Unheard voices: addressing AI bias towards accented English in educational technologies.

Artificial Intelligence (AI) has increasingly been integrated into educational technologies. From automated assessments, virtual tutoring, and adaptive learning platforms to conversational tutors and voice assistants, these tools provide personalised learning experiences, 24/7 support, and tailored access to education. However, some critical challenges emerge; these include the persistent bias towards accented English (Azizov, 2023). Native accent, according to linguistic scholars Jenkins (2020), refers to the pronunciation patterns that are associated with regions considered linguistically dominant for a certain language. English is spoken globally in a multitude of accents, shaped by geography, culture, and identity; native English accent refers to American English or Received Pronunciation of the United Kingdom which are considered the "standard" accents. This bias disproportionately affects learners with diverse linguistic backgrounds, leading to misinterpretations, reduced accessibility, and inequitable learning experiences. Learners who speak English with regional or non-native accents often encounter misinterpretations, inaccuracies, and diminished interactions when engaging with speech-based educational technologies. These errors are not merely technical flaws; they reveal deeper issues of linguistic exclusion and inequity embedded within the design and deployment of AI systems.