Why companies are investing in conversation designers
- If one company makes a chatbot, can everyone else just copy it with some tweaks?
- How frequently are companies using insights from interactional sociolinguistics, or conversational style to design personality for their bots and virtual agents?
Hundreds of questions were submitted by attendees of Botmock’s AMA about becoming a conversation designer. Five industry experts with backgrounds in conversation design answered as many as we could in a live roundtable discussion. To register for future (free!) AMA sessions and watch the full recording of this session, you can do so here.
The following experts who have designed conversation at Amazon Web Services, Facebook, and Nuance joined us to answer your questions:
Mary Parks, Voice Interaction Designer
Celine Park, Conversational Experience Manager
Rebecca Evanhoe, Conversation Designer
Brooke Hawkins, Conversation Designer
Brielle Nickoloff, Product Manager, Botmock (Have questions about conversation design? Feel free to connect with me! ⚡️ )
Q1: Why is this industry evolving? How is it not a one and done field?
If one company makes a chatbot, can everyone else just copy it with some tweaks?
Rebecca
At first it’s hard to understand how these interactions can’t all be developed from one template. It is easy to think that ordering a suitcase, or shoes, or food, all create very similar conversational structures. While there are patterns we can derive from human conversations in general, once you get into this field, you realize the impact of factors like the user’s context and their space, what they’re doing at the time, what the product is and what the content of the language is. All of these aspects of conversation create an incredible amount of variance in how people respond and what their expectations are for the interaction. For example, I have made four different chatbots for car insurance companies, and all of those chatbots had different business purposes and user profiles. So even when designing a chatbot that lets you check when your insurance premium is due, we had to create a very different experience to efficiently serve specific users, contacts, and products.
Mary
There are so many different ways to answer this. Chatbots and voice interfaces present a probabilistic input problem. We’re trying to get people to use their personal language to interact. We want to help them focus on the task and not worry about the language they’re using for their input. The diverse communities that we’re designing for should have a real impact on the design decisions we are making, decisions around not only how chatbots and voicebots behave, but also decisions around what they say and how they say it.
These systems are complex and multilayered; there are multiple factors to take into consideration. Though there are patterns and better practices that can be leveraged, we’re still learning a lot. Also language changes over time; these systems need constant updating because humans change. And even when people interact with one interface, their language will change over time. It’s the nature of language. Design and improvement are never done. I don’t think we can just copy what another interface does and use it somewhere else.
Branding also makes a big difference. Each brand needs a different approach. There’s a lot that goes into these systems.
Brooke
I think users’ and consumer’s fluency and familiarity with voice experiences are evolving constantly. I started my career voice designing IVR calls and I have been witness to all the innovations that have happened in recent years. It has been great to see how different Alexa, Google devices, and many other devices and applications that previously did not exist have launched and been adopted successfully by consumers.
I believe that the field of voice is not just evolving in response to the companies or users we are designing for, but it is also progressing in response to the changes happening in the world we live in. As the world is evolving the expectations out of voice design and implementation are also changing.
Celine
Coming from a retail company background I can confirm the fact that the needs of the business and the needs of users are constantly changing along with the use cases that have to be studied when designing a system. If you add to that all the other changing aspects of the field is not far-fetched to realize that what could have worked six months ago might not be applicable now.
Brielle
When I started as a conversation UX designer, my boss had been in this field for about 25 years and had worked a lot on IVR systems. So when I joined the team, he specifically mentioned that he was glad that I didn’t have a lot of experience. He went as far as saying that he was not even going to share a lot of the best practices that he had learned over the years, emphasizing that he did not want to contaminate me with the not-so-delightful “speak to click” model of IVR and other practices that don’t employ most of the design principles that we use nowadays. His reasoning was, let’s put to work your clean brain and new ideas, and use all these new resources to create something of our own. Hearing this from a true expert made me realize how true it is that at this time, we are all learning together.
Rebecca
In this field it is very important to stay open and understand that as the technology improves, constraints that we once had designed for will be gone. As technology evolves, we might find entirely new ways of organizing. For example, in the present we all work off of intents and utterances. In the future, there might be a different system that doesn’t use that at all.
Another thought I like to bring to the surface is directed to our peers who are new to this field. Sometimes people have the misconception that this technology is new and that we have never seen anything like it before.
In truth, this technology has been around since the ’50s. What is new is the way we are approaching its development.
At this moment the industry is being influenced by people that are experts in the social and behavioral sciences and humanities that are so necessary for driving the voice space forward. It is also very important to remember that there are people in our field that have 20 or 30 years of experience, and we need to continue to look to them for knowledge and direction.
Q2: How frequently are companies using insights from interactional sociolinguistics, or conversational style to design personality for their bots and virtual agents?
And how many resources does a company dedicate to defining that personality?
Mary
It’s important to come up with some really simple ways to promote the perception of the bot’s “personality” in the mind of end users. Language is one of the ways to identify a person, as well as a character. And one thing about identity is that it needs to be consistent. In design, we have to make a decision about the language base that the character has. And though it’s really important to have a language base, it should never be a “standard dialect” or “standard language.” We can be open to using any dialect in the design of these characters.
But dialect is not what defines the character. It’s the psychological component that defines the character.
For example, there could be a character that is supposed to have a personality trait that is “sophisticated.” The goal would be for the character to be perceived as sophisticated. Then we would definitely want to figure out how to have a lexicon that supports that notion, as well as a style of talking that supports that perception.
Sociolinguistics can help with all these things. But in terms of companies using insights from sociolinguistics, I don’t think they’re doing it enough. Moreover all of us tend to use our own dialects and the language of our own community of practice as the basis of the language that goes into bots. Sociolinguistics is not being used enough in the industry, but it’s something I’d like to advocate for.
Another interesting problem is the use of machine learning to create output, via natural language generation. If you’re using natural language generation, which tends to introduce more variation in the output, it becomes harder to control the voice of the character so that it has an identity that’s consistent. I think sociolinguistics could help here too.
Brielle
The topic of identity reminds me of a story that I heard from Deborah Harrison. Deborah is one of the original writers for Cortana (Microsoft’s Voice Assistant), and she mentioned that when designing Cortana they struggled early on trying to figure out what kind of vocabulary the bot should have. At first, they thought that since this entity was a sort of keeper of knowledge that had all the internet at her disposal, they should design a voice assistant that expressed random knowledge and used big words. Their initial thought was that this identity would make her seem like an all-knowing friendly entity. After further study, they decided that this type of personality could make her seem over-educated and even give the impression that she was talking down to its users. In the long run, they made a design decision to not add those personality traits.
Mary
I believe that something important to keep in mind when designing for conversation is figuring out what is colloquial. To do so, there is this trick I use in which I focus on only using Anglo Saxon roots and trying to avoid other types of roots like Greek or Latin. Keeping that in mind assists in giving a colloquial flavor to conversation.
Celine
It is still in the early stages for companies to know what users want in terms of the personality and identity of their bots and virtual agents. One huge thing that we did for our bot was to determine a persona before we even started to think about progressing further into other design stages.
To define the personality of our interface we got input from marketing and research and used that information to determine what our users were looking for. We found out that our users wanted a very knowledgeable bot. Knowing this opened the door to a conversation about ethics in voice and conversational design, and I wanted to bring this up because one of the biggest decisions we had to make was to use a male or a female voice. We knew that the personality that users wanted was someone knowledgeable, but how are we going to figure out which one is going to portray a better sense of that? Since females in the past have proven to show they are more of an assistant person, at first our assumption was to choose a female voice. But after doing research and taking into consideration factors like the nature of our company, the study pointed to a male voice. Going through this process with its ethical and social implications was very interesting. Differences in voices and personalities are certainly identity traits that users are conscious about.
Brooke
I believe that some of these interrogatives about bias, conversational style, and persona can be studied by looking at chat transcripts and listening to live interactions with the conversational interface. Doing this type of testing quickly surfaces if there is any bias at play as well as any stopping points that people are having with the conversational style that your bot was designed to use. The principles of user experience design are very important for determining the best decisions for these types of considerations. Putting that work in ahead of time rather than making misinformed decisions is the best way to make sure that we are covering the user’s needs and expectations
Rebecca
When it comes to persona I believe that companies are engaging with the idea of testing persona and thinking deeply about its impact on the users. All kinds of studies show that people perceive persona strongly and instinctively; so it’s a huge component of the user’s experience. Still, when defining the identity of a conversational interface a lot of companies think of persona as an extensive and overly developed backstory (something along this lines: our chatbot has a Ph.D. in Native American Studies, and grows orchids and lives in Sacramento) that doesn’t have anything to do with how the chatbot is going to connect with the user and serve the user. So when talking about persona, we all have to think about what are the benchmarks of the interaction; does it need to be efficient? Does it need to be nurturing? Does it need to support a level of intimacy because private information is being shared? What does the interaction need to accomplish? Once those questions are answered we can proceed into thinking about what kind of personality traits will support qualities like efficiency or something like creating a sense of trust.
So I think there are deeper ways to think about persona and the study of persona is going to continue to emerge as companies understand the importance and impact of this factor.
The Botmock team is excited that more of our community members are involved than ever before. There will be more panelists and attendees from diverse backgrounds contributing to these events in the future, so keep an eye out!
This article is an expansion of the first topic covered in Botmock’s first conversation design and development AMA session. You can also read read recaps of the other two sections, Measuring the Success of Conversational UIs and The Technical Elements of Conversation Design.
