Conversational AI Virtual Agents have become a hot topic because of their ability to save businesses money while presenting a much better user experience when compared to traditional IVR systems. For that to be possible, though, it is important for Virtual Agents to provide a personalized experience that helps the user and feels similar to a real conversation.
Solution Methodology Research Lead, Carol Righi and Digital Experience Designer, Jeremy Chu, both of 1904labs, recently hosted a webinar alongside STLX entitled Conversational UX Design to discuss the important things to keep in mind when designing conversations for Virtual Agents. Here are some of the top takeaways:
Conversational UX Design has come a long way but it is still evolving
Conversational UX Design is an emerging space and evolving as we work in the field. While the first bots emerged in the 1960s, we’re just now scratching the surface of what is possible with Virtual Agents thanks to the use of AI. Today’s Virtual Agents are more human, understand interactions better, and provide a superior user experience.
We have all seen the likes of Siri, Alexa, and Google Assistant work in smaller settings, and technology such as Amazon Connect is making these Virtual Agents available to businesses now.
Voice agents provide a better experience – when done right
While a lot of people like to interact with chatbots over the internet, voice agents provide a better user experience than graphical user interfaces (GUIs) when done correctly. While yes, people have experience interacting with GUIs now, they have much more experience with having conversations. They are also more accessible for users with disabilities, such as visual or motor impairment.
The voice agents need to feel natural and like a normal conversation to the user to be effective. But, if they are, they are easy to interact with for most people.
Human conversations include rules and assumptions
Conversational UX Design always starts with human conversation. We don’t think about it when we talk because it comes naturally to us, but there are rules and assumptions in human conversations. It’s vital to understand these because we want to train the technology to have conversations, not let the technology change our patterns. In particular, there are two important rules that need to be understood: the cooperative principle and turn taking.
The cooperative principle is the idea that, in order to be helpful in conversations, we subconsciously agree to be informative, truthful, relevant, and clear. Meanwhile, turn taking is the mechanism in which we have conversations and use the content that the cooperative principle represents. Understanding how people manage turn taking to naturally resolve ambiguity is important to designing a Virtual Agent that feels natural.
Conversations should be designed with six elements in mind
When building Virtual Agents, you want to think about some elements that both take place in normal human conversations and that help the voice bot collect information correctly:
- Design to give feedback by including conversational markers such as timelines, acknowledgements, and positive feedback that make the conversation feel more natural and give the user prompts to show where they are in the interaction.
- Design to help users feel understood by including explicit confirmations (asking a question that requires a response) and implicit confirmations (lets the user know what was understood without requiring a response).
- Design to handle errors because things can go wrong. The most common type of errors include no speech detected, speech understood with no match, the system mishandling what it has heard, or the system misunderstanding what was said. Confirmations can be used to counteract these issues.
- Design to provide context by using information from earlier in the conversation as part of the interaction to create a more natural feel. This includes programming the agent to remember the subject and providing options explicitly where possible.
- Design to combat ambiguity. Human conversations will always have ambiguity that prior knowledge can resolve – the information the bot already has should provide context for it to overcome ambiguity where possible, and confirmations can help overcome confusion.
- Design to be empathetic. While bots are becoming less robotic, including personalization where possible and understanding things such as the motivation for the user, their mood, and how long they would be happy to wait on the phone, amongst other things, is important for creating a better experience.
If you would like to see the full webinar, including these highlights and more, you can either follow this link or watch the video below.