[YG Conlang Archives] > [jboske group] > messages [Date Index] [Thread Index] >
pc:
> a.rosta@hidden.email writes:
>
> <<
>
> you are suggesting that {ko'a clani} is a
> proposition that is never wholly true. That may be right, but
> I would also like to introduce a further example that sometimes
> needs fuzzy values and sometimes doesn't.
>
> 1. He is in the room.
>
> If he is half in the room and half out, then he is sort of in
> the room -- the truth value of "he is in the room" is between
> pi no and pi ro. If he is entirely in the room, but only by a
> few inches, then the truth value is pi ro, but nonetheless he
> is barely in the room, and he is 'less in the room' than
> someone standing further from the doorway.
>
> >>
> As usual here, it is important to distinguish between how true some
> claim is and how much the claim is ({jei} and {ni} in Lojban). If
> the guy is only a few inches inside the door, it may be that it is
> absolutely true that he is inside the door and yet the quantity of
> his being inside is less than that for a may who is several yards
> inside the door. The two can vary independently of one another -- or
> nearly so.
How do they vary independently?
I think it is conceptually attractive to blur the difference.
I don't deny the difference altogether: I wouldn't say that the
truth of {ko'a clani} is measured in metres, even though {lo'e
ni ko'a clani} might be (I guess).
I think of degrees of absolute truth being measured in terms of
how much the world would have to change in order for the proposition
to be false.
> xorxes on &
> <<
> > I don't disagree with anything you've said (except that it needs
> > to be clarified, IMO, that .9 entails "not (wholly) true").
>
> It all depends on how you define NOT for continuous truth values.
> A value of .9 is not a value of 1 just as it is not a value of .8.
> But we probably want a softer "NOT" for continuous truth values.
> For example, a function that maps value x to value 1-x.
> Then {.9 <bridi>} does not entail {not 1 <bridi>} = {0 <bridi>}.
> >>
> If we are off in open-ended truth, which begins at 1 and goes on, the
> negation is going to be trickier than this supposes. On this view,
> tv .9 is absolutely not truth, [1+, but it is not falsehood either
> and it is closer to truth than to falsehood, [0- . Your proposal is
> Boolean, but that won't work here.
I'm not sure whose proposals are meant here.
Anyway, I don't think it is a problem to apply NOT to SORT-OF (or
specific values of SORT-OF), though I acknowledge that something
more like {to'e} for a softer "not" is also needed, but it
definitely boggles my mind to apply SORT-OF to NOT or JA'A, unless
SORT-OF NOT p = SORT-OF p. I think I'd need to set this out on
a wiki page (a sister to Jorge's on Three Valued Logic) -- it's
too difficult to keep a grip on it in email.
--And.