Ideation
Wireframes were created on a simple 1x2 grid as to not overstimulate younger kids. The application is created for a tablet since this technology is most accessible to our younger audience.
The home page consists of navigation prompts. We planned to include kid friendly illustrations here and throughout the app.
The camera page prompts the user to take a picture of the weather outside as well as AI assisted instructions.
The chatbot page would be the most used so we spent a lot of time ideating on how to accommodate the user. Since our users would be at a range of reading and writing skills, we decided to create a variety of communication styles. These included text, illustrated buttons, audio and microphone.
Finally, we created a parents page to account for ethics. The team choose to include a lock here to ensure that the child wasn't accessing anything that would be harmful to their learning experience, and that the parent would have total control and access over what occurs in this AI assisted experience.
The Sunny Sidekick Chatbot is designed to help users understand the weather, make connections and solve problems in order to be prepared for the day.
The team tested the chatbot a few times to ensure that the flow worked throughout a "normal" conversation and an "irregular" conversation. We also had a four other individuals converse with the chatbot to explore new conversations. This helped us understand where faults or challenges could happen throughout the flow.
The team made sure to complete the small talk section of Dialogflow to give the AI a well rounded personality. We stuck with an excited educator for the personality of the AI and went through multiple possible scenarios of how a kid may react while conversing with a chatbot.
Image Recognition
It is important to note that while the team did train the computer to classify images, this function was more of a step for the user to establish a connection between the weather and how to prepare accordingly.
In this dataset, there are 4 different types of weather images:
Testing accuracy increased from sessions 1-3. However, it decreased slightly in the following sessions. The top confusion was rain and clouds - the computer mixed the two up in more than one session. We presume that if we added more data, and therefore more sessions, the computer’s accuracy would increase. In other words, it would be able to differentiate rain and clouds (and all other weather types) more consistently.
Folders were created for testing and training and each of these folders contained a subfolder with weather type. The testing images are the same 10 images for weather type.
Prototypes
The team prototyped two scenarios. The first was an event where the user struggled to pronounce the “TH” sound. And the next involved problem solving. Below are video prototypes of each scenario:
Style Guide
The style of Sunny Sidekick is colorful and kid friendly. Furthermore, fun illustrations take up a large portion of the application.
Color Palette
Typography
Illustrated Digital Communication Board
Final Designs
Designs were finalized after carefully considering research insights.
Navigation included multiple visual cues (large icons and clear labels) that served as signifiers.
A parents/guardian page with several details regarding the child's experience.
