EarlyBird is serious fun, thanks, in part, to Artificial Intelligence.
At EarlyBird, we believe that assessments should be fun for children and easy for teachers, while accomplishing massively challenging and important science-based feats, like accurately predicting whether a child is at risk for dyslexia and identifying their specific early literacy strengths and weaknesses. EarlyBird accomplishes this—all within an engaging game format—by incorporating multiple forms of AI into our technology platform.
Assessing oral language with voice recognition AI
EarlyBird is unique among screeners because it includes assessment of a young child’s oral language skills—a critical, yet hard to assess, indicator of a child’s true knowledge and understanding of key literacy milestones.
Typically, oral skill assessments require the one-on-one involvement of a trained test administrator who is responsible for scoring each child’s oral performance. This 1:1 human assessment is time-consuming, especially when an entire class or grade needs to be assessed. Plus, research shows that human scoring is prone to bias, fatigue and scoring error. This can result in some students getting incorrectly flagged as being at risk for dyslexia, and other students slipping through the cracks.
Using voice AI, EarlyBird captures the child’s voice recordings as they occur and provides instantaneous analysis and automatic scoring. EarlyBird partners with a leading voice AI company, SoapBox Labs, that uses sophisticated algorithms and machine learning techniques to analyze audio input, identify the sounds (phonemes) spoken, and match what the child said to the target answer (or a set of possible answers) to automatically score these input responses.
EarlyBird incorporates the SoapBox scoring into proprietary algorithms to produce accuracy results and an extensive and growing data set to norm results against a nationwide, demographically representative population. Privacy and security are high priorities for EarlyBird and SoapBox Labs, and only anonymized, de-identified information is stored for ongoing product improvement.
Recently, SoapBox Labs became the first AI company to be certified for Prioritizing Racial Equity in AI Design by Digital Promise and the EdTech Equity Project. SoapBox’s highly accurate speech recognition engine works in the real world because it’s built on a large, proprietary, database of natural child speech, of all accents and dialects, from noisy environments, and from 193 countries.
The most important part is that it analyzes a child’s performance consistently and with no regard for that child’s race, background, age, or socioeconomic status. This removes the possibility of bias—whether conscious or unconscious—in the administration and scoring process. And, in this day of teacher shortages and limited PD time, EarlyBird’s automatic scoring saves precious teacher time and ensures consistency of results.
Algorithms built with Supervised Machine Learning models
Addressing the dyslexia screening regulations put forth in over 40 states, EarlyBird is a screener that predicts, through proprietary, multifactorial, yet easily explainable algorithms, which children are at risk for dyslexia.
EarlyBird accomplishes this with the use of Supervised Machine Learning, with algorithms that are trained on labeled data, where input-output pairs are provided and the algorithm maps inputs to correct outputs. These algorithms were trained on extensive, multi-year EarlyBird validation studies conducted on thousands of children nationwide adjusted for each grade and time of year of administration.
EarlyBird employs Computer Adaptive Technology (CAT) to personalize the assessment for each child. CAT works by utilizing AI-driven adaptive algorithms to tailor the assessment experience to each individual’s abilities. The process starts with an initial question of moderate difficulty. Based on the child’s response, the algorithm determines their proficiency level and selects the next question, adjusting its difficulty accordingly. Supervised machine learning plays a critical role in making the adaptive assessment process effective and efficient. The adaptive algorithm used in CAT relies on AI techniques to analyze the child’s responses, estimate proficiency levels, and dynamically generate personalized test items. This adaptive process continues, refining the estimation of the child’s skills, and providing a personalized, efficient, and accurate assessment tailored to their abilities.
A word about our growing, and one-of-a-kind data set
EarlyBird has been used with 40,000 students nationwide, representing a demographically and geographically diverse population, reflective of today’s students. With each testing period (three times per year), EarlyBird recalibrates our data norms and today, can provide 100% post-COVID data norming sets. Additionally, with over 15 subtests adapted for the developmental stage of the child, EarlyBird is the most comprehensive assessment available (including oral language, Rapid Automatized Naming, sound-symbol and an extensive array of phonological and phonemic awareness subtests). The quality of work done through Artificial Intelligence is guided by the data sets used for training, and we have the most comprehensive student data sets available, driving the most thorough insights.
Interested to learn how EarlyBird’s advanced dyslexia screening capabilities can improve literacy outcomes in your district? Contact our team to learn more.