This research introduces a comprehensive framework for aligning Large Language Models with specific online communities and systematically evaluating the fidelity of these alignments.

• Creates LLMs that authentically represent community values and language patterns
• Develops robust evaluation methods across multiple dimensions including authenticity and emotional expression
• Demonstrates practical applications with eating disorder communities, enabling ethical study of sensitive health topics
• Provides a foundation for using AI to understand community dynamics without direct patient engagement

For healthcare researchers, this work offers powerful new tools to study sensitive medical communities ethically, potentially transforming how we gather insights about conditions like eating disorders while protecting vulnerable populations.

Improving and Assessing the Fidelity of Large Language Models Alignment to Online Communities