Why Do We Need to Design Conversations?
Human conversation is varied, complex, and contextual. How can we teach machines to seamlessly chat with us?
If people have been talking to each other for thousands of years, why do we need to design conversations? Because now it’s not just people you can chat with, it’s technology. Language and its nuances are the cornerstones of human conversation. We interact not just with words, but with inflection, tone, and non-verbal cues like facial expressions and body movements. Teaching a machine to participate in a conversation is essentially what’s defined as conversation design.
Get started designing conversations
Most of us have had experiences with conversational products that have left us feeling frustrated. Reasons could range from machines not understanding our accent to not having an answer to your question. We’ve all been caught in endless loops of a bot saying, “I’m sorry, I didn’t understand that, can you please try again?” with no path forward. Each of these requires attention to conversational design.
For machines to understand user queries, we have to tell them what users might ask and how to respond. A huge part of conversation design is combing through call logs and chat transcripts so you can understand your bot use cases. Combing through call logs is called utterance analysis and requires an understanding of how machine learning models work.
To avoid looping, conversations need a strong backbone of interaction design. It’s not a conversation without turn-taking between a machine and a human. Designing for conversations, especially in natural language, is different than designing for static interfaces where you can keep the user interactions within bounds. Conversation design is fluid and can move in many different directions. The key is to anticipate what the user might ask and design for those interactions and conversation repairs.
A conversational product is never done. User queries change over time with influence from social, political, and environmental factors. For instance, when the Covid pandemic started, most chatbots started seeing covid-specific questions which is why continuous iteration is required even after a bot is shipped. The user’s behaviors can determine if new responses are needed, or if we should reclassify intents (goal of the user when asking a question), or retrain machine learning models.
Tools for creating seamless conversations
Conversation is a mode of interaction — similar to touch, click, or gesture. A conversational interface allows you to converse with a machine using voice or text. With the advent of graphical conversational interfaces, we can now complement words with rich user interface (UI) elements to enhance the experience for users.
Multi-modal interfaces are one of these UI elements designers can use to improve their bot conversational skills. It’s defined as an interface that uses two or more input modes to interact with machines. Even though language is a primary component of conversation design, graphics complement the words and allow designers to stretch the use cases.
A human can speak three times faster than they can type, and we can read even quicker than that. Combining graphical elements with written or spoken words can enhance any conversation when done mindfully. And when crafting conversations, all these elements come together to complement, not compete with, each other. You don’t want to litter your conversational interface with words or graphics unnecessarily.
So far we have seen how designing conversations requires an understanding at the intersection of language, psychology, interaction design, and AI. It takes an amalgamation of many skills to teach a machine to understand humans and infuse their speech with the necessary emotion and clarity to create a seamless user interaction.
Conversation Design, when done well, can create a more inclusive experience that allows people to interact with machines in the most basic form — words.
This article is part of a series called Crafting Conversations. Check back for more articles where you’ll learn how to design great conversations and get advice from our team of designers and linguists.
Check out some of the other articles in the series:
This series is a collaborative effort between a team of conversation subject matter experts. Thank you to Michal Angel, Austin Bedford, Greg Bennett, Rachel Blank, Marlinda Galapon, Denise Martinez, Jonathon Newby, Shalini Johar, Isabela Berbel, Brent Laing, Madeline Davis, and Margaret Seelie.
Learn more about Salesforce Design at www.salesforce.com/design.
Follow us on Twitter at @SalesforceUX.
Check out the Salesforce Lightning Design System.
