Grant: Project Grant

More than 30% of children in Canada grow up learning more than one language. Multilingual children often produce speech sounds differently from monolingual English-or French-speaking children, making it difficult for speech-language pathologists (SLPs) to determine whether a child's speech patterns are typical for their linguistic background or indicative of a disorder. Training SLPs to recognize the full range of speech patterns across hundreds of languages is not feasible, and current assessment practices often rely on caregiver input. While this collaborative approach can help, it remains subjective, time-consuming, and inconsistent, while creating challenges for clinicians and directing pressure on the healthcare system to manage caseloads efficiently. To address these issues, we propose testing the feasibility of an unbiased artificial intelligence (AI) tool that automatically analyzes children's speech and identifies features associated with SSD, regardless of the child's language background. This AI-supported approach aims to improve the speed, equity, and accuracy of diagnosis while minimizing over- or under-diagnoses. We will engage clinicians and parents throughout the process: sharing findings, gathering feedback, and co-designing implementation strategies to ensure the AI tool is aligned with real-world needs. By involving families and SLPs directly, we aim to support effective adoption, enhance clinical decision-making, and promote fairness in clinical care.