In "On the Logic of Demonstatives", David Kaplan points out that the sentence "I am here" is always true (he suggests it may be analytically true) but that the proposition it expresses is not necessarily true because it expresses that a certain person is at a certain place at a certain time, which is contingent. In order to deal with this situation, he proposes adding another element to Frege's semantics.

In Frege's semantics a phrase expresses a sense which has an extension (the extension of the sense is called the denotation of the phrase). Kaplan proposes breaking up the sense into a character and a content, so now the path from phrase to denotation goes like this:

phrase -> character -> content -> denotation

The difference between the character and the content is that in the character, demonstratives like "I" and "now" are unresolved and in the content they are resolved.

I find this account puzzling because I don't see why demonstratives are being treated differently from names. Names can be as context-sensitive as demonstratives. When I talk to my brother, Ed, and use the name "Bob", I'm talking about my other brother, Bob, but when I'm talking to my friend, Ray, and use the name "Bob", I'm talking about our mutual friend, Bob. If I want to mention my brother, Bob to Ray, I call him, "my brother, Bob".

Is it just that the meaning of a demonstrative is defined by the language itself so that you can write rules about how they are used such as "the sentence 'I am here now' is always true"?

I think I may have just answered my own question in the process of carefully posing it, but I'll post anyway in case someone else has a comment to make on it.