Problem Space

How can AI help students with learning disabilities (LD) improve their reading skills?

Students identified below reading level are currently pulled out of the classroom to get extra reading practice with parent volunteers, increasing barriers for students to develop socially and academically with their peers.

Design Solution

Introducing “Read With Me”: an AI reading companion for K-5 learners aimed to help students improve their reading skills early on, while also helping them spend more time in the classroom, ultimately improving their grades, sense of belonging, and social skills.

Research Process

Secondary Research

To gain a comprehensive understanding of the learning challenges faced by children with learning disabilities, the current teaching methodologies employed for our target audience, and the evolving landscape of artificial intelligence, our initial step involved conducting extensive secondary research. This research allowed us to establish a solid foundation in our problem domain, enabling us to effectively prepare for our upcoming interviews and focus group sessions.

Semi-Structured Interviews with Parents

We then conducted 5 semi-structured interviews with parents of children dealing with ADHD and/or Autism. These interviews empowered us to identify pain points via first-hand experience and stories.

Focus Group with School Teachers

Finally, we conducted a focus group session with school teachers. We were fortunate to have 5 passionate teachers with a combined teaching experience of over 140 years. They generously shared their experience and expertise in special education, providing insights into their daily interactions with students with LD. Additionally, through a speculative design activity, we collaborated with these teachers to envision the ideal teaching experience, weaving their perspectives into our insights.

1. Students with ADHD/Autism Face a Lack of Teacher Accommodations

Students with ADHD/Autism face significant challenges in school regarding teachers and accommodations due to the lack of acceptance and understanding surrounding these diagnoses. This makes the school experience even more challenging for students with LD who prefer more structure and boundaries but cannot get the accommodations they need.

“My kid has a dual ADHD/Autism diagnosis. He has a 504 plan but most teachers don’t follow it on a day-to-day basis. They just have other students to attend to.” - Parent #1, from our interview sessions

“Sometimes you will have teachers who say, “I taught it but you didn’t learn it.” Or sometimes you even have teachers who don’t believe in ADHD!” - Parent #2, from our interview sessions

 

“I wonder if there is something we can do to change the culture.” - Teacher #1, from our focus group, on resistance from teachers to provide accommodations for children who are on a special education plan

2. Loss of Self in Students

The longer a student with LD stays in an individualized pull-out program, the more they lose their sense of self and belonging in the classroom (source). Students with LD who stayed in their classroom had better outcomes in grades and scores in language arts than students with LD who were pulled out from the classroom for individualized learning (source).

“Our educational system tends to want to isolate them, often they are pulled out too many times. But they want to be part of the whole, they do. You can even say it’s segregated, definitely. And what you do to those individuals, they will hold and carry.” - Teacher #2, from our focus group

“School isn’t made for her. When you don’t meet the norm, it’s a very difficult system to be in.” - Parent #2, from our interview sessions

3. Variation of Learning Styles

Students with ADHD/Autism exhibit a wide range of learning styles and individual needs. Some may thrive with precise, step-by-step instructions, while others find visual aids and a distraction-free setting more conducive to their learning. 

Designing for students with ADHD/Autism requires consideration of their unique learning and perception styles, as well as the high degree of variation between individuals and contexts. It is important to leverage their strengths or "superpowers" to address their challenges.

“Some might walk in (to the classroom) and say it’s chaos in there. It’s not. It’s kids learning there. They just need to find their own ways.” - Teacher #3, from our focus group

4. Potential of AI

Teachers currently struggle to meet the needs of both general and special education students in large classrooms. AI technology can alleviate this burden by providing timely feedback tailored to individual students and breaking the spatial-temporal barrier.

AI technology also presents a promising avenue to address diverse learning styles and preferences, such as offering automatic generation of step-by-step guidance, adapting contents to their learning levels, conversion of text into visual representations, and seamless integration of text-to-speech functionalities.

Design Principles

Based on our research insights, we defined the following design principles and requirements as our Northstar for the design solution.

1. Intelligent and conversational
   
2. Simple and intuitive
   
3. Reliable and trustworthy
   
4. Realistic and optimistic
   
5. Ethical and safe
   
6. Adaptable and personal
   
7. Empowering and supportive

Design Process

Scoping

Our design process started from scoping down our problem space to focus on helping students with their reading skills, rather than going into the systematic issues of the educational system and lack of acceptance and awareness of learning disabilities. This allowed us to come up with a practical and actionable solution in a limited time frame.

Idea Generation & Prioritization

We then generated ideas with the Crazy 8s method and the Empathy Map and prioritized them based on selected metrics established according to our design principles. These ideas were then divided into the current spectrum and the future spectrum, based on technical viability.

The current spectrum focuses on improving established reading methods. This approach is particularly pertinent because students with learning disabilities frequently find themselves withdrawn from regular classrooms for supplementary reading sessions, which can disrupt their overall class experience and sense of inclusion. In light of these considerations, our primary goal was to craft a laptop interface, given the widespread availability and usage of laptops in the current U.S. education system.

The future spectrum envisions a more seamless and contextualized reading experience using novel interfaces, such as AR glasses and AI assistants. Given that ChatGPT was launched in December 2022, this project, initiated in early 2023, recognized a valuable opportunity to integrate large language models (LLMs) into the realm of educational technology.

Design & Prototype

Finally, we built a working prototype of our design solution using Figma, Midjourney, Unity, and inworld.ai, a service that integrates LLMs and voice recognition. The power and limitations of AI can be truly exemplified through a working prototype. Through this tangible representation, we delved deeper into understanding how an AI teacher would engage with various inquiries, explored people's perspectives on interacting with such an AI-driven educator, and assessed the boundaries of AI's capabilities.

While we did not have enough time to build a complete prototype for the future spectrum, we put our AI teacher onto the Nreal AR glasses and gave users a glimpse of what it could be like to interact with an AI teacher in AR.

Introducing “Read With Me”

Our design solution focuses on the current spectrum. Using “Read With Me”, students can practice reading with an AI teacher at their convenience, anytime and anywhere.  This approach provides them with timely and personalized feedback, all while ensuring they remain an integral part of their regular classroom experience.

Our solution also aims to communicate students’ progress with school teachers and empower them to help students stay on track with their learning goals (e.g. special education plan IEPs and/or 504s). This is especially important in a mixed-ability classroom.


High-Level User Flow



Key User Flow

Read With Me leverages the integration of AI to personalize students’ learning experiences by processing student input, generating questions based on measured abilities, and producing progress reports for teachers. This enables teachers to instruct and supplement education in a mixed-ability classroom while providing personalized tracking, feedback, and resources to students.

Our solution also enables teachers to better accommodate and monitor student progress in a mixed-ability classroom through the detailed dashboard. Meanwhile, students are empowered to practice reading on their own time, build confidence, and enhance their reading skills.




Key Features For Teachers

1. Generate and review reading materials of various levels



2. View student status at a glance and track student progress in a mixed-ability classroom



3. Customize the AI teacher, set restrictions, and monitor how students interact with AI


Key Features For Students

1. Read aloud to the AI teacher and get timely feedback


2. Accommodations (e.g. Focus Mode) to minimize distractions


3. Practice reading comprehension with AI-generated questions based on their learning levels


4. Gamified system to encourage good reading habit


5. Encourage exploration of genres and subjects in reading that align with student’s interests.

Reflections

Explore AI With A Working Prototype & AI in Its Nascent Stage

Although our proposed solution incorporates AI, we acknowledge that this technology is still in its nascent stage of development. While it holds immense potential for the future, it is crucial to recognize the potential risks and negative consequences that may arise, particularly in the absence of adequate regulations and widespread awareness. Consequently, compared to only using static mock-ups, our working prototype serves not only to vividly illustrate our idea but also to empower users with a better understanding of AI, fostering meaningful discussions around this transformative technology.

Incorporate Design To Raise Awareness Of AI

There are risks of using AI with young children as they may be unable to distinguish AI from real humans. This can result in children trusting what the AI says even when the algorithm produces incorrect information. A potential change we could make here, aside from expecting AI technology to improve, is to have the AI avatar be non-human to make it easier for children to distinguish the AI from a real human.

To bring this design to the real world also requires more awareness and education on artificial intelligence. The design should incorporate more education on this, such as including a video or a Q&A section on how the AI works, what it can or cannot do, etc. This will allow teachers and parents to have a better understanding of the technology that their children are interacting with and help address their concerns.

High Variability And Complexities Of Learning Disabilities

Learning disabilities are extremely complex and their symptoms and severity can vary widely between individuals. They are not one condition, but many, and different contexts can also lead to within-child variability. While AI may be able to adapt to different scenarios and contexts, how well can the technology adapt remains a question, and it is even more challenging when the scenarios involve vulnerable users and reactions/behaviors that may not be well-defined for a computer to understand.

Future Work

1. Conduct evaluations on our prototype through usability testing and the established AI guidelines and iterate on our designs
   
2. Evaluate and incorporate more accessibility features
   
3. Involve parents in the process, empower them to also be able to monitor their children’s learning progress as well as establish timely communication with teachers
4. Incorporate features that aim to improve the awareness and understanding of the AI technology
   
5. Develop the future paradigm more to explore emerging interfaces and their potential, as students with LD may benefit from these novel interfaces
   
6. How can we utilize AI further? Potentially, we can use AI perception technology to observe and predict the emotions and attention levels of students and provide real-time support accordingly. We can also utilize AI’s multi-modal capabilities, for example, converting texts to images or graphs that are easier for students to digest.