Designing for Common Ground

Common ground is the foundation of a good conversation. Every time you agree to something or stir curiosity to "tell me more," a conversation adapts. Designing for common ground is more than designing for when things go wrong. It is being polite, building trust and making a lasting impression.

During a conversation, information is constantly updated with each piece of dialogue, whether that dialogue is coming from a human or a machine. For humans, the ability to interpret and respond to new information is core to our natural understanding of language. However, when it comes to designing for machines like conversational assistants, it requires a calculated effort to consider all possible outcomes.

The goal of a conversation designer is to guide users to the next step, which is easy to visualize using Voiceflow’s intuitive mapping interface. While the ability to "repair" conversations is crucial, Voiceflow also offers more subtle methods of designing to create more natural conversations.

Know the User

Designing with the user in mind is a central conversation design principle. In Voiceflow, prototypes allow for personalized experiences through custom greetings, conditional settings and external integrations. The ability to design for any conversational channel opens the possibilities to prototype for various use cases and its users.

Let's take the example of a smart fridge. With the proliferation of voice devices in new channels, it is important that interactions not only help users accomplish tasks, but are also delightful and trustworthy. Given the average person's daily trips to the kitchen, a smart fridge is a great place to establish and maintain common ground.

In the example above, if the time of day is before noon, the user is taken down one path and if it is afternoon, there is a different greeting. If it is breakfast, the smart fridge will greet the user with a message that sympathizes with early risers. Since it is early in the day, there is also the option to give breakfast recommendations. By creating situational context, Voiceflow prototypes robust experiences that simulate reality. 

Remember the past

Smart devices are considered smart because they act in ways that save us time and money. Prototyping default settings adds a layer of personalization and efficiency. Take the following example, which includes the user’s name, time of day and favorite food options. 

Using the Set and If step

Using this information,  the “other greeting” block in the very first example can be further developed.

Using variables to create a more personable experience

Greeting someone using their name is pleasant. Since the default “1” indicates that Sally’s favorite food, pizza, is in the fridge, this recommendation also creates a tailored experience. Using a list of expiring items, perhaps connected through a Google sheet integration, could add more complexity when it comes to food suggestions. 

Changing it up

Hearing the same message loses its charm. Voiceflow knows this and is equipped with features that let designers create what sounds best for a given situation. Varying the welcome message is a simple and effective way to add personality to any system. Beyond creating variants of a speech block, you can add and test different tones, pitches and speed using SSML (Speech Synthesis Markup Language).

Customizable sSpeak variants in Voiceflow

Whenever context changes, the system should respond accordingly. There are several ways to approach this in Voiceflow, one of which is by creating flows. Flows are reusable components that when tailored to specific situations make it easier to know when and how users will accomplish certain tasks. 

An example of a command in Voiceflow

In the above example, the user is not Sally, despite what  the default settings indicated. By invoking the “change username” command on the home canvas, the user is brought to this flow. Once a user gives their name, the system provides a chime sound as confirmation followed by a new greeting. If the “change username” command was started accidentally, the user can “go back” by saying so and keep the original settings. 

The implications go beyond changing a name. Setting contextual commands within certain flows help reduce errors and bring users back on track from a conversation that may be going off on a tangent. Whenever adding new intents or commands, it’s best to use this opportunity to try the NLU/NLP training feature. Knowing which intents are being matched and at what confidence interval ensures that users and the system are on the same page. 


You can activate Voiceflow's debugger mode when prototyping to evaluate intent confidence scores

Final Thoughts

These examples are a small portion of what is possible in Voiceflow. When designing, we should remember there is more to a good conversation than completing a task. The best conversations are the ones that delight in unexpected ways. It is as if the other person or system understands what you're really saying. Designing to establish common ground can accomplish that and more. 

This post was written by Julia Anderson. Julia is a Conversational UX Designer based in Los Angeles who is fascinated by Conversational AI's impact on our everyday interactions with machines and each other.

You can find her on LinkedIn & Twitter or reach out to her via email.