The Toulmin test: Framing argumentation within belief revision theories

Arguing on the Toulmin Model: New Essays in Argument Analysis and Evaluation

Volumn , Issue , 2006, Pages 359-377

  Paglieri, Fabio ^a   Castelfranchi, Cristiano ^b  

^a UNIVERSITY OF SIENA   (Italy)

^b INSTITUTE OF COGNITIVE SCIENCES AND TECHNOLOGIES   (Italy)

Author keywords

Argumentation; belief revision; computational models; data structures; multi agent systems; Toulmin's layout of argument

Indexed keywords

EID: 60449088562     PISSN: None     EISSN: None     Source Type: Book    
DOI: 10.1007/978-1-4020-4938-5_24     Document Type: Chapter

References (8)

1. Here we intend argumentation as 'persuasion to believe' (possibly as a means for 'persuading to do'), in opposition to both (i) forceful coercion via threats and promises, and (ii) other conversational contexts in which arguments may be employed. This restricted definition of argumentation serves to make more explicit the connection with belief revision. Nonetheless, it should be noticed that belief revision is a pervasive phenomenon in every instance of persuasion, i.e. not only in argumentation, but in coercion as well. Even when an agent is forced to do the bidding of another by means of threats or promises, such incentives are effective only if the agent believes they are enforceable. Putting a gun to my head is not enough to force me to do something: I must also believe that the person who holds the gun is ready to pull the trigger-and I must prefer life under coercion to violent death. Therefore, even coercion requires belief change, although only to a limited extent, and the distinction is more subtle: while threats and promises can be used in 'persuasion to do' (provided they are actually believed), they cannot be used in 'persuasion to believe', i.e. argumentation. This depends on the non negotiability of beliefs (Stalnaker, 1984; Bratman, 1992; Castelfranchi, 1996), i.e. the fact that agents cannot deliberately decide to believe or disbelieve something just because it is convenient to do so. As for other conversational contexts in which argumentation is used not necessarily to foster persuasion, such as inquiry, deliberation, negotiation, etc. (Walton and Krabbe, 1995; Walton, 1998), belief change is nevertheless involved. For instance, inquiry can be characterized as an attempt to gather new information on a certain issue and to test against each other alternative hypotheses, which more often than not will result in a change in the agent's beliefs, and that will certainly produce a change in the underlying data structure (cf. 3); as for deliberation, we can argue about our decisions and our motives only by supporting or challenging the beliefs underlying each of them (for instance, when we argue that a certain action is not worth the effort, we are actually handling a complex argument that involves beliefs about means-end relations, perceived values and expected costs, and alternative lines of action available to the agent). More generally, to the extent that the function of argumentation is to increase or decrease the acceptability of a controversial standpoint (van Eemeren et al., 1996, p. 5), this will usually results in a corresponding belief change.
2. An important rationality postulate in the original AGM approach, and one of the most debated in subsequent works on belief revision, is the so called success postulate: this principle states that new incoming information is always to have priority over previous beliefs-in AGM terminology, the new evidence is always maximally entrenched, hence taken as valid by the receiver. This requirement is not only unrealistic, it also makes the original AGM belief revision very ill-suited to address argumentation, since the main point of every argument is precisely to assess whether or not a given claim should be accepted, and for what reasons. If we simply postulate priority of new information over old beliefs, this crucial dimension is lost. This is also the reason why more recent works on belief revision have successfully improved the AGM framework, by keeping intact the general set-theoretical approach, but relaxing some of its most idealistic assumptions (for an interesting review, see Wassermann, 2000). The success postulate was one of the first axioms to be relaxed, giving rise to the sub-field of non-prioritized belief revision (Hansson, 1999), i.e. AGM models of belief change in which incoming information is not necessarily preferred over older convictions (see also Pagnucco, 1996, for connections with abduction).
3. The so called foundations theories of belief revision exactly aims to provide a precise account of the reasons supporting a given belief, either using Truth Maintenance Systems (Doyle, 1992) or probabilistic models (Boutilier, 1998). Recently, similar proposals have also been advanced in the field of multi-agent systems (Dragoni and Giorgini, 2003). Although a closer comparison between the model presented here and foundations theories of belief revision is a future goal of our research (cf. 6), it is a fact that none of these theories explicitly addresses argumentation so far.
4. Individual variation in MAS is a major concern also for argumentation studies, e.g. as a way of framing a theory of personality in multi-agent platforms (see for instance Kakas and Moraïtis, 2003).
5. This is exactly why we become so annoyed, and yet vaguely uncomfortable, whenever faced by the everlasting question of any curious child: "Yes, but why is it so?". In principle, the child is perfectly right in raising the question-actually, we are not annoyed by any lack of rationality on her behalf, but rather by her being so damnably rational. In contrast, what we want her to understand is just a matter of practical wisdom, not a theoretical law: that it is pointless to spend the rest of your life, or even a small part of it, going on to chain one 'why' to another.
6. Additional results in integrating DBR and argumentation have already been discussed in Paglieri and Castelfranchi (2005a), in particular with reference to the impact of individual variation over argumentation-a point also discussed in Kakas and Moraïtis (2003).
7. The connection between the DBR understanding of focusing strategies and the pragma-dialectic notion of strategic maneuvering (van Eemeren and Houtlosser, 1999) is bound to be deep and insightful. However, detailed analysis of this issue is left to future work (cf. 6), due to space constraints.
8. Verheij (2002) would say that statements which are argued for are issues, while presupposed statements are assumptions: although here we use a different terminology, the notions are basically synonymous.