July 15, 2019

The reason most chatbots are poor listeners

Article

Using intelligent listening to provide a superior chatbot experience

I want to spend a minute discussing the importance of intelligent listening when building effective chatbots. Anyone that has frequently interacted with chatbots has probably had some rough conversations in which it seemed like the chatbot just wasn’t listening at all. Well, the truth is, you’re most likely right. That chatbot wasn’t listening to you at all. This can cause frustrations for users that are used to a certain cadence and flow in their everyday chats whether it’s via Slack, Facebook Messenger, or just plain old texting. The issue arises from the fact that we humans rarely complete a full thought in a single chat message. Instead, we break that whole thought into smaller messages. Some of us even change our minds on the fly and swap a detail that is crucial to the meaning of what we’re trying to articulate. The example conversations below show the difference between a chatbot with intelligent listening capabilities and another without.

Chatbot Intelligent Listening

In our example comparison of two different chatbots that assist with travel bookings, the chatbot with intelligent listening capabilities waited to learn whether or not the user finished their complete thought. The typing indicator bubbles provided the user with confidence that the chatbot had successfully received their messages and was preparing a response. What is happening behind the scenes here is that the chatbot is waiting for three seconds to pass, without the user typing, before determining that the user is done with their response. This allows users to chat with the bot in small, segmented bits that make up a complete thought. This is ideal since it is how we most commonly interact in our everyday chats with other humans. The chatbot without intelligent listening tried to respond immediately to each message the user sent without knowing whether or not the user was actually finished with their response. Not only did this break up the natural flow of the conversation, but it caused the user to have to repeat things several times.

Clearly, the user interacting with the chatbot without intelligent listening got quite frustrated at the end. To further distinguish between these two chatbots, the chatbot using intelligent listening was also leveraging conversational context by using the user’s phone number to look up their past travel history. This is another critical component of effective chatbots that we highlighted in a previous blog post on the importance of context. So if you’re looking to build a chatbot that users will love talking to, make sure it listens to them or prepare for some angry responses like the one above.