Using this framework, we can formally characterize the
(normative) conditions under which information exchange
should take place. Information I should be sent from one
agent A to another agent B when: 1) agent A knows the
truth-value of I, 2) agent A believes that agent B does not
currently know I, and 3) B has a current goal G, the
achievement of which depends on knowing I, i.e. if B does
not believe I, then it will never be able to accomplish it’s
goal, but if it knew I, it would be able to:
(BEL A I) Ù (BEL A Ø(BEL B I)) Ù
(BEL A (GOAL B G)) Ù
[Ø(BEL B I) ® Ø(DONE B G)] Ù
[(BEL B I) ® Ø Ø(DONE B G)]
® (GOAL A (Inform B I))