Artificial Intelligence (AI) technologies serve as foundational enablers for modern social media and digital health services, yet they simultaneously act as vectors for misinformation. A significant challenge arises in non-English contexts and low socioeconomic classes, where limited data hampers the training of AI models for effective detection. Consequently, culturally and linguistically diverse (CALD) communities struggle to access trustworthy health information through AI-driven tools. Current AI tools underperform due to a lack of training data and are largely unable to account for language nuances and traditions in non-English settings.
This research addresses these gaps by proposing a CALD-friendly AI-based health misinformation detector and providing a dashboard for medical professionals to analyze this misinformation, which is crucial for mitigating concerns among CALD populations. We conduct a series of experiments using a Bangla-translated health misinformation dataset to evaluate the performance of various Small Language Models (SLMs). SLMs are particularly relevant here due to the frequent underperformance of Large Language Models (LLMs), which often arises from insufficient domain-specific knowledge and high costs associated with resource-intensive fine-tuning.
The results demonstrate that the Phi-4 model achieves an ideal balance between precision and recall in claim extraction. To address the limitations of SLMs, we design and test a novel health misinformation detection framework grounded in Responsible Natural Language Processing (NLP), incorporating cultural sensitivity, potential for harm, and communication quality, thereby providing a holistic perspective for evaluating misinformation in low-resource languages.
Blogger's Review: This paper offers an innovative approach to tackling health misinformation in low-resource languages by integrating small language models with a culturally sensitive NLP framework. The application case in Bangla highlights how responsible AI design can enhance the credibility and accessibility of information in data-scarce environments, presenting significant social value and practical implications.