What is conversational UI?
A conversational UI is a User Interface that uses a conversation as the main interaction between the user and the software. It’s an umbrella term that can encompass both chat-based interfaces, such as a Slack chatbot and voice interfaces, like Alexa. A conversational UI is possible only if the software manages to do two main tasks. First, it needs to capture the user’s intent even if expressed as natural language. In other words, a conversational UI should be able to understand, to a certain extent, what is that the user wants. The user should be able to express himself in the most natural way possible. This task is often achieved by using a Natural Language Processing engine. Secondly, the software powering a conversational UI should be able to express itself using natural language. In other words, the software should be able to sustain the conversation. Note that usually the conversational UI is built in such a way that the software ‘talks’ using a set of phrases designed by its creators. There have been attempts to use a Natural Language Generation bot in commercial use, with scarce results – such as the spectacular fail of the Microsoft Twitter bot.
A conversational UI is just another type of interface. You can think of the different, existing platforms as you think of devices in responsive web design. In other words, Slack, Facebook Messenger, Amazon’s Alexa and others allow you to release your own conversational UI on their platforms. Your conversational UI design will need to adapt to the different nuances of the different platforms. On the other hand, you want to maintain a consistent user experience across different devices and platforms. In that regard, the conversational UI needs to be adapted the same way that a website’s design needs to be adapted and changed for mobile applications.
Best practices for Conversational UI
Recently, I got an opportunity to design a chatbot for GameOn Technology, a pioneer in building intelligent chatbots for world’s largest enterprises. After two months of intensive research (although limited resources available online), great feedback from GameOn, multiple design iterations and critique sessions, I present my usability findings in the form of UX practices that can help you design better bot experiences.
I understand that designing a conversational UI for the first time can be overwhelming. Some of the key points mentioned below can help you get started quickly in the right direction.
Dialog based on the Personality
Most blogs suggest to add a personality to your bot and I’d agree. As we all know
Personality captures the heart.
Since words are the only way of going about it in conversational UIs, use your words wisely. For example: A professor bot can have an authoritative personality, while an assistant bot might sound rude with a similar tone.
Beauty attracts you, Personality makes you stay.
Sometimes people ignore minor flaws if you have a great personality. This applies to our design too. More trust and retention is possible if your script makes your user want to stay.
Difficulty in mapping conversations? Start here!
This diagram can greatly help in bringing the anxiety down. It scopes down what seems like infinite possible interactions to a finite number of features the bot can offer. I would suggest building on this diagram as you progress in your design process. It helps you record your features, interactions, script, and flow for all possible scenarios.
For more help on developing a conversation diagram, check the article I referred to for project.
An introduction should be as small as an elevator pitch — People don’t like to read long passages. Also, it must include what the bot can do (think of it as avoiding the initial awkwardness/small talks).
Note: Things that do one thing well are more helpful than things that do multiple things poorly — Remote control is a good example.
First impressions influence overall opinions
Making it clear that the user is chatting with a bot and not a human is very important. We want to design smart conversation, as human-like as possible but wait until Artificial Intelligence is more advanced or pay the trust tax. Trust tax says one undesired interaction with the device loses 3 times the trust that can be achieved by 1 desired interaction — Heavy duty stuff!
We want to reduce the TRUST TAX
Once the trust is restored, the users will give a second chance to your brand (if not the bot)
Introduction should be as small as an elevator pitch but also include what the bot can do (think of it as avoiding the initial awkwardness/small talks). And things that do one thing well are more helpful than things that do multiple things poorly — Remote control is a good example.
Communication is a two-way street — Listen before you speak!
Heard about this multiple times? Now let’s implement it. We don’t want the user to feel the bot is forcing it’s offerings on them. At most points, the user should feel in command of the conversation and should be able to direct the conversation according to his needs.
A good balance between the bot suggestions and user inputs is essential.
Visual Hierarchy Helps!
As we all know, people don’t like to read long passages. Using simple language and short messages always help. But something else that can help but is not implemented yet is a visual hierarchy. We can learn a lot from typography here — Emphasizing the text, highlighting your point, and fonts can help portray emotion and context.
Designing text interactions — Keywords are not just links but are the new buttons that perform the required tasks. Every interaction with your bot does not have to be conversational.
6. Design for Failure
As they say, you should learn from your failures. Let your users learn from it too. Chatbots have limited scope and will FAIL for out of context conversations (unless linguistics and natural language learning capabilities come into the picture).
How you handle those scenarios and bring the user back to what the bot has to offer is designing for failure. This can be a point where you lose your users and hence designing for these use-cases is important.
I have found these points very helpful while designing for conversational UIs. It improved the success rates of tasks and increased satisfaction among the users as gauged from the user tests. Once you have everything figured out, I would highly encourage prototyping and testing your design with real users to find more insights in your context.
Also, as a learner, I’d love to hear what variations or new concepts helped you in your design. Leave your questions and comments below.
Also published on Medium.