Wolfgang Schwarz

Trust and Dyadic Conventions

I had another look at Lewis's trust condition on linguistic conventions. It says that the members of a linguistic community generally take utterances of a sentence as evidence that the sentence is true. My opinion up to now has been that insofar as this condition is correct, it is redundant, and insofar as it is not redundant, it is incorrect.

The condition seems mostly redundant because the convention of truthfulness already requires of everyone to impute truthfulness to others. To be truthful means to try to utter sentences only when they are true. So by partaking in the convention of truthfulness in English, I already expect you to utter "it's raining" only when you believe that it's raining. So unless I believe your opinions about the weather are unreliable, I will take your utterance as evidence for rain. No need for an additional convention of trust.

Lewis's trust condition does not include the proviso "unless I believe your opinions to be unreliable", and that is why it seems too strong to me. If you tell me that I was in Berlin last week, that elephants are remote controlled robots or that the world is a construction of our thoughts and concepts, I have no inclination to take these utterances as evidence for their truth. At best (which is bad enough), I take them as evidence that you believe in their truth.

Now take a case where I do believe your opinions to be reliable. Does this together with the convention of truthfulness really entail that I trust you, in the sense that I assign higher conditional credence to you uttering the sentence given that it is true than given that it is false? I think not. What follows is only that I assign high credence to the conditional that if you utter the sentence, it is true. But that is not the same. Maybe I do that because I'm pretty sure that the sentence is true anyway, or that you won't ever utter it.

This is odd. If there is a (commonly known) convention in a community to do A only if B, shouldn't the members of the community regard A as evidence for B?

Consider a convention to wave a red flag whenever a tiger attacks. As tigers are quite rare hereabout, the people in my neighbourhood all follow that regularity: they wave a red flag whenever a tiger attacks, which happens never. I know that, so I do assign relatively high credence to it. Moreover, most people hereabout probably agree that even though tiger attacks are very rare, it is good to have some way or other to indicate if one takes place. So in particular, we prefer everyone to wave a red flag if a tiger attacks given that this is how everyone else indicates tiger attacks. All this might well be implicit common knowledge among us. So on Lewis's terms, there is a convention in my neighbourhood to wave a red flag if a tiger attacks. (There's also a convention to wave a blue flag etc.) That is silly.

If there really were a convention to wave a red flag if a tiger attacks, this should manifest itself in two conditional attitudes: people would intend to wave a red flag given that a tiger attacks, and they would expect the others to wave a red flag given that a tiger attacks.

Maybe the problem lies in Lewis's monadic conception of conventions. Most conventions regulate actions under certain conditions; they are conventions to do A if (or when) B: to say "hi" if you meet a friend; to knock on the table if a university lecture has ended (in Germany); to restore a telephone connection if it got interrupted and you're the original caller; to wave a red flag if a tiger attacks, etc. (Many conventions are conventions to do A only if B: wave a red flag only if a tiger attacks; say "it's raining" only if it's raining; these can presumably be understood as conventions to do not-A if not-B: to not wave a flag if no tiger attacks.)

Lewis's analysis hides this dyadic nature of conventions. For him, a convention is basically a regularity R such that

1. everyone conforms to R;
2. everyone prefers everyone (including themselves) to conform to R given that everyone else conforms;
3. there is an alternative R' to R meeting the previous condition;
4. everyone believes that conditions (1)-(4) obtain.

A regularity here is a property -- like the property of waving whenever a tiger attacks --, instantiated by those who conform. It is a property everyone values to have, and believes the others to have.

But valuing to [wave if a tiger attacks] is not the same as valuing to wave conditional on a tiger attack, I guess. I don't really know anything on these matters, but I suppose if conditional values were values of conditionals (on some reading of the conditional), people wouldn't have bothered with dyadic systems of deontic logic. At any rate, valuing to A conditional on B is certainly not the same as valuing the material "if (now) A then B" or its universal quantification "whenever A then B". And believing that everyone does [wave if a tiger attacks] is not the same thing as believing about everyone that they wave if a tiger attacks.

I'm not sure how to best state a dyadic analysis of conventions. Maybe somehow like this:

There is a convention to do A if B in population P iff, by and large, among the members of P

1. everyone prefers to do A (over ~A) given that B;
2. everyone prefers everyone (including themselves) to do A given that B AND that everyone else does A given that B;
3. there is an alternative A' satisfying the previous condition;
4. everyone believes that (1)-(4) obtain.

(Maybe the second and third condition could be merged. I've weakened Lewis's condition (1) because it seems to me that there would still be a convention to do A if B even if the members of the community overlook many occasions of B and thus fail to do A, as long as they intend to do A whenever B and expect this of each other.)