Request and propose

In Section 5.5.3, it was shown that in a cooperative setting, once an agent agrees to always, unconditionally execute the object of a request act, there are times when it can be more efficient to use than a propose act. However, suppose that agent 2 were to make such an agreement, and then, due to differences in belief, agent 1 requests an alternative that agent 2 does not prefer. Agent 2's rational choice is then to break the agreement. It would seem then that the rigid pragmatic form of the request act is impossible. The experiment has shown that the strong form of request can be superior to the weaker propose. However, there is probably a trade off between the utility offered by this strength and the utility offered by allowing the second agent to break the agreement. There might be some continuum of acts ranging in strength between propose and request, and the agents should plan an expectation that the agreement will be broken, especially in a self-interested setting. Further experiments might show that each act on the continuum is dominant in some region of the belief space. The contrast between the two acts also needs to be explored in the self-interesting setting, since self-interested agents are more likely to prefer autonomous forms of proposing.