Local Interpretable Model-Agnostic Explanations for Online Maternal Healthcare

In culturally conservative communities, access to authentic sexual, reproductive, and adolescent information is scarce, particularly in low and middle-income countries. This has led to an over-reliance on social media and online communities to obtain such information, hence leading to the proliferation of fake and inappropriate healthcare advice. Moreover, there is no regulatory body to verify and validate shared healthcare information on online platforms. Individuals often disguise their identity while seeking sensitive information on sexual, reproductive and maternal health online. This has facilitated untraceable spread of incorrect information and harmful medical advice among social groups. These variations in social dynamics result in healthcare disparities, which reinforce health inequalities. In this paper, we propose the use of interpretable machine learning to evaluate online maternal medical advice for authenticity. We report on the negative results of Machine Learning Models attempt to distinguish between authentic and fake medical advice and urgently advocate for the establishment of a sexual, reproductive and maternal health corpus for machine learning models to learn, filter and detect medical imposters or misinformation. Our work highlights the insufficiency of explainable AI in medical contexts and underscores the need for establishing regulatory bodies to ensure the authenticity of sensitive healthcare information via social media and online platforms.