Real talk: Chatbots need some human coaching

Greetings, earthlings!

Building a chatbot shows us how much we take for granted in our daily interactions. While bots may seem far removed from human conversations, many components of successful communication apply to both settings.

Think about the last conversation you had at work about an upcoming project: You probably defined the problem, chatted about possible solutions, and then (hopefully) agreed on a solution to move forward. Customer-facing chatbots can be trained to narrow in on specific goals and conversation paths. Like people, they need to exhibit empathy, transparency, and avoid making assumptions in order to make a meaningful connection with the customer.

Bot guidelines: be nice, be clear, and ask questions

#1: Empathy

Have you ever been stuck in a one-sided conversation where you couldn’t get a word in? Great communicators understand the balance of questions and answers, and bots too must be prepared to interact with both customer questions and statements appropriately.

#2: Transparency

It is crucial that the bot relays both its capabilities and also limitations. Jessica Collier, Bot Designer of All Turtles, said it best in a past presentation: It is the bot’s job to clarify its purpose so the user understands its deficiencies. In human interactions, this might occur during a job interview. Candidates introduce themselves to their potential future managers, explain their areas of expertise, and demonstrate their value to the company. In this scenario, things can turn sour if there is lack of consistency in expectations versus reality. Bot users will be just as disappointed if they feel they’ve been misled about the technology’s capabilities.

#3: Avoid Assumptions

Humans and bots share a limitation: they cannot read minds. It’s true that, between people, asking many questions at the onset of a conversation seems strange; however, bots that make assumptions can do more harm than good. It is accordingly helpful to pose questions early on to ensure that users have a personalized experience.

User Flows & Timing

For a bot to be a great communicator, the foundational layer of empathy, transparency, and lack of assumptions must be present. This layer is essential in how a bot designer thinks about the user flows and timing.

User flows for bots need to account for different ways a conversation can develop. Just like an application or a website, bots maintain user interfaces; however, they are are made up of dialogs instead of screens. Dialogs help designers separate different areas of functionality and determine how conversations might progress. For example, we would create one user flow for browsing flights and a different flow for purchasing a flight or checking out.

Timing is also crucial to the bot’s efficacy; it’s not just about what you say, but when you say it. Well-designed user flows help create smooth conversations that are logical and achieve a goal; however, without timing, even conversations with great copy are unhelpful. For example, in a job search bot, the bot should ask for a user’s age and occupation in the beginning rather than the middle of the conversation while discussing past positions. By asking important questions upfront, the bot can deliver the intended goal with fewer interactions.

Beware the pregnant pause

Furthermore, just as saying something at the wrong time triggers a negative emotional response, so can saying nothing when there should be an immediate response. Conversation analyst Professor Elizabeth Stokoe gave a popular TedTalk titled “The science of analyzing conversations, second by second.” In this presentation, she plays a recorded phone conversation between a young couple and demonstrates how a 0.7-second delay in the conversation translates to trouble. After Gordon says “hello” to his girlfriend, Dana, she hesitates. Sure enough, Dana skips the neutral return of “how are you?” and asks “where have you been all morning?” in an annoyed tone. Gordon likely noticed the delay in response and predicted her negative reply.

Conversely, users expect longer delays while speaking with automated systems. In order to engender trust, the bot should acknowledge this delay by inserting some sort of “filler” that shows that the bot is thinking about its next response. This could be a simple “hmm” or “hold on just a sec.” Used sparingly, these responses build trust with users and put them at ease. This naturally occurs during a phone call with a customer support agent when he/she looks up information. The agent usually says “hold on just one moment while I look that up for you” to let you know there will be a slight delay before continuing the conversation.

It can seem like a paradox: talk like a human but don’t pretend to be a human. The key is figuring out the right balance of personality and clarity to make users feel both acknowledged and assisted.

What will your bot be?

As a software delivery and product company, our team at Wizeline is obsessive about building engaging customer experiences. We work in teams of copy editors, UX designers, bot trainers, and engineers to craft customized bot conversations for brands. If you’re interested in learning more about Wizeline, check out our websiteor a recently published bot success story with our customer, PmNERDS.