We present the results of an experiment in support of our position that communication between Humans and Artificial Agents will only succeed when agents follow the unwritten rules of human social interaction. We have examined two situations of simulated human-agent interaction, where the agent interrupts with sensible but unsolicited suggestions. Our examination of the humans? reactions show the importance of not criticising people?s existing view of a situation, but rather offering information to persuade them to change their view voluntarily.