Language technologies, particularly generative artificial intelligence (AI), are emerging as powerful tools in the field of public health. From real-time outbreak detection systems that scour global news reports to chatbots offering mental health support and innovative conversational diagnostic tools, these advancements are addressing numerous health challenges that societies face today.
At the core of these innovations is natural language processing (NLP), a subfield of AI focused on enabling machines to understand, interpret, and generate human language. By processing and analyzing vast amounts of health-related data, NLP can provide insights that far exceed human capabilities. This is particularly valuable in areas with limited healthcare resources and weak public health infrastructure, as it allows for swift, data-led responses to pressing health issues.
Recently, an interdisciplinary team comprising experts in computer science, human geography, and health sciences conducted a thorough review of how language AI is utilized in public health across African countries. An analysis of nearly a decade’s worth of academic research revealed that practical applications of this technology are still limited. Of 54 publications examined, only two (around 4%) demonstrated measurable improvements in public health, such as enhancing individuals’ mood or increasing vaccine intent.
A significant challenge that emerged from this review is that the majority of projects tend to stagnate at the technology development phase, seldom transitioning to real-world applications. This stagnation could limit opportunities for improving health and well-being across the continent.
Current Limitations
The rapid increase in AI language technologies for public health in recent years can be largely attributed to the renewed focus on health during the COVID-19 pandemic. Tools such as health chatbots and sentiment analysis systems have been developed both in Africa and elsewhere.
Supplied
Health chatbots are designed to engage users by providing trusted health information in a user-friendly, conversational format. Sentiment analysis tools sift through social media content to gauge public sentiment and concerns. Together, these technologies can uncover misinformation and gauge shifts in public opinion, providing timely, accurate information in response.
However, these advancements are not without limitations. The majority of health technologies in Africa predominantly operate in a few languages inherited from colonial times, mainly English and French. This language barrier means that crucial health messages are often inaccessible to many local communities, leaving millions without essential information or resources.
From the findings, it’s evident that very few projects have been rigorously tested in real-world settings. Alarmingly, only one out of the reviewed systems demonstrated a measurable impact on public health.
A Successful Model
A notable success story in this arena comes from a collaboration between the Center for Global Development, the University of Chicago, and the Busara Center for Behavioral Economics. They developed a chatbot on Facebook Messenger aimed at addressing vaccine hesitancy around COVID-19 in Kenya and Nigeria. The application, however, was only available in English.
With over 22,000 users interacting with the chatbot, it provided tailored, evidence-based responses to common concerns surrounding vaccinations, including effectiveness, safety, and misinformation. The results were promising, showing an increase in users’ willingness to get vaccinated by 4%-5%, with the most significant impact observed among those initially hesitating.
The root of this success lay in the researchers’ commitment to understanding the local context. Before the rollout, extensive discussions with focus groups and social media users in both countries were conducted to identify the specific concerns that influenced attitudes toward vaccination. By addressing these particular worries, the chatbot successfully provided relevant guidance—highlighting the importance of a user-centered approach in technology design.
From Lab to Life
Though the initial development of AI technologies in public health has been accelerated by the COVID-19 crisis, they are still in the nascent stage. Emerging advancements in models like GPT-4 are lowering technical barriers to developing language solutions, allowing for more accessible creation of yet-unexplored applications with minimal data and effort. This could potentially empower smaller teams or individual developers to craft tools fine-tuned for their communities’ particular needs.
Support from investors, accelerators, and government initiatives could facilitate this critical transition from development to real-world applications.
Moreover, technology developers should emphasize the importance of community involvement, engaging health workers, and fostering multi-disciplinary collaborations to ensure the design aligns with the real-world challenges and needs of the population.
To unlock the full potential of language technologies in public health, it is crucial to:
-
Involve communities and health professionals in the process of designing natural language processing tools.
-
Enhance the availability of resources in indigenous African languages.
-
Integrate these language technologies within existing healthcare systems.
Future research must expand its focus from merely testing technical prototypes to robust real-world evaluations that can authentically measure health outcomes and impacts.
The other co-authors behind this research are: Abigail Oppong, Ebele Mogo, Charlotte Collins, and Giulia Occhini.