AI may soon detect dyslexia early from children’s handwriting

Researchers are exploring how artificial intelligence can analyze children’s handwriting to identify early signs of dyslexia and dysgraphia, potentially revolutionizing screening processes and offering timely support to young learners. This innovative approach could make diagnostic tools more accessible and efficient, especially in areas facing shortages of specialized professionals.

The challenge of early detection in learning differences

Dyslexia, a learning disorder primarily affecting reading and language processing, and dysgraphia, a condition impacting writing abilities, can significantly hinder a child’s academic journey and socio-emotional development if not identified and addressed early. Current screening methods, while effective, often come with drawbacks: they can be expensive, require considerable time from specialists, and typically focus on diagnosing only one condition at a time. This can lead to delays in children receiving the tailored support they need to thrive.

Adding to this challenge is a nationwide shortage of speech-language pathologists and occupational therapists. These professionals play a crucial role in diagnosing dyslexia and dysgraphia, respectively. Their limited availability can create bottlenecks, preventing many children, particularly those in underserved communities, from accessing timely evaluations.

“Catching these neurodevelopmental disorders early is critically important to ensuring that children receive the help they need before it negatively impacts their learning and socio-emotional development,” explains Venu Govindaraju, PhD, SUNY Distinguished Professor at the University at Buffalo’s Department of Computer Science and Engineering and the study’s corresponding author. “Our ultimate goal is to streamline and improve early screening for dyslexia and dysgraphia, and make these tools more widely available, especially in underserved areas.”

This significant work is part of the broader mission of the National AI Institute for Exceptional Education, a research organization led by the University at Buffalo. The institute is dedicated to developing advanced AI systems designed to identify and assist young children with speech and language processing disorders.

The concept of using AI to analyze handwriting isn’t entirely new. Decades ago, Professor Govindaraju and his colleagues pioneered groundbreaking work in this field. They employed machine learning, natural language processing, and other AI techniques to analyze handwriting – an advancement that organizations like the U.S. Postal Service still leverage today for automating mail sorting.

The new study, recently presented in the journal SN Computer Science, proposes a similar framework. It adapts these established AI methodologies to identify specific indicators of dyslexia and dysgraphia in children’s writing. These indicators can include spelling difficulties, inconsistent or poor letter formation, issues with organizing thoughts on paper, and other subtle clues that might otherwise be missed without specialized assessment.

The research aims to expand upon previous efforts. Historically, AI applications in this domain have more frequently focused on detecting dysgraphia. This is partly because dysgraphia often presents with more overt physical differences in a child’s handwriting, making it somewhat easier for AI to observe. Dyslexia, on the other hand, is more intricately linked to reading and speech. While certain writing behaviors like persistent spelling errors can offer clues, identifying dyslexia through handwriting alone has been considered more challenging. The new framework seeks to address this by incorporating a broader range of analytical capabilities.

Addressing data needs

A common hurdle in developing robust AI models is the availability of sufficient and relevant data for training. The study acknowledges a current shortage of handwriting samples from children, which is crucial for teaching AI systems to recognize the nuanced patterns associated with these learning differences.

To overcome this and ensure their AI models are practical for real-world settings, the UB computer scientists, led by Govindaraju, collaborated closely with those on the front lines of education. They gathered valuable insights from teachers, speech-language pathologists, and occupational therapists. This collaborative approach helps ensure that the AI tools being developed are genuinely useful and viable in classrooms and clinical environments.

“It is critically important to examine these issues, and build AI-enhanced tools, from the end users’ standpoint,” emphasizes Sahana Rangasrinivasan, a PhD student in UB’s Department of Computer Science and Engineering and a co-author of the study.

A key partnership in this endeavor was with study co-author Abbie Olszewski, PhD, an associate professor in literacy studies at the University of Nevada, Reno. Dr. Olszewski co-developed the Dysgraphia and Dyslexia Behavioral Indicator Checklist (DDBIC), a tool designed to identify symptoms that overlap between the two conditions. The DDBIC focuses on 17 behavioral cues that can manifest before, during, and after the writing process.

To gather the necessary data, the research team collected both paper and tablet-based writing samples from students in kindergarten through 5th grade at an elementary school in Reno. This data collection was conducted with ethical board approval, and all student data was anonymized to protect privacy. This rich dataset will serve multiple purposes:

• Further validate the DDBIC tool itself.
• Train the AI models to autonomously complete the DDBIC screening process.
• Compare the effectiveness of the AI models against human experts administering the same screening tests.

Apple research paper unveils Matrix3D for 3D content generation

How AI uncovers insights from handwriting

The proposed AI framework is designed to perform a multi-faceted analysis of a child’s handwriting. The study outlines how the team’s models can be utilized to:

• Detect motor difficulties: By analyzing writing speed, the pressure applied to the pen or stylus, and the fluidity of pen movements, the AI can identify potential challenges with fine motor skills often associated with dysgraphia.
• Examine visual aspects: The system scrutinizes the visual characteristics of handwriting, such as letter size, consistency in spacing between letters and words, and overall layout on the page. Irregularities in these areas can be indicative of underlying issues.
• Convert handwriting to text for error analysis: The AI can digitize handwritten text, allowing it to spot misspellings, letter reversals (like ‘b’ for ‘d’), omissions, and other errors that might point towards dyslexia or dysgraphia.
• Identify deeper cognitive issues: Beyond the mechanics of writing, the AI can be trained to look for patterns related to grammar, vocabulary usage, and sentence construction, which can offer clues about cognitive processing related to language.

Ultimately, the research envisions a comprehensive tool that integrates all these analytical models. This tool would summarize the findings from each aspect of the analysis, providing a holistic assessment that could greatly aid educators and specialists in early screening.

“This work, which is ongoing, shows how AI can be used for the public good, providing tools and services to people who need it most,” states study co-author Sumi Suresh, PhD, a visiting scholar at UB.

The potential impact is significant. Early intervention can dramatically improve a child’s learning trajectory, boost their confidence, and prevent the long-term academic and emotional difficulties that can arise from undiagnosed learning differences. As AI technology continues to evolve, its role in creating more equitable and effective educational support systems is poised to grow, offering promising solutions to long-standing challenges in how we identify and assist all learners.

Featured image credit