Re: SNAF, NAF, and monotonicity [was: Comments on * DRAFT * Rules...]

Ok, let me rephrase what I tried to say:

1. There is obviously something about SNAF, CWA and non-monotonicity 
that led Sandro to write a contradictory draft, Dan to refuse that 
007car be yellow and me to make a fool of myself (well, the latter may 
not have anything to do with SNAF, CWA or non-monotonicity, actually);

2. A useful way to think informally about CWA is as if it were an 
inference rule (just out of curiosity, is there a short explanation, or 
a pointer towards a long one, why it could not be an inference rule, 
actually?): that way, I can see clearly why adding a fact can make 
previous consequences false (from a purely deductive point of view, so, 
without having to think about interpretations);

3. A fallacious way to think about CWA --a big no-no, don't do that, 
ever, it doesn't make sense, anyway; but, still, sometimes you cannot 
help-- is to assume that the world is really closed, and that we have 
complete knowledge about it, and thus that we can act as if all the 
missing negated facts were there, and then we just work with good ol' 
classical negation. And then, of course, it happens that we forget that 
it is just an assumption and, so, we are not happy when reality breaks 
in and adds some facts that contradict the assumed negated facts (we are 
unhappy because it contradicts our initial closed world assumption, not 
because of the non-monotonicity). Could it be that Dan was careless 
enough to trip into that kind of fallacy, at early some point in the 
discussion?

4. On the other hand, this is what monotonic inference mechanisms do, 
for all practical purpose: within the scope of the assumed closed world, 
they reason monotonically; when the closed-world assumption is broken or 
before it breaks, for whatever reason (and, possibly, because of actions 
triggered by the inference system itself, as in a production rules 
engine), the engine is made to stop and, if and when more inference are 
needed, to restart with the new initial conditions. That's 
non-monotonicity all right, if you consider the whole system, including 
the part that causes the engine to stop when or before the CWA is 
broken; but if you consider things from inside the inference engine, 
there is nothing but monotonicity.

(see attached use case scenario: thanx to point out what is wrong with 
it and why, if something is).

5. It is clear to me that NAF is needed if we want the RIF to be of 
practical use. It is clear to me as well that the scope of NAF may need 
to be specified in some way when sharing rules.

Now, what is not clear to me, is: what kind of features specific to 
non-monotonicity does a rules interchange format need to have, to deal 
with that use case, besides the capability to unambiguously specify the 
scope of a NAF? More generally, what is not clear to me is: what kind of 
features are needed in a RIF wrt non-monotony, besides the capability to 
specify and scope unambiguously the nonmonotonicity-generating features 
in a rule (such as NAF, default, etc)?

If the answer to the first question is: none, couldn't we repair the 
draft charter (wrt that issue) by removing section 3.1 and rephrasing 
section 2.9 to say that SNAF is in scope and the RIF must offer ways to 
specify the scope of NAF. That is, after section 2.3 has been amended to 
avoid the contradiction between such a statement and the RIF being full 
FOL, of course.

Michael Kifer wrote:
>
> [...]
 >
> We don't need to care about anything. We just need to understand what the
> issues are and be aware of simple things that have been known for decades.
> 
> This whole thread started because the draft charter contained several
> contradictions. Clearly, it is not healthy to base one's work on a document
> that is internally contradictory. All we did was to point this out.

Once you pointed out that, the issue becomes that there must be a reason 
why the draft charter says that SNAF is in scope, but that NAF and 
non-monotonicity are out.

So, the interesting questions are: can we remove the contradiction and 
keep the intention,? And, if yes, how? And, btw, what was the intention?

Just for the avoidance of doubt: my postings on this issue are meant to 
make progress on these questions (whether they do is something 
different, yet), not to question the foundations of formal logic, model 
theory or whatever.

Christian