Creating a New Kind of Doctor
We recruit and train physician leaders as comfortable taking on systemic challenges in health as caring for individual patients.
ARE YOU ONE?
Radical Collaboration. Real-World Impact.
Texas expertise fuels the discovery, delivery and diffusion of the next generation of preventions, diagnoses, treatments and cures.
LET'S GO
World Class. Close to Home.
We’re working to make person-centered, integrated care the standard in Central Texas and beyond.
Health Starts Here
More Information
GET CARE
Meet Dell Med
We’re rethinking the role of academic medicine in improving health — and doing so with a unique focus on our community.
ABOUT US
More Information
EXPLORE
Make an Appointment Give Faculty Students Alumni Directory

How Do We Think About AI in the Context of Environmental Health?

April 18, 2024

This blog is authored by Roger Peng and Sarah Chambliss, members of the Center for Health and Environment: Education and Research.

One of the most exciting developments across the world today is the advancement of artificial intelligence technology and its application to all aspects of our lives. While many aspects of how AI can be used productively are still being worked out, there have already been impressive achievements made by the various companies. People today are familiar with chat services like OpenAI’s ChatGPT or Google’s Bard, which appear to have human-like abilities to respond to a wide variety of questions. One issue that many scientists are grappling with is whether or how we can use AI tools in the context of scientific research. In particular, how can AI tools be used to improve or change the nature of environmental health research?

Defining Artificial Intelligence

The first challenge to answering this question is defining exactly what we mean when we say AI. AI has been historically difficult to specify because as concepts of technologies become more familiar to us, we tend to be less likely to attribute them to any sort of “intelligence.” Hence, technical feats that might have appeared magical just five years ago, but today are in everyday use, are no longer categorized as AI. Tasks that are handled with statistical machine learning models might have previously been considered AI are now relegated to the category of “machine learning.” That is not to say that these models are unimpressive but rather that they are well understood.

Today’s AI is generally centered around large language models that use transformer-based neural network models at their core. These models are trained on truly massive amounts of information from the web and other sources and can respond to natural language queries by returning natural language responses. This ability to “comprehend” natural language denotes a shift away from models that traditionally took numerical data as inputs and provides a more intuitive interface for human beings.

AI in Environmental Health

AI techniques have the potential to improve and enhance environmental health research in a variety of ways: We can better characterize a person’s environmental exposures by using the massive power of AI techniques to summarize and process huge amounts of data collected, such as using low-cost sensors placed inside or outside the home. We can also explore ways that AI models could be embedded in control systems that regulate our indoor environments by, such as sensing a decrease in indoor air quality and filtering the air in response.

Moving beyond the individual level, we can use AI techniques to improve population health by integrating disparate datasets to identify areas of cities or states that should be prioritized for intervention. For example, California’s Air Resources Board currently uses CalEnviroScreen to identify historically underserved neighborhoods that can be prioritized for investment, and AI could be used to better aggregate and synthesize relevant demographic and land use-based indicators.

While the use of AI tools promises to add much to environmental health research, their use in decision-making should be viewed with caution. Biases in data collection and assumptions made in the modeling process can lead to inequity in downstream decisions about resource or treatment allocation.

The environmental health science community stands to be affected by the AI revolution much like many others. AI tools deserve to be studied and adapted to the needs of environmental health researchers to improve the quality of research and to protect the public’s health. However, researchers should not blindly adopt these tools; rather, they should carefully consider their impacts on the populations being studied.