Dytective

Detecting the Risk of Dyslexia with an Online Game

an image of a computer and a mouse

The State of the World

About 10% of the population have dyslexia—an invisible disability. Dyslexia is responsible for kids dropping out of school, and what is worse, is that it usually goes undiagnosed, leaving the children with low self-esteem. Medical tests to detect dyslexia are expensive (~$1000), they are boring and intimidating—especially for children.

I drove from Baltimore to Pittsburgh in June of 2015. On a Monday in the labs of the Human Computer Interaction Institute (HCII), me, Dr. Luz Rello, and Dr. Jeffrey Bigham came up with the idea of working backwards from some of Dr. Rello’s research on the type of spelling mistakes Dyslexic people typically make. Using a large data set Luz had collected, we set out to train a machine learning algorithm to detect patterns of spelling and letter recognition mistakes consistent with those made with people who had an official diagnoses of Dyslexia.


The Team

Dr. Abdullah X. Ali

Artifact Creation

Concept design, interaction design, UI design, naming and branding

Programming

Front end-development of a web-based video game using HTML/CSS and Javascript. Back-end development: Creating a PHP-based system to store data in a MySQL database and run an ML model that would communicate the results back to the front-end site.

Knowledge creation

co-authored multiple publications.

Linguistics and Communication

Demo’d Dytective at multiple scientific conferences and competitions.


Dr. Luz Rello

Human Behavior: Knowledge Creation, Data Analysis and Statistics, Linguistics and Communication.

Dr. Miguel Balesteros

Programming: Machine learning expert.

Dr. Jeffrey Bigham

Human Behavior: Project supervisor.


What I did

I drove from Baltimore to Pittsburgh in June of 2015. On a Monday in the labs of the Human Computer Interaction Institute (HCII), me, Dr. Luz Rello, and Dr. Jeffrey Bigham came up with the idea of working backwards from some of Dr. Rello’s research on the type of spelling mistakes Dyslexic people typically make. Using a large data set Luz had collected, we set out to train a machine learning algorithm to detect patterns of spelling and letter recognition mistakes consistent with those made with people who had an official diagnoses of Dyslexia.

The Research

This was a case where used a prototype as a design probe to test a hypothesis. Given that we were not working from the absolute beginning, rather working from a position where Dr. Rello had done her entire Ph.D. in HCI and Linguistics on Dyslexia and intelligent detection tools.


Iterative Design and Testing

Because our target audience were children who may or may not have Dyslexia, we wanted to create a screener that is engaging. On that same Monday, we came up with Dytective, a web-based game that has timed levels challenging children to do a number of tasks like find all the ‘E’ letters in a sea of ‘F’ letters in 30 seconds, or cut up a long string of letter without spaces into words in a sentence. By Wednesday we had a working prototype, tested with 40 users, and submitted for publication at the ASSETS conference. Our prototype had 81% accuracy in detecting whether the user playing the game performed in a manner consistent with that of a person with Dyslexia or not. The interface itself received overwhelmingly positive feedback. Parents were telling us that their kids were upset they couldn’t play the game again after their code expired—we limited each kid to a single game session to ensure accuracy in the data for our ML model.


Following the Dytective pilot’s success I spent the rest of the summer expanding coming up with new levels, and improving the Dytective game.

I created the bubble’s background and used a soft and light color palette to strike a balance between a minimalist design and a kid-friendly game. Most video games, especially ones geared towards children have screens that are over-crowded with elements with highly saturated colors, which tend to be distracting. The design decision I went with here was create a calm bright environment and let the task at hand be the focus of the User Interface. This was a test of performance, removing any distracting elements was a key design consideration.




What I delivered

A virtual Experience, the Dytective game. I am officially a co-inventor of the IP of Dytective, which is now a part of Dr. Rello’s Change Dyslexia, a non-profit org working to reduce of dropout rates due to dyslexia.



We entered Dytective into the UAE AI and Robotics for good Competition in 2016 and made it all the way to the semi-finals placing 7th out of 664 international submissions.