the difficulty in supporting such beliefs -- they are grounded in cultural convention, not logical reasoning. Attributing the contrary belief to the opponent in a question either eliminates the necessity for supporting the core belief (if the opponent denies the attribution), or puts the burden of support on the opponent. A declarative assertion by the Arab that the blockade was a legitimate act of self-defense would be more likely to require support.
There are several claims, one quite general and two fairly specific, that we have tried to establish in this chapter. The general claim is that the processes and representational structures we have presented constitute a plausible fragment of a computational theory of argumentation. That is, the arguing and reasoning structures and rules described here at the very least serve functions that any process model must accomplish.
The first of the more specific claims concerns the interactions between reasoning processes and argument processes. Engaging in an argument requires several distinct kinds of knowledge: knowledge of the domain, knowledge of how to reason, and knowledge of how to argue. The proper coordination of these disparate knowledge sources can have a mutually beneficial effect in reducing undirected processing. This interaction can also explain the distinction between rebuttals that are discovered as side effects during understanding, and those that require more serious use of argument structures and rules.
We have also argued that personal attacks play a logical role in argumentation, not merely a rhetorical one. They can serve to support other propositions or justify courses of action, and as far as our model of argument structure is concerned, they cannot be distinguished from other kinds of propositions.
There are numerous problems that we have not addressed in this chapter. Perhaps the most important of these unaddressed problems concerns how argumentation is used to persuade, particularly in problem-solving arguments. We have not constructed even a partial theory of how beliefs or opinions are changed by argument (see Dennett, 1978). This is a result of our concentration on adversary arguments, in which typically no one is ever convinced.
We thank Mark Burstein, Gregg Collins, Ernie Davis, Natalie Dehn, Michael Dyer, Jerry Samet, and Roger Schank for useful discussions and comments on earlier drafts of this chapter. This work was supported in part by the Defense Advanced Research Projects Agency, monitored by the Office of Naval Research under contract N00014-75-C-1111, and in part by the National Science Foundation under contract IST7918463.