Conditions for rule application

[Maxim Tarasov]

I'm trying to understand if there are additional conditions for when certain rules can be applied that I missed when reading the book. Here's an example.

Let's take rule `T1 --> M, T2 --> M |– (T1 U T2) --> M`

Can it reasonably be applied to two premises

<(*,{tom},{sky}) --> likes>
<(*,cat,[blue]) --> likes>

To derive `((*,{tom},{sky}) U (*,cat,[blue])) --> likes`

Technically it matches the pattern of the rule but does it make sense? 

In the chapter on compound terms it says that terms cannot contain each other and that they cannot have overlapping evidential bases. Both of those are satisfied here. Should there be some additional check to decide that a conclusion is valid?

Also, we can take another rule `P --> M, S --> M |– S --> P`
The same premises presumably would also match and produce a conclusion
`(*,cat,[blue]) --> (*,{tom},{sky})`

Is that a valid conclusion? Perhaps I'm overthinking it and it *does* make sense but I wonder if I'm missing some key element here.

[stephen clark]

Maxim,

Usually ONA responds that these statements are similar:

`(*,cat,[blue]) <--> (*,{tom},{sky})`

Have you tried out these statements in Open Nars and ONA?

--
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 on the web visit https://groups.google.com/d/msgid/open-nars/1afc48a4-a1c2-4ba2-9ef1-4a576735b173n%40googlegroups.com.

[Maxim Tarasov]

I'm talking more conceptually than about a specific implementation. 

My guess is ONA applies another rule like `P --> M, S --> M |– S <-> P`. 

The bigger question I'm trying to ask is if results like this are valid and what they mean. Especially in the first example with a compound.

`((*,{tom},{sky}) U (*,cat,[blue])) --> likes` How can we interpret this result? It's no longer a relation strictly speaking.

[Pei Wang]

Hi Maxim,

Many valid conclusions are not useful, and this is especially the case for compound-terms, as in your examples. However, this usually cannot be checked from their formal features. What NARS does is to produce many of them with low priority, and leave it to the system's experience to filter out the ones that turn out to be useful.

Regards,

Pei

[Maxim Tarasov]

So in other words, as long as the pattern of the rule matches and the standard conditions for compounds are satisfied (does not contain each other and no overlapping evidence) then the rule can be applied and the result is valid. Its usefulness is a separate topic and is something the system will determine later. Thank you for clarifying.