AI Enhances Real-Time Selection of Public Health Messages for Government Campaigns
Public health campaigns often face challenges in efficiency. They tend to be costly and rely heavily on creative intuition without clear evidence on what drives behavioral change. A new study from the University of Pennsylvania, alongside government and community agencies and researchers from the University of Illinois and Emory University, demonstrates how artificial intelligence (AI) can improve message selection using theory and real-time data.
The research team, led by social psychologist Dolores Albarracín, developed computational methods that automatically generate HIV prevention and testing campaigns tailored for U.S. counties. They tap into real-time social media to source messages, creating a dynamic repository that reflects community-generated content.
How AI Selects Actionable Messages
Social media serves as a continuously updated pool of messages created by communities. The researchers designed AI tools to identify HIV prevention and testing posts from U.S. social media platforms. These messages were then curated based on their “actionability,” meaning their potential to drive people toward taking action. The AI also filters content to ensure relevance to specific target groups—in this case, men who have sex with men (MSM).
Validating AI-Selected Messages
- Computational Analysis: Confirmed the AI effectively selected messages with the intended qualities.
- Online Experiment: Tested with MSM participants who rated AI-selected messages as more actionable, personally relevant, and effective compared to control messages.
- Field Experiment: Involved public health agencies and community organizations across 42 U.S. counties. Agencies using the AI selection process posted significantly more HIV prevention messages on social media.
The study also compared AI-selected messages that were reviewed by humans versus those that were not. While all AI-selected messages outperformed control messages, human vetting improved effectiveness further. The researchers emphasize the necessity of a brief human review to eliminate harmful content and misinformation.
Implications for Government Public Health Campaigns
This research marks the first empirical evidence supporting automatic public health message selection for community and government use. AI can provide a cost-effective, scalable method for agencies to continually update and disseminate impactful health messages.
Integrating AI-driven message selection enables public health teams to rely more on data and theory rather than guesswork. This approach can accelerate the creation of campaigns that resonate with target populations, improving outreach and engagement.
For government professionals interested in expanding their AI capabilities for public health or other communication efforts, exploring specialized AI training courses can be valuable. Resources such as Complete AI Training’s government-focused AI courses offer practical knowledge to apply these technologies effectively.
Further Reading
The study, titled "Living health-promotion campaigns for communities in the United States: Decentralized content extraction and sharing through AI," was published in June 2025 in PNAS Nexus. It provides a detailed account of the AI methods and their implications for public health messaging.
Your membership also unlocks:
