Language Use Without Trust
The main difference between Lewis's account of language use in Convention and his account in "Languages and Language" (and later works) is that in the latter the convention required for a language L to be used is a convention of truthfulness and trust in L, whereas in the former it was only a convention of truthfulness. I wonder if there are any good reasons for this change.
Suppose in a certain community there exists a convention of truthfulness in L. On Lewis's analysis of conventions this means that within the community,
- Everyone only utters sentences that are true in L;
- Everyone expects the others only to utter sentences true in L;
- This expectation gives everyone a good reason to only utter sentences true in L;
- Given that almost everyone only utters sentences that are true in L, everyone prefers that everyone only utters sentencs true in L;
- The last two conditions are also be satisfied by a different language L';
- All this is common knowledge.
Now consider what is needed to turn this into a convention of truthfulness and trust in L, where to be trusting in L means to believe that the others only utter sentences that are true in L.
As for (1), everyone must believe that the others only utter sentences true in L. This already follows from (2) above (which in turn follows from (1) and (6), incidentally).
As for (2), everyone must expect the others to believe that the others only utter sentences true in L. This follows from (2) and (5) above.
As for (3), the belief that [everone only utters sentences true in L and believes that everyone else does so], must give everyone a good reason to believe that the others only utter sentences true in L. That's trivial.
As for (4), given that almost everyone [only utters sentences true in L and believes that everyone else does so], everyone must prefer everyone to [only utter sentences true in L and believe that everyone else does so]. This doesn't strictly follow from the convention of truthfulness in L. A counterexample would have to be a community where everybody wants everybody to be truthful, but not everybody wants everybody to believe that everybody is truthful.
As for (5), the revised (3) and (4) must also hold for a different language L'. For (3), that's again trivial. For (4), the only way for it to fail given a convention of truthfulness in L and the revised (4), is that while [everyone wants everyone to be truthful and trusting in L given that most are], and [for some L', everyone wants everyone to be truthful in L' given that most are], still for no such L', everyone wants everyone to be trusting in L' given that most are.
As for (6), all of the above must be common knowledge. This follows from the old condition (6) together with the assumption that it is common knowledge that none of the two kinds of situations mentioned as counterexamples in the previous two paragraphs are realized.
To sum up, a convention of truthfulness in L entails a convention of trust in L unless either two rather far-fetched possibilities are realized or it isn't common knowledge that they aren't realized.
Did Lewis add the requirement of trust to rule out those far-fetched possibilities? That sounds incredible. For it is not at all obvious that if they were realized (or if it was unknown that they aren't realized), L would not be used in the community. At least my intuitions are completely silent on that matter.
Lewis mentions two reasons for adding trust (LL, p.170f.), both of them having nothing to do with the far-fetched possibilities: First, it is implausible that language conventions only coordinate the behaviour of speakers, rather than the interaction between speakers and hearers. -- It would indeed be implausible if our language conventions assigned no role at all to the hearer. But that doesn't follow from our language conventions being conventions of truthfulness alone. To the contrary, a convention of truthfulness (by condition 2) requires everyone to believe that what the others say is true. It seems that Lewis has forgotten that a convention of truthfulness is more than just regularity of truthfulness.
The other reason Lewis mentions is that there might be members of a language community who never say anything.
To explain how this other reason is supposed to work, and why it doesn't, I have to point at another difference between Convention and "Languages and Language": In the latter, "truthfulness in L" is used roughly in the way I have used it so far, as demanding only that one must try not to utter something that is false in L. (I have so far omitted the "try not to" complication -- I think that's irrelevant for what I want to say here.) In Convention (somehwere around p.180), however, Lewis also demands that to be truthful in L, you sometimes have to say something in L. The difference is particularly obvious when you compare "Languages and Language" with its Italian predecessor, "Lingue e Lingua", written in 1968 (published in 1973). Here is the definition of truthfulness in "Lingue e Lingua" (p.4f.):
Essere veridico in L significa agire in una certa maniera: emettere all'occasione enunciati di L, ma cercare di non emettere alcun enunciatio di L che non sia vero in L. Questo vuol dire emettere qualche enunciato di L che sia vero in L secondo le proprie credence, e non emettere qualunque atro enunciato di L.
Here is the corresponding definition in "Languages and Language" (p.167):
To be truthful in L is to act in a certain way: to try never to utter any sentences of L that are not true in L. Thus it is to avoid uttering any sentence of L unless one believes it to be true in L.
So Lewis has carefully removed the requirement that one must sometimes say something in order to be truthful.
This seems reasonable, for indeed there could be members of a language community who never speak, perhaps because they have nothing to say. But why was the requirement there in the first place, and why did Lewis only remove it when he added the trust condition, counting that as an advantage of adding the trust condition?
The answer (LL, p.170 and Convention, around p.180) is that apparently he thought that if truthfulness doesn't require saying anything, and truthfulness is all it takes to belong to a language community, then it gets far too easy to belong to a language community: I've never said anything in lots of languages, but I don't belong to all these language communities -- I don't use all these languages.
True, but it seems that, again, Lewis has forgotten clauses (2)-(6) of conventionality. You can't conform to a convention of truthfulness in L simply by never saying anything. You also need to have the right expectations and believes and preferences concerning what everybody else in the community will say and believe under various circumstances. This suffices to rule out that I take part in the convention of truthfulness in all those languages I don't speak. So Lewis could just as well have dropped the requirement that one must not be silent without adding the trust condition.
(PS: All this only applies to indicative sentences used to convey information. Other speech acts require conventions other than truthfulness, and other than truthfulness and trust -- unless, with Lewis, one redefines "truthfulness and trust" so as to cover whatever else is needed in these cases.)
Later in LL, when he discusses Schiffer's objection, Lewis defines trust not merely as believing that the others tend to be truthful. Rather, to be trusting is to assign a sufficiently high conditional credence to a sentence's being true given that it is uttered. Understood in this way, there are probably some more ways a convention of truthfulness could prevail without a convention of trust. Since the difference between the two readings is minute, I don't believe any of these ways are less far-fetched than the three ways I mentioned.