Hi!
There can be modalities (usually spelled as diamonds or boxes in modal
logics):
DUTY_TO_PERFORM_ACTION(agent, action, time horizon) - agent should perform
action in time horizon
BELIEF(agent, fact, time instant) - agent believes in fact at time instant
Such modalities are necessary to model robotic rules (e.g. AGI safety
question) or express how one agent understands another agent beliefs and
hers reasoning type (i.e. knowledge that is strongly required for
communication).
There are two questions about use of modalities in atomspace:
*1) how to express modalities in Scheme?* Maybe one can introduce new links
for each type fo modality, but my guess is that we can simply remain at the
PredicateNode level - there can be predicate that expresses the existence
of the modality among instances of agent, action and time moment:
PredicateNode "exists_belief"
ConceptNode "Erving"
ConceptNode "Money is good"
I guess - it can work.* But my question is - an be use any other Node (like
ConceptNode, like Instance of some concept, result of SatisfyingSetLink and
so on) instead of the literal constant "exists_belief"?* Use of other Nodes
in the PredicateNode would allow great flexibility in modelling expanding
set of modalities!
*2) how we can introduce new meta-rules with each new modality.*
https://en.wikipedia.org/wiki/Modal_logic
has list of modal logic axioms that expresses how modalities behaves in
inference process. Such axioms/meta-rules are quite debatable but
nonetheless they should be used in some case. Some popular example is
IS_DUTY(agent, action)->IS_PERMISSIBLE(agent, action). *The question is -
how can be express such meta-rules in OpenCog?* If we model modality as the
predicate then there should be some patterns which expects certain type of
predicates in its body and creates certain new predicates in its action:
pattern/rule_body(predicates_with_belief_modality_expression)->conclusion/rule_hear(predicate_with_some_other_modality)
Of course, I am thinking and experimenting with this, but - maybe others
have had similar thoughts or experiences in such direction, so - it could
be nice to hear them.
Thanks!
--
You received this message because you are subscribed to the Google Groups
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit
https://groups.google.com/d/msgid/opencog/a67ba437-0674-4370-b2a3-fdccb1a85db7%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
