The goal of this next assignment was to understand the differences between prototyping for voice versus screen-driven interactions. This was an opportunity to practice prototyping voice interactions and avoid key pitfalls.
Presenting Movie Finder, a voice user interface (VUI) TV add-on for any movie watcher. Movie Finder allows users to ask any questions regarding a movie, genre, or where to watch the movie.
The ideal user group for Movie Finder includes movie watchers at any level of frequency (daily, weekly, monthly). The idea of the Movie Finder is to be an add-on speaker for existing TVs similar to a Google Chromecast. Movie Finder will be connected to the TV with a VUI where the user’s voice can also be detected.
The purpose and function of Movie Finder is to help find movie information faster, minimizing the time spent trying to find a movie and maximizing the time watching movies with friends & family.
According to a recent Netflix study, Netflix users spend an average of 17.8 minutes per day searching for something to watch on its platform (source). The overall goal to minimize movie search time makes Movie Finder desirable to users looking to keep movie searches simple & easy and feasible in that the user will only need to speak to the VUI.
Users can ask Movie Finder any questions relating to movies. Key tasks include:
- Find the highest rating for [genre] movies.
- Who stars in [movie]?
- What is [movie] about?
- Where can I watch [movie]?
- How long is [movie]?
These dialogue prompts were created based on commonly asked movie questions quickly surveyed from friends and family. Below is a dialog flow chart that depicts multiple types of interactions along with potential error handling states.
The goal of the flow chart is to demonstrate the different types of interaction with Movie Finder, along with potential interconnected error handling states. Some interactions may overlap with one another, but could not be depicted visually aesthetic.
Most dialogs are focused on narrowing down the user’s question by asking for more details, offering new suggestions, and asking if they would like more options. The error handling states include reclarifying the question and narrowing down a search with a confirmation question.
The overall tasks above demonstrate potential dialogs that help narrow the movie search while catering to the user’s preferences.
Here are my initial results after conducting testing and receiving critique in class.
Things that worked well:
- The VUI was straightforward and simple to use/understand.
- The suggestions and error handling by prompting different movie genres worked well to narrow down movie choices.
- Tester personally mentioned they would want to buy this for themselves.
Things that needed improvement:
- Some sample dialogs were very specific and might not be a typical question that would be asked from users. It was suggested to make the dialogs generic first and more specific if prompted by the user itself.
- Adding more potential error handling where the VUI repeats what the user said as confirmation.
Overall the effectiveness of the design was indicated through testing and class critique sessions above. The concept was desirable and feasible based on both feedback sessions.
After concluding my testing and gathering feedback, here are things I would change and things I learned.
Changes I would make:
- I would consider adding more conversational prompts such as “good morning” or ending with “have a good day.”
- I would also add more error handling to confirm what the user had just said. This is in the case when the VUI is unable to detect the user’s voice fully due to background noises or soft-spoken words.
Things I learned:
- VUI design is tough. Designing for all types of users can be difficult if some users like conversational dialogs versus straightforward dialogs (ex. question then answer).
- During the in-class critique session, I was surprised to see the variety of dialogs from my classmates. Some were very detailed and conversational whereas others were straightforward and short-ended conversations.