Towards Explainable AI Writing Assistants for Non-native English Speakers

AI writing assistants lack transparency for vulnerable users: The study found that non-native English speakers struggle to evaluate paraphrasing suggestions because AI tools don't explain why they made specific changes, creating a significant usability gap for this population.

Explanations are essential, not optional: Interview participants with varying English proficiency levels consistently identified the absence of explanations as their primary barrier to confidently using AI writing assistants, suggesting that explainability is a core accessibility need rather than a nice-to-have feature.

Users develop workarounds when systems fail them: When faced with unexplained AI suggestions, the 15 non-native speakers in the study developed their own strategies to assess quality—a finding that highlights how lack of transparency forces users to shoulder additional cognitive burden.

Design solutions can democratize AI tool access: The researchers proposed four user interface designs that incorporate explanations specifically tailored to non-native speakers' needs, demonstrating that responsible AI design can make powerful tools more equitable and trustworthy for underserved populations.

Responsible AI requires considering diverse user needs: This research emphasizes that AI systems designed without input from non-native speakers—a significant global user population—may inadvertently exclude or disadvantage users who need support most.