KH: Tell me about your journey to conversation design.
EW: Becoming a designer was never part of the plan. When I got into university my goal was to study microbiology with a focus on the brewery industry. I loved the whole concept of making alcohol with bacteria and fungi.
While I was studying, I got a job as a music scheduler at a radio station. My job was to “clean” music of vulgar words before they aired. I got tired of how repetitive the work was, so I taught myself Python. Eventually, a software engineer friend asked me if I’d considered pursuing product design. I started looking into it and discovered how multi-faceted design could be (and much more fun than cleaning songs).
Siri is what got me initially interested in CxD. I was interacting with her and began questioning how she works and who’s behind her. She also frustrated me at times—many chat assistants' responses are off-putting or frustratingly off-topic. When I began to research Siri, I got caught up in the web of conversation design—building realistic, useful assistants that could mimic real human conversation. That’s when I found the Daily Conversation Design Challenge with Hillary Black and decided to start on my own project. The rest is history.
Why did you create a mental health conversation assistant?
As humans, we don't like feeling like burdens to our loved ones. We don’t want to bring too many problems to them. Even our friends might get tired of hearing about the same issues over and over. But not a chatbot.
One of the limitations of this technology is that it's not a diagnostic tool. It's not going to cure my mental issues. But if I need someone to talk to who will never tire of our conversations or make me feel like a bother—that’s what Ava can do. The goal of using chat assistants for people experiencing mental health issues is to help users understand and manage their mental state.
What was your process for building Ava?
In Nigeria we say ‘there's smoke coming out of my head’ to mean we’re reading incessantly. During that project, I was doing tons of research on chatbots and AI—going down one rabbit hole to the other.
Because of my experience in product design, I did a competitive analysis of chat assistants and the role that conversation design plays in the mental health industry. I discovered that the main issue users have with mental health chatbots is that they lose the focus of a conversation. I would tell a bot what I'm going through—anxious feelings, for instance—and it would pick up on a phrase and start talking about something seemingly related, but not particularly helpful.
I wanted Ava to be different. So I went to work on my sample flow and conversation scripts. I also created a personality for my bot. I didn't want it to be genderless, I wanted her to feel relatable.
Because I was teaching myself everything, I had issues understanding how to build proper intents and account for utterances. But I watched Voiceflow tutorials and read a lot of Medium articles to get a better grasp on CxD concepts. After all my research, creating the actual prototype only took me a day. I'm very quick when it comes to learning new things.
How could Ava support those experiencing mental health issues? How do users react to her?
Ava focuses on what you’re going through at that moment and gives you counsel. She will “check in” on users in three scenarios:
- Ava drops a “fun fact”. Her prompt of a funny video or image can lead to a wider conversation about the user’s current mental state.
- Ava prompts you by asking about your day, to which the user will respond with a reflection.
- Ava checks in on you to continue a previous conversation you both had. She connects a user’s current mental state to previous utterances. This scenario is part of Ava’s development roadmap.
The best part of this project has been hearing about how people feel when they talk to Ava. My friends, colleagues, and family have all loved using it. One friend couldn’t believe it was a prototype and he wanted to have more conversations with her. Another friend said that Ava was already better than the current mental health chatbot he downloaded from the app store. Even Hillary Black said it's a very valuable product. So, the positive responses have led me to begin building an MVP that will cater to Nigerians—for now.
[An example of a conversation between a user and Ava:
- User: Hi Ava
- Ava: Hi Blessing, how are you doing today?
- User: I feel so bad. I'm having issues with work.
- Ava: Okay, I'm sorry about that. What’s bothering you?
- User: I’m feeling burnout at work, my boss doesn't listen to my ideas, and it’s really stressful
- Ava: That sounds really difficult. Would you like some resources that might help you manage work stress?
- User: Yes
- [Ava shares mental health resources relevant to work stress]
- [The following day Ava checks in] Hi, Blessing, I hope you're feeling better about work stress. How has today been?]
How do you think conversation design will impact healthcare tech and vice versa?
Mental health is a very complex issue. It should be treated in several ways including medications and therapy. Conversation design can be an effective tool for healthcare professionals to offer their patients as they manage their mental health issues. Virtual assistants can be programmed to provide support, resources, and information. They will never be a complete solution, but they can offer support where our overburdened healthcare systems may not be able to in our everyday lives.
It’s also true that assistants struggle to detect when a user needs to talk to a real human. That’s why on my app there will be an SOS button to get in touch with a therapist or crisis hotline.
On the other hand, healthcare tech has the incredible potential to contribute to large language models (LLM) and personalization. With a language model, conversations can be used to generate empathetic responses. These responses help users to feel heard and understood. Using the anonymized mental health data that Ava collects, we could help these LLMs. Eventually, these conversations could be fed back to the model to make future health assistants better and better over time.
Do you have advice for emerging conversation designers?
Keep learning. I'm only two years into product design. Now imagine adding conversation design to it! I still feel there's a lot to learn.
My second piece of advice is to bring your questions to industry leaders and mentors. I'm already in two communities, the Daily Conversation Design Challenge and Women in Voice. You'll find there are people who are willing to help. Nobody's withholding answers, you just have to ask questions.
How did Voiceflow help you with building this project?
I started with a proposal—what Ava does, the project’s limitations, and goals. But later I realized it wasn't enough to capture the vision. There was a lot of functionality that I was missing without a working prototype. I started scribbling, writing, making sketches, and then went back to researching AI. That’s when I discovered Voiceflow and decided to build my prototype there.
I couldn’t have completed the prototype without the Voiceflow platform. Because of that prototype I can share Ava with colleagues, receive instant feedback, and continue to iterate her conversational abilities until I build the app.
Ebere and Ava—The future of mental health support
For Ebere, there’s been no shortage of applications for conversation design in the healthcare space. As she continues to extend the flows of her Ava prototype, Ebere already has her sights set on her next project—an assistant that will support people fighting malaria.
In the meantime, Ebere keeps endlessly occupied with freelancing, running her NGO supporting hospital patients, and writing poetry. It all goes back to her learning mindset, “Having varied interests makes you a better designer and a better person. You can see connections in the world that other people can’t. In a world where everyone wants me to specialize, I refuse.”
Connect with Ebere Wilson on LinkedIn.