Re: object algebra

> > Given EyeColor = {red, brown, NOT_APPLICABLE}. Given the table:
> > Person EyeColor
> > ------ --------------
> > Mary NOT_APPLICABLE
> > Bob NOT_APPLICABLE
> >
> > Then, Mary.EyeColor equals Bob.EyeColor leading an AI program to
> > conclude that Mary and Bob have the same EyeColor.
>
> let color(X) be the color of an eye
> let M be the set of Mary's eyes
> let B be the set of Bob's eyes
> for all m in M, for all b in B, color(m) = color(b)
> Note that this works perfectly with having a N/A value in the
> eye color domain. So this hypothetical AI program will not be confused.

The robot would not be confused internally, but the validity of his conclusions about the "real" world could be incorrect or not quite correct (ie John and Mary have the same EyeColor). If a human boy sees that John and Mary are missing their eyes, and we ask him if they have the same eyeColor, he should respond "They don't or can't have the same eyeColor because they are missing their eyes".

> If, perchance, you are not comfortable with nullary logic, you might be
> tempted to introduce additional clauses to the above:

The problem with this approach is that it is difficult, if not impossible, for an AI-type program to know in advance if the "additional clauses" will apply to new set elements. Basically, it will have the assume the least specific existing relationship between the set and its elements.

> > This examplifies
> > Date's conclusion: "3VL suffers from the very serious ("showstopper")
> > problem that it does not match reality, that is results that are
> > correct according to 3VL are sometimes incorrect in the real world".
>
> If N/A is a value in the domain, then this is not an example of 3VL.

Technically you are correct, but they are closely related and the end result is similar. Masks for NULL can produce mismatch with reality. Received on Sun Feb 29 2004 - 06:38:48 CET