How Are Chatbots Different From Apps?
This article discusses some of the meaningful differences that set chatbots apart from mobile apps and other common smartphone interactions.
About me
I’m Rucha, a user experience researcher at Syllable in Sunnyvale, California.
What Do We Mean When We Say Chatbots?
When most people think of robots, they think of these images.
Some robots are controlled by human operators, and some robots are programmed machines that carry out physical actions autonomously with some level of artificial intelligence, or AI.
Chatbots, however, are not mechanical robots and therefore have no physical manifestations. Chatbots, or bots, are computer programs that talk to people. They are closer to apps on your smartphone — it is software with intelligence.
Fun fact, Chatbot is short for “ChatterBot” and was originally coined by Michael Mauldin in 1994 to describe these conversational programs.
Today, many software applications behave intelligently. Some of the most popular examples of this are Siri and Facebook News Feed. Siri can answers questions about local businesses, call a ride-share service for you, and even translate English into Mandarin.
The Facebook News Feed shows you articles based on your past behavior of reading articles and liking posts on Facebook.
So many apps are already showing the traits of chatbots.
A Meaningful Difference — How Chatbots Respond to Users
Developing a chatbot involves creating a computer program to complete tasks that would otherwise be done by a human. Chatbots are designed to respond to user input and to learn about the user’s interests. They are conversational, which creates a stronger connection with the user. These user connections are beneficial in areas such as healthcare.
An important aspect of how chatbots are different from apps is how they respond to user input. To explain this further, here is an example of how two popular healthcare apps behave when you are trying to login unsuccessfully.
The apps, as shown above, only provide an error. There is no explanation of what went wrong, no option for help or how to contact a human, no phone number, and no email address. Chatbots, by nature of being conversational, generally provide more information and try to help the user. In essence, they are more empathetic. Below is an example of a chatbot that is trying to understand why the user provided negative feedback to one of the interactions. More on that later.
More About Chatbots and Their Functions
Bots can live anywhere. They can live inside Facebook Messenger, mobile apps, Google Home, Alexa, or a website. Different channels affect how chatbots are presented, and chatbots can also present different content depending on the channel.
Let’s delve deeper into this using a healthcare chatbot that lives in a chat window on a website.
The chatbot greets the user with a call to action, and it invites the user to click on the icon. Once the user clicks, a window pops up with a friendly welcome message that invites interaction with the chatbot.
At first, the chatbot teaches the user how to do quick replies.
It is important to get the user familiar with quick replies because they are a new style of button that appears inside of the conversation. Quick replies are different from regular menus and buttons in apps because the conversation with the chatbot will not continue until the user clicks on the button.
Quick replies put a break in the conversation so that the chatbot can solicit input from the user. More importantly, quick replies provide a break in the interaction so that users can pay more attention to what the chatbot is telling them.
Chatbots and Educating Users
An important aspect of chatbots is educating users. We call this aspect the flow or linear content.
Most content on a website is static — the user scrolls through the webpage and decides what to read and what to ignore. Showing conversational-style content in a chatbot is more effective at educating the user because it is designed to teach the user something.
It is almost like having a one-on-one conversation, pausing now and then to see if the audience understands what you are saying. Linear content can also have branches like a tree, reacting to different input from the user. However, in general, we prefer linear content to be simple and straightforward.
Some Chatbots Can Answer Questions
At Syllable, our chatbots can answer questions that users may have. If the AI is sure that it has the right answer, the chatbot provides the answer without further guidance.
How do we ensure the AI can provide the right answer? We train our AI to calculate a confidence score that determines if the chatbot has the right answer. If the confidence score is low, the chatbot can present the user with up to three suggestions for similar questions that could satisfy their request.
So, how do we know if the AI is doing well? In some situations, we present a thumbs up or thumbs down to the user to solicit their feedback. If the user gives us a thumbs up, we know that we are doing well and that we can stay on this topic and provide the user with more information.
If the user gives us a thumbs down, we know we are not doing well, so we provide them with the option to speak with a human over email or SMS.
This provides the opportunity to collect feedback from the user and helps the AI to learn more about how to handle users’ questions.
I hope this article has shown that chatbots can be designed and trained to develop a connection with users through conversation and to exhibit empathy while improving the user experience.
I enjoyed writing this post. If you enjoyed reading it, please follow me on Medium.
