Alexa Custom News
A Voice Application for Amazon Echo
TYPE: individual (personal interest)
DURATION: 1 week
TOOLS: SaySpring, paper + pen
MY ROLE: Researcher, information Architect
Build an Alexa application/skill that allows a user to customize their news feed
Most people that have an Echo use it for only a few things: weather, news, Jeopardy. One of the most common complaints I hear when talking with people about their Alexa usage is the inability to customize or contextualize most information that comes from the device. For example, if you ask about the temperature, Alexa will repeat the full forecast. That's not ideal.
One interesting complaint I heard was something I had not thought of: the selection of news stories for the Flash Briefing. Currently, only NPR and Stitcher can deliver news; this often means receiving a narrow spectrum of opinions and topics. This is partially because APIs for VUIs are still in their infancy, and partially because there's no good way to tell Alexa what sources you're looking for. A user must go into the Alexa app and change settings to accomplish this. My goal is to turn on/off news sources using voice only.
Research, Research, Research
While Voice User Interfaces are a part of UX, voice design is still a young, yet growing specialization. It was particularly challenging finding Voice UX processes to model this application from, so much of this skill is pieced together from pre-project research. I opted to use as much of my existing process where possible. So, this meant Pre-Research > Market Research > User Research > Ideation > Prototype, with some different techniques than I normally use.
DISCOVERY & RESEARCH
Understanding the Alexa User
And gaining context around news habits
This project was spurred by an off-hand comment about the Alexa News app (Flash Briefing), so the original problem was understood. However, I wanted to get a deeper understanding of the habits of the typical Alexa user.
I interviewed several Alexa users and began interviews with the following pre-selected questions:
- What do you typically use Alexa for?
- Do you read news on social media?
- Do you watch cable news?
- What are your top news sources?
- Is there anything you wish Alexa did that it doesn't already do?
I also observed subjects interacting with Alexa, both in the news app and outside of it.
And understanding human behavior
I expected to have much more content around the types of news sources available, what words might trigger Alexa to add or delete a news source, and gripes about what Alexa can or cannot do.
Four main motifs were found:
- Users sometimes do not use the same phrase to start an Alexa app. It is truly conversational.
- "Alexa needs to be way less transactional and more conversational"
- People hesitate to speak over or interrupt applications for fear they may not hear them and will have to ask questions or say commands all over again
- People hesitate to give natural commands/questions
Limited options vs. many options
One thing in particular that might be driving the comment "I wish I could customize my news feed by just talking to Alexa" is the fact that Alexa doesn't currently have many options to choose from.
The user experience may be vastly improved once a greater variety of news sources are able to be fed into the system.
The challenge here is this: how might we have a user customize their experience (as much as allowed by current technology) through voice alone?
No assistant allows a user to add or delete news sources via voice alone
And, Alexa is less conversational than Google, which may create more challenges as we build out the phrase map and app scripts.
using all the methods
Trying on techniques for size
As my first voice project, I have little to imitate in terms of process or research techniques. For this reason, I chose to employ as many techniques as I could in order to understand what would or wouldn't work for a certain design problem.
I utilized storyboarding "lite," text scenarios, flow maps, and phrase maps in my ideation phase.
Where, when and how?
Checking the news is generally a solo activity. Thus, the story is less conversational, and more contextual. Our character listens to the news in the morning, and at other times where he has been away from the internet or social media for several hours.
Rumblings of a Script
During storyboarding, it felt more natural to expand upon the speech bubbles and begin building out some natural-sounding "proto" text scenarios as I had the image in front of me.
This would become the basis for the plain-text scenarios in the next section
User FLOW MAPS
How many different ways can you ask to hear the news?
Turns out...quite a few. I mapped out several key phrases within the application and arrived, through user research, at various different ways one could ask to launch the app, hear the news, interrupt Alexa, and say yes and no.
Luckily, the SaySpring application accounts for minor differences in speech and can accurately decipher intent (within reason). So, in many cases, it was not necessary to add separate logic for a sentence that contained the word "can't" vs. "cannot." Alexa would interpret these as the same thing, and rightfully so.
Fixes and additional features
This is just the beginning, but, in the next iteration:
- More support for alternate phrases
- Visual (device) feedback for errors
- Increasing support for various edge cases
- Logic around handling errors