Planning can support believability because it allows agents to choose reasonable actions and to respond appropriately to the environment.
However, for believable agents, it is not enough to select correct actions; they must also engage in those actions in a way that is understandable to human users.

Current planning approaches can actually undermine believability because they are focused on internal problem-solving rather than external,
communicative effect. Agents select actions based on what will help them solve their goals, rather than on what will make sense to the user. I will describe a system, the Expressivator, that takes the opposite tack: the actions agents select are based on the user's likely current interpretation of the agent, and are chosen to most adequately communicate the agent's thinking and behavior to the user.

The Expressivator is based on the insight from narrative psychology that agent behavior will be most comprehensible if the agent clearly demonstrates not only what it does, but why it is doing it.  The Expressivator reduces the apparent randomness of agent behavior choice common in most behavior-based architectures by adding *transition behaviors*, special behaviors that function to explain to the user the agent's motivations in changing from one activity to another.  In addition, the Expressivator offers a *sign-management system* that keeps track of the visible signs the agent's behavior has produced, allowing
the agent to make decisions based not only on its internal idea of what it is doing, but also on the likely user perception of its behaviors.