Here’s How to Design Human-Like Chatbot Conversations
User interface design has always been a crucial part of the digital product development process. UIs represent the connection between people and technology, they’re how we interact with digital products. And those interactions form the experiences we go through while using a software or an app.
As users we naturally gravitate towards interactions that feel natural to us. We want frictionless, lightweight, and enjoyable experiences that represent the extension of our daily lives. In essence, we want the interface to get out of our way and let us do “our jobs” in the most efficient way possible.
This need opened up a new frontier of interaction design — conversation as a UI. Messaging is the new platform and conversational interfaces, such as chatbots, are surfacing as a way for companies to satisfy user need for more natural human-to-software interaction. At least they’re trying to.
Companies are Getting Chatty
All the cool companies today want to use chatbots to deliver personalized customer experiences at scale, something that was nearly impossible to achieve before.
Despite their exploding popularity and an infinite number of chatbot use cases, majority of bots fail to deliver on the promise of smart, personalized, and engaging interactions that fit the messaging platforms people are using daily.
More often than not, we can track this failure back to the bot’s inability to converse with customers in a meaningful way (e.g. natural, contextual, problem solving), while appropriately communicating brand personality and voice. And this, of course, has everything to do with design.
It’s quite simple — language rules the world of conversationals UIs. When designing a chatbot you’re designing for conversation and words are the biggest asset you have to create amazing experiences for your users. The language you choose and how you decide to structure the conversation flows will help drive (or decrease) user engagement and retention. Users want to be hooked into a conversation which only happens if it flows naturally and efficiently, and the other side — a bot in this case — shows personality traits that they’d find appealing.
Focus on Designing Human-Like Interactions
We’re more likely to emotionally attach to lifelike characters, but creating a bot that “lives” within the constraints of a conversational UI and that people can interact and connect with it in a meaningful way is not an easy task.
Conversations are easy for us humans, but they’re very hard for bots. You need to work around their limitations and design your bot so it’s capable of empathising with users.
To do that you’ll need to really know your users. Who do you imagine will make up the majority of your audience? Why would they use your bot? What problem/need are they trying to address? Answers to these questions will inform how your bot “talks” with users. So many chatbot designs fail because conversations are based solely around a brand’s goals, neglecting what the customer is looking to get out of those conversations.
We all know that the bot isn’t a person, but we still want to a) connect to our image of the brand that we’re interacting with via chatbot, and b) feel that it empathises with us and understands the context of our situation.
Build Strong Bot Personality
When we built SavaBot, we knew from the start that we wanted him to be a witty, goofy, and playful character that’s always ready to help. A bot that would lighten up your day while you’re suffering through a hectic conference schedule. We recommend that you do the same.
Establish core personality traits for your bot first, and then plan the design around them. This will make it more authentic, relatable, and relevant to users. Is your chatbot going to be funny, sarcastic, serious? Knowing this will help you choose the language you want to use — the right words and phrases for all future interactions with users. This will also be the right time to decide what type of content you want it to deliver. You can play with the combinations of text, emojis, videos and photos.
Once you’ve set those personality traits or human-like characteristics for your bot, stay consistent and never stray from them. It would be very weird for a banking chatbot that uses a formal, serious tone to start telling weather jokes in a middle of conversation about your finances, right?
So don’t confuse your users with out of character or ambiguous messages — this will only increase the cognitive load, frustrate your users, and make them abandon your chatbot.
Go for Engineered Spontaneity
Conversation is spontaneous rather than planned communication between two or more people. This quality makes it difficult to replicate in a human-to-software interaction. So how do you create spontaneity in your bot’s conversations with people?
Planning for spontaneity sounds like an oxymoron, but that’s exactly what you’ll need to do to train your bot accordingly.
The following tips will help you design a bot that will appear more spontaneous and more human-like while interacting with your users.
1.Keep it short and sweet, with a hint of unpredictability. No one wants to read big chunks of text, especially not on a phone! Like in real-life conversations, this gives an impression of talking at users not with them. By keeping it short and providing varied responses that users won’t always expect, your bot will better mimic the dynamics of human-to-human conversation.
2. Prepare fallback responses. You can’t predict everything that bot users will say or ask, but you do need to plan what your bot’s responses will be like if it doesn’t understand the user. Written well, fallback responses will help keep the conversation on track and guide the users towards the next step. Your bot will need to respond in a way that matches the user’s input and based on the words or phrases that triggered your fallback.
3. Be prepared for interruptions. Interruptions are a part of normal conversation. Don’t think this wont happen becuase your users know they’re chatting with a bot. When interruptions happen, your bot needs to be able to respond in line with its personality. That’s why it’s crucial to create a set of interruption rules that the bot will follow to respond to certain interruptions, no matter when they happen in the conversation flow.
4. Design natural-sounding conversations. Sounding natural takes more effort than you might expect. It requires a well-defined bot personality and skilled UX or design writing to make a chatbot sound natural while conversing with your users. If you make this magic happen, however, you’ll see amazing results in terms of user retention.
5. Use guided responses. This is a type of predetermined copywriting that you’ll use to offer multiple choice questions to users. By giving your users options (e.g. buttons to push) and not letting them guess what to ask next you’ll create a better experience and this will show in increased engagement rate.
6. Nail down the conversation flow. This is probably the most difficult part of chatbot design. Conversations between people are typically non-linear, often unpredictable, and it’s not easy to replicate that experience within a conversational interface.
Mapping up an entire conversation can be challenging, but what you can definitely do is create a so called “happy path” or an ideal scenario and work out other possible variations of what users may say or ask. You can’t predict every user question, but you can have mechanisms in place to get your users back on track without them even noticing.
Takeaways
While it’s relatively easy to build a chatbot, designing one that users actually want to chat with is everything but. There’s a lot of awkwardness that you need to help your users overcome so they can feel comfortable interacting with a bot.
So when designing your bot, make sure to add a human touch, emulate conversations between people, and use the language that feels most natural to your users.
Thanks for reading, you made it to the end! If you enjoyed reading this, please give it a clap and share! :)
Originally published at spicefactory.co on May 18, 2017.
