Anexo 6. Estrategia de imputación de casos perdidos
3.2.3. Docentes
It is tempting, but ultimately unsatisfactory, to limit one’s attention to the conditional probability, since it seems to offer the basis of a theory of meaning and reasoning. Other factors – we might argue - are ‘just pragmatics’, and should be treated elsewhere. This argument is reminiscent of a common – and often
We start with a formal theory which, though elegant, cannot explain certain crucial phenomena. We then declare these phenomena to be pragmatics, often solely on the grounds that they do not fit within the theory, and take no further responsibility for explaining the phenomena. Applying this strategy more generally results in a pragmatics populated by discarded, unconnected phenomena. Specifically for the conditional, the approach is unsatisfactory for three reasons. Firstly, as we will see below, there is no well-developed pragmatic theory of the conditional to explain what is leftover. Secondly, if we are interested in the process of learning a
conditional, then what is learnt in a given context will not, presumably, be limited to purely semantic information. Thirdly, pragmatics is intimately bound up with
reasoning.
While the first two reasons are self-explanatory, the third requires elaboration. The essence is that pragmatics is crucial for discerning the premises and the
operations on those premises. Consider, for example, the following sentence:
(3) Syntactic Structures is Chomsky’s book.
Here, semantics alone does not yield an evaluable proposition: with the semantics alone, we can only tell that there is some relation between Chomsky and the book. A mandatory pragmatic process must specify what that relationship is: the process is mandatory because it is driven by the linguistic structure; there is a slot, here
provided by the genitive ‘s, which needs to be filled pragmatically (Recanati, 2004). Is Chomsky the book’s author, borrower, or owner? Consider, now, sentence (4):
(4) If Syntactic Structures is Chomsky’s book, then he’ll be angry that you lost it.
Pragmatics, again, is crucial. The conditional probabilities – and, hence, any
subsequent reasoning - will presumably differ depending on whether Chomsky is the author, borrower, or owner, not to mention the necessary reference assignments21. Sometimes, semantics delivers an evaluable proposition, but that proposition may not be the intended one. Consider sentence (5):
(5) Peter has had breakfast.
Here, there is no linguistic slot to be filled, but the proposition may optionally be specified further, for instance, by adding ‘today’ (Recanati, 2004). Consider, now, sentence (6):
(6) If Peter has had breakfast, he won’t be hungry this morning.
Again, the conditional probabilities – and, hence, any subsequent reasoning - will presumably differ depending on whether the breakfast in question was eaten today, yesterday, or twenty years ago. These pragmatic concerns will apply, too, to the conditionals themselves – indeed, to any operator. If researchers want a theory of reasoning which can predict people’s behaviour across contexts, then they will need a component which explains how the operators are interpreted and why. It is
important, then, to ask how people change their beliefs when they learn a conditional from assertion. Is the change localized on the conditional probability, or do beliefs change more broadly? If belief change is broader, then there may be a profound effect on reasoning: for instance, by modifying belief in the other premises in an argument.
21 I am not, here, committing to conditionals expressing propositions. I will assume
Perhaps surprisingly, there are limited data, thus far, on the pragmatics of the conditional. Much of these data come from studies on conditionals with valued antecedents and consequents, known as utility conditionals. Such studies suggest that people draw on the utilities of antecedent and consequent and on simple mentalistic assumptions to interpret conditionals, to infer other people’s intentions, and to predict other people’s behaviour (Bonnefon, 2009, 2012; Bonnefon, Girotto, & Legrenzi, 2011; Bonnefon, Haigh, & Stewart, 2013; Bonnefon & Hilton, 2004; Bonnefon & Sloman, 2013; Haigh & Bonnefon, 2015). Utility conditionals will be the topic of a later chapter. It suffices, for now, to observe that utility conditionals are an important subset of conditionals, and that more remains to be done on the pragmatics of the conditional per se.
A more general topic is invited inferences. Geis and Zwicky (1971) suggested that conditionals tend to undergo ‘conditional perfection’, a pragmatic enrichment, and become bi-conditionals practically automatically (Noveck & Bonnefond, 2011). This practical automaticity has since been questioned. For instance, Evans and Over (2004) have observed that the inference ‘If p then q; q; therefore p’ tends to occur more with abstract conditionals, and that the inference ‘If p then q; not p; therefore not q’ tends to occur more with everyday conditionals. There is arguably, then, no strong connection between the two inferences (for discussion, see Bonnefon & Politzer, 2010). Moreover, as Noveck and Bonnefon (2011) observe, the idea of default pragmatic inferences has lost favour as, contrary to prominent theories (e.g. Levinson, 2000), even seemingly routine inferences have been shown to require effortful contextual processing. Conditional perfection arguably arises out of the expectation that, when we hear a conditional, we will then hear information about the antecedent being satisfied (Noveck & Bonnefond, 2011).
When we do not hear such information, we display a surprise response, and attempt to accommodate this information by pragmatic enrichment (Bonnefond et al., 2012; Noveck & Bonnefond, 2011). On this account, once this accommodation has occurred, the classically fallacious inferences can still be inhibited by people with high cognitive abilities, but this process is effortful and time-consuming (Bonnefond et al., 2012; Noveck & Bonnefond, 2011).
While it is worth bearing this pragmatic account in mind, it applies in the context of reasoning: the pragmatic inferences arise out of a conflict between the expectation of information about the antecedent and the actual appearance of a minor premise which does not satisfy this expectation. There is a more fundamental
question to be asked: namely, what is learnt from the assertion of a conditional. The following three chapters will address this question.