Anti-goal reasoning in NAL

[bowen xu]

Hi all,

when researching the goal-reasoning mechanisms in NARS, I found that there seems to be ambiguity on "negation of a goal". 

When NARS is pursuing a goal (that is, to realize the event), it is natural to use the rule

```

A =/> C, C! |- A!

```

(if A implies C, C is to be realized, then A is to be realized)

However what if the system is avoiding an event? For example, to avoid hurt?

In intuition, the rule seems to be

```

A =/> C, avoid C |- avoid A

```

but the current NAL does not support this rule, and I don't think designing deliberately the desire-value function of rule `A =/> C, C! |- A!` can really solve the problem.

So I wrote a paper to address the issue --
Anti-Goal Reasoning: Rethinking the Theory of Goal Reasoning in Non-Axiomatic Logic

(link: https://zenodo.org/records/21207706)

It begins with a counter-example that reveals the ambiguity in the goal reasoning theory in NAL; after analysis, it provides a potential solution that extends NAL, following the overall methodology in the NAL book (that is, defining IL first and then extending to NAL). Four case studies are discussed:

1. Act to Realize

2. Not Act to Avoid

3. Act to Avoid

4. Not Act to Realize

I would like to hear from you if you have ever encountered the same or similar issues. Any feedback on the paper is appreciated!

[David Ireland]

Hi Bowen, 

I had this in my version of NARS (Common Lisp and Erlang)  which was a pattern that would come up a lot in the problems I was dealing with (ethical decision making). 

  (:P (:S =/> (-- :P))
      :infer (with-terms (S)
               (substitute-with-keywords-vars `(,S) bindings))
      :finally
      (lambda (P1 P2 R)
        ;; INVERT DESIRE
        (let ((f (frequency R)))
          (setf (frequency R) (- 1 f))))
   :desire #'truth.deduction))

This rule was applied only to goals. 

I didn't give it too much thought, but anti-goal reasoning does seem under-examined in NARS. 

Thanks for the paper. I'll give it a read. 

Regards,

David

--
You received this message because you are subscribed to the Google Groups "open-nars" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-nars+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/open-nars/8d38c96d-8a9d-4b65-9efc-24dc92e75354n%40googlegroups.com.

[bowen xu]

Hi David,

If I understood correctly, does your code snippet mean such a rule?

```

{ S =/> --P , P! } |-  --S!

```

If so, this corresponds to Case 4 "Not Act to Realize". But in the paper, the prevention relation is explicitly represented. For example, "{ (light, ^press) =/> ^prevent(food), food! }   |-   --(light, ^press)!".

[David Ireland]

Hi Bowen,

Yes, that would be correct.  I'm hoping to go back to my NARS work later this year. Your paper is a nice addition. 

Also, you got a small spelling error, Section 6, 2nd paragraph "imolemented" => "implemented"

Regards,

David

To view this discussion visit https://groups.google.com/d/msgid/open-nars/f26c41f0-e146-4760-9206-a8606c0ef6a3n%40googlegroups.com.

[bowen xu]

Got it. I'll fix the typo and revise it according to other feedback.