of persuasion, was supported by three strategies: pathos or appeal to emotion, ethos or appeal to authority and logos, the appeal to reason. With the coming of writing, logos began its rise to dominance over the other two. Reason was more easily captured by the written word and only reason could stand up to the extended scrutiny which the permanence of the written word allowed. In a literate world, reasoned arguments were the most persuasive. Indeed, acceptance of the principles of reason became, and remains, a minimal condition for meaningful discourse.
There are many domains of knowledge where discourse takes place within clearly articulated principles of reason. Perhaps, the epitome of reasoned discourse occurs in science, based on the scientific method.
More recently, the claims of complete rationality, even for the scientific enterprise, have been challenged, but these claims are, of course, supported by reasoned arguments. Our basic commitment to reason seems secure. At the same time however, we see the return of classical rhetoric in the post-literate broadcast media. Pathos (Sally Struthers?) and ethos (Candice Bergen?) are back.
In this truncated and personal view of the landscape, critical thinking is taken to mean:
To capture this interpretation of critical thinking in a program, these aspects must be represented in some model. Formal logic is an obvious candidate.* Considerable effort has been dedicated to extending proof by formal logic into informal 'real' domains. To deal with the incompleteness and inconsistency of the real world (which are not tolerated in the mathematics), the somewhat artificial non-monotonic logics have been developed. The Peitho project rejects this line of development and the price for this decision is the lost deductive power of formal logic.
Instead, we have drawn on rhetorical models of reasoning - developed in the pragmatic real world - which provide a more natural basis for software tools of critical thinking.
A survey of the literature of rhetoric produces many models, each describing the art of persuasion from a unique perspective. The closest to a general theory is probably Stephen Toulmin's (e.g., Toulmin, Rieke & Janik, 1979) but it too is incomplete as a basis for a software tool. He has little to say, for example, concerning multi-step chains of reasoning.
To formulate yet another theory of rhetoric would be presumptuous and futile. Instead, we identify the common features of all the models of reasoning and consider those as a primitive basis upon which any of the particular models, including formal logic, could be described as special cases. The common features define a very simple structure:
This model assumes very little. The unit element is an argument which must begin from a set of accepted conditions and use them to support a claim. If one of the conditions is not accepted, then it too must be established by another unit element in which it is the supported claim. Other elements may also be attached to extend the chain of reasoning. See Figure 1. {EMBED PBrush \s \* mergeformat|}
As stated above, this model does not have the power of inference of formal logic; a program based on it cannot do critical argument. This is not the intent. The model is a structure for describing a variety of specific theories of reasoning in common terms. The software based on it is a support tool for a person doing critical thinking.
In figure 1 the unit elements of argument are just syllogisms but these could take the form of any strategy of argumentation.
The software tool, Peitho, uses the model defined as a concrete representation of an argument. The software gives the user the means to manipulate this representation so critical thinking becomes a hands-on activity.
Because it is process-oriented, the tool is appropriate for learning. It makes explicit the activities involved in framing an argument. These include (i) the gathering and creation of statements, (ii) the selection and organization of the statements into arguments in support of a claim, and (iii) the critique of the argument. The same process applies whether the user wants to analyze and evaluate an existing line of argument or to frame and test her own chains of reasoning. In brief, the user proceeds as follows.
Other features will be identified by the program. For example, Peitho can recognize syntactic problems like "begging the question", a circular argument that is self-justifying.
Still other features are content-based. A well-structured argument can still be strong or weak. Here, Peitho focuses the user's attention with questions about, for example, how well the statements fulfill their roles, or whether an exception could be found to challenge a general claim.
The user can move easily among the activities of information gathering, organizing and analyzing. When finished she can produce a linearized version of the argument which could serve as an outline for writing a paper.
The real power of Peitho comes from its construction as a "design-in-use" tool which enables an instructor to create reasoning environments based on particular models of reasoning such as Toulmin or Aristotle, or the fundamental syllogism of formal logic as in Figure 1. An instructor can provide unit element templates appropriate to the conventions of a discipline and the sophistication of the learners. Figure 2 shows an argument template for Toulmin's basic pattern of analysis. {EMBED PBrush \s \* mergeformat|}
In creating the templates, the instructor defines roles and the relationships among them. She has the opportunity to specify requirements on particular roles. For example, a fact which is being used in justifying a claim should be supported somehow outside the argument. In Figure 2, the grounds, the backing and the rebuttal are shown as needing external support. In the software, there is a hypertext link to that supporting argument if it exists.
The instructor can also frame questions which will be posed to the user/learner at appropriate points in the analysis phase. For example, she might want the learner to consider whether the warrant in Figure 2 really applies to the grounds. This would prompt the user to consider the relationship between the statements attached to these two roles.
Experimentation with Peitho so far has shown that it is capable of representing various approaches to thinking about critical thinking: the unit elements can capture strategies of reasoning. Manipulating a set of statements within the system does help discover structure and does show up weaknesses in an argument.
However, the software itself is still in evolution. This prototype demonstrates the representational power of the model but it can be made simpler and more flexible for the learner to use.
An interesting side effect has been observed from the instructor activity of creating templates for unit elements of argument. The differences among authors writing about critical thinking are clarified by representing their works in a common model and the exercise leads to critical thinking about critical thinking. Approaches vary in many dimensions, such as focus on chains of argument or individual strategies, but it is not yet clear if an improved definition of critical thinking will emerge.
The design of Peitho has been influenced by the family of programs called hyperintelligence tools. (Carlson, 1990) These programs employ strategies of artificial intelligence in a hypertext environment. They support activities such as planning, negotiation, group decision making and engineering design. These are creative, problem-solving processes.
In contrast, Toulmin, for one, is clear in limiting the use of models of reasoning to the construction of rational and convincing arguments in support of claims which have already been discovered.
With Peitho, it is possible to bring these two activities together. The experimental manipulation of the statements in a variety of argument structure elements is typical of the hypertext tools, and allows the user the opportunity to find an 'aha!' connection among the statements.
Exploring this connection between creative and critical thinking is one of the next goals of this exploration.
The project described here is being developed in preparation for my dissertation for the Ph.D. degree under the supervision of Dr. Gary Boyd, Educational Technology, Concordia University, Montréal.
1998 NOTE: Ph.D. completed November 1996 - Thanks Gary et al.
* Mathematical reasoning has been modeled using formal logic. These models have enjoyed substantial success as theorem provers.