4 Designing effective dialogue

This chapter covers

Writing dialogue that supports your process flows
Creating dialogue with tone, empathy, and variety
Asking questions that increase your chances of getting a useful response
Responding to conversational mishaps

In chapter 3, you learned how to design effective processes that your conversational AI assistant will implement. After you design all the flow charts and user-to-system interactions, you need to breathe life into them by writing the dialogue text your assistant will read or speak.

The assistant can respond in several ways to a user who needs to reset a password, as shown in figure 4.1. A response of “I need your user ID so I can reset your password” is direct but has a nearly accusatory tone. A response of “I’m sorry you can’t access our system. I’ll try to help you. To start, what’s your user ID?” conveys a softer tone and lets the user know what’s happening next. From a programming standpoint, both of these responses are identical, but users will not perceive them the same way.

Figure 4.1 Motivating example for the chapter. This process flow is a good one, but what will the assistant actually say at each of these prompts? How can we prime users to answer questions in a way that moves us through this process? What tone will the assistant use at each prompt?

4.1 Writing dialogue

4.1.1 Take a conversational tone

4 Designing effective dialogue

This chapter covers

Figure 4.1 Motivating example for the chapter. This process flow is a good one, but what will the assistant actually say at each of these prompts? How can we prime users to answer questions in a way that moves us through this process? What tone will the assistant use at each prompt?

4.1 Writing dialogue

4.1.1 Take a conversational tone

4.1.2 Don’t repeat yourself (much)

4.1.3 Acknowledge the user

4.2 Asking questions

4.3 What if the assistant doesn’t understand?

4.3.1 Reprompting

4.3.2 Disambiguation

4.3.3 Escalation

Summary