[Attempto] [CNL] Should I allow for non-monotonic reasoning?

[Norbert E. Fuchs]

> On 25 May 2015, at 10:40 , > <mottdh at googlemail.com> wrote:
> 
> Good luck with that. I am currently wrestling with the same issue in the CNL that i use. there is already a way to run nonmonotonic reasoning, based on a sentence such as "it is assumed that ...". But i have been working with folks from Aberdeen University on the use of argumentation, and I realise the logic is potentially more complex than I had thought, with different options for the underlying semantics.
> 
> Whereas it is possible to implement such semantics, I think the real challenge comes when making sure that the users, who may not be logicians, understand from the CNL itself what is going on when the system makes non monotonic inferences. I have found that users are happy with the concept of assumptions, I'm not yet sure how happy they will be in distinguishing between different semantics, eg the difference between negation as failure and negation as classical logic. This is a simple case, Im sure there are more complex ones hiding away. But since, again, argumentation is intended to mirror human reasoning, perhaps concepts from argumentation theory is the way to go?

David

This gives me food for thought, specifically concerning negation-as-failure that is part of ACE, but that RACE does not handle. Also I agree with your statement "the real challenge comes when making sure that the users, who may not be logicians, understand from the CNL itself what is going on when the system makes non monotonic inferences".

In this vein I tried – but until now was not completely successful – to get default reasoning with the help of modality as in

  Tweety is a penguin. Every penguin is a bird. No penguin can fly. If X is a bird and X can fly then X flies. |- Tweety does not fly.

which succeeds but does not report the correct minimal number of axioms needed for the proof, while

  Tweety is a penguin. Every penguin is a bird. No penguin can fly. If X is a bird and X can fly then X flies. |/- Tweety flies.

correctly fails.

Best regards.

   --- nef

PS. Since you are not subscribed to "attempto at lists.ifi.uzh.ch" this copy of your message is discarded automatically.