Autism Help was inspired by a study from researchers at Newcastle University which showed that children with autism were better able to handle their fears and phobias by using virtual reality. Daphne Mallory, the Creator of Autism Help, wanted to explore the use of voice interface technology to make an impact. Creating an Alexa Skill for low functioning children with autism, including non-verbal children, was a no-brainer. Thanks to Pavel Anni, the Application Developer whom she met during a webinar they both attended announcing the Alexa Skills Challenge: Kids, Autism Help was born.
What It Does
Autism Help takes the guess work and hassle out of compiling, breaking down, and teaching skills (functional, verbal and social) to children with autism. It helps children acquire skills in a convenient and manageable way, and it's accessible to use daily. Each skill has steps, and each step is separated to allow time for children to learn and practice. During that time, Alexa plays music or is silent, depending on the skill. Daphne wrote a task analysis for each of the 50 skills, breaking down the main steps based on research, anecdotes, and her own experience. Daphne also created the sequence for when each skill is learned.
It was important to us that Alexa interacts primarily with the caregiver and not the child learning the skills. This is so a child does not become frustrated when she does not clearly answer Alexa's questions (which is needed to move along a task chain.) This also makes Autism Help inclusive, because it accommodates non-verbal children.
One of the main value propositions of Autism Help is that it’s based on real world experience. Daphne has first-hand knowledge of each skill created for Autism Help (and we plan to add more in the future with the help of users). As a mom of children on two opposite ends of the autism spectrum, it was important to create task chains that were practical and realistic. For example, the amount of time allocated to complete each step, and the incorporation of specific music and applause effects, are strategic. These are based on personal experience with her children and other children she interacts with. For example, it may take the entire 3-minute song to clip one fingernail due to sensory issues and meltdowns. A child may have to work on putting the toothpaste on the toothbrush for a week before proceeding to the next step.
The design and portability of Echo devices is critical to Autism Help, because parents and caregivers will be motivated to use it. Each Echo device is easy to transport and use around the home and at other locations. The use of the echo device also reduces some of the social challenges associated with acquiring skills from adults. One-on-one interaction with adults in learning situations can cause stress for children with autism, due to cognitive, perceptual and sensory issues. There's now a third party that's less threatening: Alexa! For some children, it will be easier to talk and listen to Alexa than a human voice. That's the power of enabling and using Autism Help daily.
How We Built It
The skill is written in Python using a framework called Flask-Ask. We used DynamoDB to store the skills and each user's progress. For each Alexa account (one per household) multiple users (children) can learn skills. Autism Help keeps track of their progress for each type (verbal, social, functional). We used Zappa to package the whole application, upload it and use it as an Amazon Lambda function. Finally, we used Amazon S3 to store static content (media files).
Challenges We Faced
The biggest challenge was to make sure that the Voice User Interface (VUI) we created is easy to use. We had to re-design the whole dialogue flow a couple times, trying and testing what worked. It's very different from the Graphical User Interface (GUIs), that require us to think and program differently. There is still room for improvement and we'll be working on making the dialogue smoother and easier to use as we collect feedback from users.
Accomplishments We're Proud Of
We're proud to create an Alexa Help that will have an impact on the lives of children with autism, who struggle daily. Many families suffer alone and in isolation, and the smallest improvement in speech or actions can greatly improve their quality of life. We're also proud of our ability to collaborate remotely. Despite living miles apart, and never meeting in person, we managed to create a very productive team: A Content Creator (Daphne Mallory) and an Application Developer (Pavel Anni). The key that has been a mutual respect, understanding and a shared vision.
What We Learned
The biggest lesson learned which we take going forward is how to find the balance between the Minimal Viable Product and the Minimal Marketable Product. As a Content Creator, Daphne wanted a Skill that parents would want to use daily, and as a result wanted features that can be developed down the road, but not right now. As an Application Developer, Pavel wanted to ensure that Autism Help worked as intended and was delivered on time with the core features present. Finding the right balance that produced a win-win outcome was crucial, and has increased our emotional intelligence around developing Alexa Skills.
We also learned how to develop Alexa Skills using Flask-Ask and how to deploy, test and debug them using Zappa and Amazon Lambda diagnostics (logs, etc.). We learned how to store the information from each session, both inside and outside the application, using DynamoDB. We also learned how to prepare and upload text (YAML) files to a DynamoDB database.
What's Next for Autism Help
We are going to create a companion web application to keep track of each child's progress, and custom learning programs for key skills. We also plan to offer a rewards program for achieving skill levels, a service web application to add more skills to the application. And, of course, we'll improve our VUI and dialogues based on users' feedback.