Measles cases are rising. Vaccination rates are falling. And the gap between public health messaging and parental concerns keeps growing.
The typical response is more information: fact sheets, debunking articles, CDC statistics. But information alone doesn't change minds—especially when trust has eroded.
What we learned
I've been exploring how the behavioral science framework might apply to vaccine hesitancy, specifically for MMR (measles, mumps, rubella) vaccination.
The core insight from this work: a chatbot isn't enough.
Damon Centola's research on complex contagion predicts exactly this. Behavior change requires multiple reinforcing interactions from trusted sources. A single AI conversation—no matter how sophisticated—can't provide that.
What's actually needed
A complete system would need to include:
Different interfaces for different stakeholders
- For parents: Conversational tools that meet them where they are—addressing their specific concerns, providing evidence in accessible language
- For nurses: Dashboards showing which families are hesitant, suggested talking points, conversation tracking
- For school administrators: Vaccination rate data by school, progress toward herd immunity targets, notification tools
Visible community data
Parents respond to what others in their community are doing. A visualization showing "87% of families at your child's school have vaccinated—here's what it would take to reach 95%" creates social proof and shared purpose.
Trusted messenger activation
The pediatric nurse who's known the family for years has more credibility than any chatbot. The system should equip these trusted messengers with the right tools and information—not replace them.
Multiple touchpoints over time
A single conversation rarely changes deeply held beliefs. The system needs to support ongoing engagement: follow-up messages, check-ins, new information as it becomes relevant.
The DAWN prototype
As part of this exploration, I built a conversational prototype called DAWN (Dynamic Advocacy for Wellness Network). It demonstrates one piece of the puzzle: an AI system that can engage with vaccine-hesitant parents in a thoughtful way.
Key features of the prototype:
- Concern identification: Categorizes parent concerns (autism fears, ingredient worries, natural immunity beliefs, etc.) and tailors responses accordingly
- Adaptive personas: Adjusts tone and approach based on how the parent communicates—more analytical for data-driven parents, more empathetic for worried parents
- Stage awareness: Recognizes where parents are in their decision process (pre-contemplation, contemplation, preparation, action) and meets them there
- Evidence integration: Draws on peer-reviewed research and trusted sources (CDC, WHO, Children's Hospital of Philadelphia) to address specific concerns
But I want to be clear: the chatbot alone isn't the solution. It's a prototype of one component in what would need to be a much larger system.
The evidence base
The approach draws on research about what actually changes minds on vaccine hesitancy:
- Motivational interviewing works better than argumentation
- Trusted messengers matter more than information quality
- Stories are often more persuasive than statistics
- Addressing specific concerns beats general messaging
- Multiple conversations over time beat single encounters
The MMR vaccine is a particularly urgent case: measles is one of the most contagious diseases, requiring 95% vaccination for herd immunity. Current rates in many communities have fallen well below that threshold.
The hard problems
This work raises difficult questions:
- Ethics of persuasion: Where's the line between helping people make informed decisions and manipulating them?
- Trust recovery: What do you do when institutional trust has broken down?
- Misinformation asymmetry: Debunking false claims often takes more effort than making them
- Identity and community: For some, vaccine skepticism is tied to group identity—changing the behavior means risking social bonds
These aren't solved problems. Any serious effort needs to grapple with them honestly.
Current status
This is exploratory work. The DAWN prototype demonstrates what's possible with conversational AI for this domain. The broader framework suggests what a complete system might look like.
The next steps would involve:
- Partnerships with healthcare systems to understand nurse workflows
- School district collaboration for vaccination data integration
- Testing conversational approaches with real parents
- Measuring actual behavior change, not just conversation quality