We consider defining executable dialogues for communicating agents. Towards this end, we introduce agent classes whose communication primitives are based on deduction. Their operational semantics are given by an abstract logical machine that is defined purely in sequential terms. These agents communicate under the control of plans requiring a synchronization flag. These plans can be rewritten as dialogues with an implicit synchronization. Reversibly, dialogues can be compiled back into plans and then executed on the sequential machine. Sub-dialogues can be entered from any dialogue, such achieving dynamic conversation structures.
展开▼
