From OpenCog

This page captures a suggestion regarding a tweak to AttractionLink, as discused in


The suggestion came up in a discussion of reducing the number of "shuffling-around" inference steps in the following example inference:


When we say

AttractionLink X Y

we are considering X and Y as sets and calculating Inheritance relations between X and Y, and not-X and Y, on this basis….

What comes up in the above inference chain, is that we may also want to say

AttractionLink x Y

considering this as the same thing as

AttractionLink (SetLink x) y

To do this in one step without extra shuffling-around rules, would require a new AttractionLink-like type, say

ElementAttractionLink x Y

I suppose…