• 

  Agent Architecture - Slide 3

• 

  Agent Architecture - Slide 1

• 

  Agent Architecture - Slide 2

HackVHS: Profiling Vaccine Hesitancy

Inspiration

Vaccines save millions of lives each year, yet vaccine hesitancy remains a persistent challenge, driven by a combination of misinformation, emotional barriers, and mistrust of authorities. Our inspiration came from the need for a personalized, data-driven solution that addresses these barriers directly. By understanding individual concerns and tailoring outreach to meet people where they are, HackVHS seeks to build trust, foster informed decision-making, and ultimately increase vaccine confidence.

What it does

HackVHS is an AI-powered tool designed to combat vaccine hesitancy through a combination of data analysis, user profiling, and personalized engagement. It starts by analyzing real-world data, such as vaccine-related Twitter posts, to identify common reasons for hesitancy. This data is displayed on an interactive dashboard where users can explore the findings and fill out a survey. Based on their survey responses, users are classified into three categories of persuasion: logic (logos), emotion (pathos), or peer-pressure/authority (ethos).

A Retrieval-Augmented Generation (RAG) agent powered by a Vectara LLM and trained on WHO data then provides tailored responses. For instance, users classified as logic-driven will see data-heavy graphs and statistics, while emotion-driven users will receive anecdotal stories, and authority-driven users will encounter endorsements from trusted experts. As users interact with the system, a patient journey agent tracks and analyzes their responses, visualizing their journey and highlighting key touchpoints where their opinions shifted. Medical Affairs teams can access these insights to refine public outreach strategies and better address vaccine hesitancy in the future.

How we built it

HackVHS combines multiple technologies to deliver its features. We analyzed data from vaccine-related Twitter posts to identify reasons for vaccine hesitancy and integrated these insights into an interactive dashboard. The dashboard is linked to a survey, which allows us to categorize users into logos, pathos, or ethos-based persuasion profiles. A RAG agent powered by a Vectara LLM, trained on trusted WHO data, is responsible for generating personalized responses to convince users of vaccine efficacy. Additionally, the patient journey agent uses survey responses and interaction data to plot detailed visualizations of each user’s journey, which can then be accessed by Medical Affairs teams.

Challenges we ran into

Integrating diverse data sources presented significant challenges, particularly when processing unstructured social media data and ensuring its relevance. Programmatically integrating the active agents into the workflow of our front end posed another challenge.

Accomplishments that we're proud of

We successfully built an interactive dashboard that integrates real-world vaccine hesitancy data and user interaction. It classifies users into persuasion categories and delivers tailored messaging using a RAG agent, enabling more effective engagement. The patient journey visualization highlights key insights for Medical Affairs teams to refine outreach strategies. While there is more to improve, HackVHS demonstrates potential in addressing vaccine hesitancy and fostering informed decisions.

What we learned

We learned that vaccine hesitancy is a deeply nuanced issue, requiring a blend of logic, emotion, and authority-based approaches to address effectively. Real-world data, like social media posts, provides valuable insights but requires extensive preprocessing to yield meaningful results. We also discovered the importance of balancing personalization with scalability in AI-driven tools, as well as the need for empathy in designing solutions for sensitive public health challenges.

What's next for HackVHS

The next step for HackVHS is expanding its data sources to include more scientific publications. Another focus would be deploying HackVHS at scale by partnering with public health organizations, clinics, and governments to target regions with high vaccine hesitancy. As we gather more user data, we plan to leverage machine learning to refine our patient classification system and optimize the patient journey visualization for even greater impact. Ultimately, our goal is to make HackVHS a cornerstone tool in global efforts to improve vaccine confidence and uptake.

HackVHS isn’t just a tool—it’s a step toward building trust, improving public health outcomes, and saving lives. Together, we can help people make informed decisions about their health and create a more resilient world.

Built With

• fetch.ai
• llamacloud
• openai
• python
• tavily