Age of Invisible Machines

CHAPTER 17Best Practices in Conversational Design

As an interface, conversation can readily simulate human connection, making it easier to gather feedback. Imagine a conversational experience that helps new employees find the right classes to help them grow in their position. Instead of designing an interface that renders a browsable list of courses with general ratings, a conversational interface can ask a series of questions that disambiguate, contextualizing who the user is and what they might be interested in. Conversational AI can offer suggestions and get feedback on those suggestions, forging a process of discovery. As certain patterns are identified, users can be mapped to personas that help determine the class they should take. Through building thousands of conversational AI applications over the past decade, I've identified some best practices for conversational design. Here are 57 of the most valuable lessons I've learned. These aren't ranked in order of importance; instead, we'll start with the more general concepts and work toward the more specific ones.

1. Remember that consistency is key: Expectations don't stay flat, they increase or decrease. The safest bet is consistency across the experiences you offer. All of the goodwill and dazzle generated by an IDW that is super advanced in predicting and personalizing can be quickly squandered by another IDW that is dumb.
2. Prioritize personalization over personality: Creating a more personalized experience is more important ...