On Jan 17, 2008, at 10:38 PM, John Black wrote:
On Wed. Jan. 16, 2008, at 9:18 PM, Pat Hayes wrote:
...
In other words, my intuition says that in this case: (ist today (and A (not A))), today is not a valid context for that tautology
Small point of logical terminology -- "tautology" is reserved for logical *truths* of propositional logic.
because there is no other context, which if it took the place of "today", would change the value of that tautology. This is pragmatic, to count as a context it must make a difference, and there must be an alternative context that makes a different difference.
So you appear to be saying that logical truths and logical falsehoods can never legitimately be asserted to be true or false in a context because their truth values never change. Trouble is, there is no general algorithm for saying when a sentence is a logical truth or a logical falsehood so no general way to be sure when it's ok to use "ist". Why not just say that logical falsehoods are false in (and logical truths true in) every context rather than trying to spell out when a context is or is not valid for a given proposition? What's the harm?
For example, to me right now, 'yesterday' is a valid context, because I can think of things that are different between that interval and 'today'. But 'yesterday-10:32am-to-10:42am' is not a valid context because I cannot think of anything (without further investigation) that falls into that particular interval.
So what counts as a context is dependent on someone's ability to remember something that happened during the interval in question? If so, you seem to be wanting ist to be a 3-place relation that includes a parameter for an agent: (ist c John A) -- "John recalls that A is true in context c", or the like. But if so, then what you *really* need is a 4-place relation (ist c John t A) that includes a time parameter -- "John recalls at time t that A is true in context c" -- since of course you might come to recall something (after further investigation) at some later time t that occurred in context c that you don't remember at an earlier time. That doesn't strike me as a sense of context that is of much use for knowledge engineering.
Chris Menzel