Dysolve: Using AI To Beat Dyslexia Through Generative Games

While large language models and predictive generative transformers have enabled AI to unleash a torrent of words on the world, Dr. Coral Hoh has been working to harness the power of AI to ensure the accessibility of those words. Dyslexia, a neurobiological disorder characterized by difficulties with accurate and fluent word recognition and by poor spelling and decoding abilities, strikes as many as one in five people. For linguists like Dr. Coral Hoh, dyslexia presents a frustrating paradox. It is the most commonly misdiagnosed learning disability, and through readily correctable, parents and teachers are often not given adequate resources to tackle the problem.

Seeing Dyslexia As A Computational Problem

Dysolve is the first AI computer program for dyslexia and language-related disorders that does not use preset testing kits. Instead, the AI-generated algorithm uses interactive games to assess and target each student’s unique dyslexia case.

Dysolve starts with the premise that dyslexia is fundamentally a coding problem—something about the brain’s information processing is not operating as it should, resulting in reading being a struggle.

For most literate people, the act of reading happens automatically, almost magically. A common internet message illustrates how readily reading occurs for most people, even when the letters of the words being read have been jumbled: “it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteer be at the rghit pclae. The rset can be a toatl mses and you can sitll raed it wouthit porbelm. Tihs is bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe.”

It is not only understandable but understandable with minimal effort.

Dyslexia represents an interruption of these automatic systems at a more fundamental level, and that is the level where Dysolve goes to work. Dr Hoh uses the metaphor of an operating system to characterize the visual language processing system. Dyslexia happens where there are errors in the code and an individual’s difficulty reading words results from these underlying errors.

With dyslexia as fundamentally a computational problem, it becomes natural to deploy a computational solution to address it. The challenge is the need to quickly process large amounts of information, the sheer volume of which makes it impossible for a human coach to observe, identify, correct, and then address problems in a constructive manner. This is especially true given that “code errors” are unique for each person.

In this context, the goal of the AI is not to generate words but to help shape the signal carrying information about the words, so that the intended result is what we get.

Adjusting the signal of a sensory process happens all the time in the case of vision and glasses. Yet anyone who has juggled prescriptions will know that the fix does not apply evenly everywhere and will change over time. Efforts to take a more computational approach to poor vision, such as compensating by adapting the image on a monitor by a monitor so that the eye without glasses would see a corrected image, have been explored but have yet to result in a consumer product. Not even the Apple Vision Pro has managed to solve this problem computationally.

In the case of Dysolve, the mission of the AI is to identify the nature of the altered signal by dynamically constructing games. Constructed new each session, these allow the program to determine the nature of the coding errors, and to then to train each student’s brain on how to work around them.

The challenge is to do all this while ensuring that the games retain enough “gameness” to remain attractive for the student, while simultaneously providing the input needed to diagnose and treat the problem. The result is a system where the games go fast–typically lasting 1 to 3 minutes–and where they are constantly changing.

Promising Results

Regarding impact, students who begin below the 25^th percentile typically progress to above the 50^th percentile within one year of using the program. Hundreds of students have been helped by Dysolve, with users catching up to their grade’s reading level in 1-2 years on average,

The best thing is that these gains persist. One does not need to keep using the program to preserve the gains. Rather, the result is a shift in how the user’s brain is processing language. To lean into the coding metaphor, it is like the errors in the code have been corrected.

The potential impact of Dysolve is significant. With the typical cost per student for dyslexia services reaching upwards of $20,000 per year in states like New York, the system-wide expense is enormous. At under $1,000, Dysolve represents significant potential savings for schools and districts.

Clinical trials using Dysolve have been underway since 2022 with 200 participants. Dysolve is set to publish the preliminary results of its clinical trials in the next few weeks.