VUI prototype: Q, a device for getting the news
Designing and testing a voice-user interface to encourage critical thought + a better-informed public
In the digital age, there are many ways to stay informed—YouTube videos, TikToks, podcasts, you name it. According to a 2018 NPR report, 70% of people who own smart speakers have reported using their devices to listen to the news. Although news is not one of the main features for voice assistants, 37% of users listen to at least two hours of news content a week. Smart speakers now have the capability to respond to political queries, as well.
Because of these features’ rising popularity, I wanted to create a VUI that specializes in news-getting and handles political queries. By creating sample dialogues and conversational flow charts, I prototyped “Q”: a VUI to encourage critical thought and a better-informed public.
Design
Q is for people who want to stay informed about news and politics, but don’t have the time to sit down and read articles on their own. Using the voice assistant, users would be able to integrate the news into daily tasks like cooking or cleaning. The VUI would work by searching the internet to respond to queries. Additionally, it allows users to personalize their experience by setting their preferred news sources and favorite podcasts. When reading articles, it makes sure to inform the user of the source and any potential bias. With Q, users could save time yet remain informed, critical thinkers.
Prototype
To begin designing Q, I created sample dialogues. Sample dialogues are like the “wireframes” for a VUI; conversational scripts between the VUI and user to test and get feedback before prototyping. When planning out these conversations, I chose to focus on five core tasks. For each task, I marked possible error states and confirmation states, as shown in the key above.
1. Asking for today’s top news
I figured this query would likely be the most popular feature for the average user of Q, so I spent a significant amount of time figuring out how to streamline the interaction. When flipping through a newspaper or scrolling on social media, people read headlines before dedicating themselves to reading the whole story. I replicated this experience by having Q read out three headlines at a time, then checking in with the user to see if they’d like to read a story or move on.
For the majority of my sample dialogues, I opted for implicit confirmation, which involves the system answering the user’s question by repeating key words from the user’s query in its response. For low stakes interactions, this method proves successful. Although it does not explicitly ask for confirmation, the system shows the user that it understood their question.
2. Setting preferred news sources
Because users who would purchase a news voice assistant likely consume news regularly, I decided to provide a way to personalize Q by setting their preferred news sources. When using this feature, the user can ask for information on news sources. When providing answers to these questions, Q references the internet and provides a bias rating from All Sides. Once a user sets their preferences, Q will give priority to the user’s preferred sources when completing news-related queries. This feature has a similar flow to the first task.
3. Asking for election information
I also created a sample dialogue for Q’s secondary purpose: political queries. During elections, the top searches on Google are for voting results. Instead of hitting refresh on their phone or computer, users can ask the voice assistant. When given an ambiguous query, the VUI will assume the user is asking about the most prominent election race. If this is in error, it will say “I’m sorry” then provide the correct answer to a user’s more specific query.
4. Setting favorite news podcasts
After doing background research on how people consume their news, I decided to also include sample dialogue for podcasts. When setting their favorites, the conversation flow is similar to setting preferred news sources.
5. Listening to a news podcast
Q will keep track of new episodes for the user’s favorite podcasts, and can notify them when they are released. The VUI will play the episode through its connection to the user’s Podcast app. This is a simple conversation flow.
Flow Diagrams
After I tested and refined my dialogues, I situated them into flow diagrams. I also marked decision points with diamond-shaped boxes, and termination commands with ovular boxes. For each dialogue flow, I identified points where the user might have alternative responses. For example, the user has opportunities throughout each flow to terminate the conversation early. Laying out the dialogues visually allowed me to imagine different paths more easily.
Analysis
Next, I did a “table read” with a partner to assess the conciseness and intuitiveness of my VUI. During testing, the conversation flowed naturally for the most part. However, at decision points the flow was broken. These dialogues were too wordy and seemed awkward. My partner mentioned that they might want to choose an option that wasn’t listed by the VUI. If I had time to rewrite my dialogues based on this feedback, I might give the user more freedom at these decision points. Instead of giving the user two options, I might leave it more open-ended and allow the user to make decisions organically. Despite these decision points, the rest of the dialogue made intuitive sense and provided useful information.
