3 accepted papers
Understanding the prevalence of misinformation in health topics online can inform public health policies and interventions. However, measuring such misinformation at scale remains a challenge, particularly for high-stakes but understudied topics like opioid-use disorder (OUD)—a leading cause of deat
This study introduces ValueScope, a framework leveraging language models to quantify social norms and values within online communities, grounded in social science perspectives on normative structures. We employ ValueScope to dissect and analyze linguistic and stylistic expressions across 13 Reddit c…
Large language models (LLMs) have emerged as an integral part of modern societies, powering user-facing applications such as personal assistants and enterprise applications like recruitment tools. Despite their utility, research indicates that LLMs perpetuate systemic biases. Yet, prior works on LLM…