UX Guidelines for Conversational Design
Today I want to write down a list of UX guidelines for conversational design.
During my research, I’ve highlighted the main sections to whom chatbot conversations should be categorized using a UX approach.
Those categories are eight:
- Acknowledgments
- Confirmations
- Prompts
- Empathy
- Inferring solutions
- Legibility
- Errors steering
- End of the dialogue
Intro — Why UX guidelines are important in conversational design?
UX is an important driver for success. It integrates into AI-enabled products with UX Frameworks.
A UX Framework is a method of considering user experience while designing an AI application. It puts the user at the center, not the technology.
The context includes information about the user, why and how they are making the request, and the external world.
The interaction is the way our AI engages the users, it could be a text message, a call to action, a sound, or a smartphone notification.
Trust is when the user feels that an AI-enabled application will perform a task that a user wants to perform. It is sticky, if they don’t trust it, they will continue to mistrust it (and the other way around 😉), that’s why it is vitally important to user adoption.
Read this book to get more info about AI-UX frameworks:
UX guidelines are important in conversational design because they allow product teams to work on the same UX framework, to guarantee high quality and consistent interactions with users, and gain trust.
UX is crucial to determine the failure or the success of a digital product. Read these two articles about the relation between UX and Siri and Alexa:
- How Siri Got Off Track — The Information
- The inside story of how Amazon created Echo, the next billion-dollar business no one saw coming).
01. Acknowledgments
An acknowledgment assures a user that the input is received.
They work as positive reinforcements, which is one of the keys to a good UX. Read these articles to deep-dive the relation between Positive Reinforcement and UX:
Types of acknowledgments
There are three main categories of acknowledgments in conversational design:
- Words-based acknowledgments
- Visual acknowledgments
- Audio acknowledgments
Words-based acknowledgments
Words-based acknowledgments are expressions like Ok! / Cool! / Sounds good! / Sure! / Alright!
Use them at the beginning of each answer. Load a subset and set them to appear randomly, this will help the chatbot to appear less robotic.
Visual acknowledgments
When I say “Ok Google”, at first, it uses a light-blink as a visual acknowledgment to assure me it got my input, and ease me to make my request.
Audio acknowledgments
Many IVR systems or voice chatbots use audio acknowledgments like pop-sounds, to assure the user that the input is received.
02. Confirmations
A confirmation assures the user that we got the request.
We can do it explicitly, or implicitly if the sentence is contextual and already includes a confirmation.
Words-based confirmations
Audio-based confirmations
If I ask: “Ok Google, turn on the light”.
Then Google turns on the light in my living room and closes the dialog with a pop sound (audio acknowledgment), to assure me that the action was done.
03. Prompts
We can say it’s always better to end a sentence with a prompt. It helps the user to know when it’s his time to speak and balances the conversation.
This guideline refers to Grice’s Maxim. Read more about this:
How many suggestions can we propose in a prompt?
It’s better to propose a short amount of buttons to avoid cognitive overload:
- Vocal conversation: max. 3
- Display conversation: max. 5
See Alexa’s UX Guidelines to read more about the maximum number of suggestions to propose in a conversation.
Take this as a base rule for your conversation designs: end your dialog always with a prompt.
This helps the user to easily find a way-out and helps the natural flow of the conversation.
04. Empathy
When a user asks us to solve a problem we first need to empathize with him. Then we try to solve the problem or we hand over with a human assistant.
The goal is to makes users understand that we got their frame of mind and that we are going to make the best to help them to solve their problems.
How can we do that?
Mapping use cases we know are more delicate for our users and add a sentence that expresses our empathy for them.
Remind empathy, not sympathy!
Can we use emoticons?
Yes, emoticons are commonly used in our conversations and they help to express our emotions through text.
[On social media, they rise like rate of 57%, and comment rate of 33%, so we assume they would rise the engagement with our conversations.]
05. Inferring solutions
Remind this UX rule: it’s better to infer than ask.
Imagine proposing a pizza to a friend of yours. Which conversation do you prefer?
06. Legibility
It’s better to divide messages and use one message for each topic and bold help the legibility.
Here I compared the same sentence written in one and multiple messages:
07. Errors steering
It’s usually hard for a user to ask for support clearly at first. Sometimes they are on a bus, at work, or playing with their children. It’s quite hard for them to be 100% focused on their request.
If they are not clear, we must help them to get back on track.
Setup a dialog for not-routed user requests, keep the control, and kindly ask to repeat. Never repeat the same sentence, it’s quite annoying for the users. Change it, always!
See this example:
Conclusions
These guidelines help Product Teams to rise experience KPIs as NPS, CES, and completion rate of tasks.
I really hope this will help you to build amazing conversations for your users.
Keep the conversation going 🖖
Do you agree with these principles? Please, applaud the article to increase its visibility and get more feedback from the community.
Please, leave a comment below and help the Conversational Design Community to improve.
