Five Best Practices for Conversational UX Design

If you are creating a dialog for the VUI of an Intelligent Personal Assistant (IPA) or a chatbot, your main philosophy must be to implement a customer-centered voice solution. For a voice experience (or even for a multi-modal experience guided by voice), such goal maps into the following main design actions:

  1. Create a conversational interaction as natural as possible
  2. Think about the problem that needs to be solved
  3. Define a persona for the IPA/chatbot
  4. Provide the IPA/chatbot with conversational memory
  5. Contextualize universals in a natural way

1. Create a Conversational Interaction as Natural as Possible

Alexa, will you be able to speak like a human one day?

As a conversational UX designer, you must think about the entire conversation, from beginning to end, from the point of view of how the human will experience such interaction with the IPA/chatbot.

The materialization of such perspective translates into the application of conversational practices and structures used by humans into the design of the human-to-machine dialogs.

The design actions you can implement in your solution related to this point are:

  • Complete “one-offs” voice commands with more complex models of interaction/conversational elements extracted from human-to-human conversation.
  • Include welcome back intents, so the conversational competence of the IPA/chatbot would include the capacity of applying conversational opening practices.
  • Provide a variety of agreement tokens (ok, perfect, I think so, got that) in the machine’s answers. Their presence is needed for creating more multi-turn sequences with a higher level of internal cohesion between intents.
Your agreement tokens tight your intents together.
  • Use repair sequences for re-establishing the intersubjectivity/logical exchange between the human and the machine after a hearing, speech or comprehension error. Repair solutions are also more efficient for recovering errors when extracted from models of human-to-human conversations.
  • Design multi-modal experiences conducted by voice where the transitions between verbal and non-verbal channels (haptic/visual) are implemented in the same synchronized and multidimensional way as they happen in human-to-human communication.
  • Use linguistic models where politeness and sociocultural dimensions of the conversation are included: acknowledgement tokens as please and thanks/thank you are expected to be used both on the human and the machine sides. This decision maps the formal tone that transactional exchanges among humans normally have, where we tend to interact informally but, at the same time, keeping some social distance.

2. Think about the Problem that Needs to Be Solved

What is the final goal of this conversation?

Focus into the final goal that needs to be achieved with the dialog. Are you, for example, designing a multi-modal experience for purchasing clothing items in an online store? Then provide your IPA/chatbot with the psychological traits of a seller and transfer to the human-to-machine conversation the typical interactive practices belonging to a transactional exchange in the physical store and in the GUI version of the service. And insert all these practices in the new interactive context provided by the voice- or text-based conversational exchange.

In the context of multi-modal purchases guided by voice, you may:

  • Provide detailed item descriptions to the customer.
  • Ask the customer his/her opinion on the item in order to provide other options in case of a negative response or to push forward the purchase in case of a positive one.
  • Offer complementary options in order to get a larger sale.
  • Push forward the materialization of the purchase with diverse marketing techniques (limiting the number of available items, etc.).
  • Include (or not) confirmation intents at the time of pursuing the payment.

3. Define a Persona for the IPA

Everybody wears a mask, your VUI/chatbot too.

In Conversational UX Design, creating an adequate persona for the voice service represents an essential decision, so users could socially interact with it in a more effective way.

When humans interact with IPAs and other computer systems, they interpret their agency as a social action. This human psychological elaboration of the machine social capabilities gets even more emphasized in human cognition when conversationally interacting with an IPA/chatbot.

This means that you, as the designer, must pay even more attention to the definition of a proper persona for the VUI/chat interface and its later implementation in the dialog.

The design solutions you can follow related to this point are:

  • Implement sentiment analysis in the IPA/chatbot’s intents according to the persona you have created for it.
  • Create a personality also coherent with the service the IPA/chatbot is providing.
  • Use language for personalization of the service and fostering user’s engagement with the voice-driven (multi-modal) experience.

4. Provide the IPA with Conversational Memory

Creating both a long and short-term conversational memory for the IPA/chatbot constitutes another fundamental design decision for fostering user’s satisfaction.

In order to be fully conversational, the IPA must be able to remember previous conversations with the user in the same context (the application) and what it has been previously shared in the ongoing exchange.

5. Contextualize Universals in a Natural Fashion

“Alexa, help”

Universals are defined in Conversational UX Design as actions that should be present at every state (main menu, help, goodbye…).

IPAs’ conversational interfaces and chatbots do not keep these concepts in a strict sense, but it is important to ensure that the user can get help when needed and in a natural conversational fashion, especially when exploring such an innovative way of communicating with the machine as multi-modality represents.

In this sense, when the user encounters interactive multi-modal contexts not explored before (first-time user experiences), you may choose to follow an approach where the machine locally recommends the most efficient way (device/channel) of communicating with her.

To Summarize

As a Conversational UX Designer, your main philosophy must be to implement a customer-centered voice solution. That means to understand (1) how the human user is cognitively wired to experience the voice interaction with the machine, (2) which conversational structures does apply in his/her daily conversations, (3) how interprets the social agency of the machine, (4) what expects the machine to remember and to know about both current and previous interactions, and (5) what does he/she need to learn in order to navigate him/herself efficiently through a multi-modal experience.

Especially about this last point, you will find yourself in the situation of having to develop interactive solutions for new models of human-computer interaction that no designer has faced before.

So, one more time, designing conversational experiences, you will find yourself in the “new worlds, where no man has gone before”. And that is just a lot of fun :).


More From Medium

More from Dr. Carmen Martínez

More from Dr. Carmen Martínez

Conversational UX: Repair Sequences

Related reads

Also tagged Chatbots

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade