OpenPsi assumption of context-free goal fulfilment seems incorrect

[ferrouswheel]

In the OpenPsi README.md it states:

Related/associated goals are those that are the implicand of the
    psi-rules that have the given action in the implicand. That is, if
    more than one rule has the same action, helping achieve different
    goals, then these goals are related/associated. Thus, the execution
    of the given action should result in achieving these goals as well,
    regardless of whether the rules that resulted in the association
    are selected or not. See `psi-related-goals` in [main.scm](main.scm).

However, this is non-obvious to me. I can imagine situations where the same action, but in different contexts, can have a very different impact on goals.

Is this fundamental to Psi theory, or is it an assumption that was added because it made sense for a specific use-case?

[linas]

Don't say "Psi theory". A core unresolved issue of openpsi is that it's a mashup of two unrelated things. These are:

• A mechanism for selecting rules and performing actions, according to a certain priority order based on achieving goals. This is a classic optimization problem. (in the sense of "mathematical optimization", the kinds of articles you can find published in the "journal of optimization" or whatever its called. The famous one.)

• A (not very good) theory of human psychology. (I've read papers offering far superior models of human psychological and emotional and affective states that are mostly just as simple, and more believable.)

Anyway, part of the cleanup is to clearly split up these two components, and make a clear note of how they can be used together, or used independently.

[amebel]

@ferrouswheel you are right, that is a misstatement. It should have stated,

... Thus, the execution of the given action should result in achieving these goals of the selected rules as well..

And, it is related to the mechanism rather than the theory.