Deep Learning

Fairfield researchers are ethically applying AI to environmental challenges.

On a monitor in the School of Engineering and Computing’s Artificial Intelligence Lab, Nusrat Zahan MS’25 peers at a murky underwater image clouded by silt and shadow. As her lines of code run, the footage sharpens—revealing the vibrant reds, blues, and yellows of invasive fish species that threaten the fragile reef ecosystem.

A few feet away, a graduate student is surrounded by glowing screens of geospatial and soil data. He is training a machine-learning model to help monitor geotechnical infrastructure—analyzing ground conditions to determine whether soil can safely support buildings and other foundations, particularly as extreme weather events become more frequent.

This kind of transformation—turning raw, unreadable data into actionable insights—is at the heart of the work underway in Fairfield’s AI Lab. Leading the charge is Sidike Paheding, PhD, associate professor and chair of the Computer Science Department. As an AI ethics researcher, Dr. Paheding teaches his students to tackle global environmental challenges by building AI models that account for bias and risk from the earliest stages of design.

“AI ethics sets the compass, establishing a solid foundation for developing future AI systems that are safe, secure, and trustworthy,” said Dr. Paheding, who serves as principal investigator on collaborative projects focused on advancing AI ethics education, funded by the National Science Foundation (NSF).

With support from the U.S. Geological Survey and under the mentorship of Dr. Paheding, Zahan’s cutting-edge underwater image enhancement techniques—particularly focused on illumination, correction, and color restoration—are powered by the in-house development of deep-learning models. Her research has been published in several international journals.

Deep learning—a subset of machine learning—draws inspiration from the structure of the human brain. The Fairfield researchers explained that deep learning models are trained through layered neural networks that ultimately recognize patterns and process complex data. Dr. Paheding and his students emphasized that deep learning models vary based on the problem being addressed and the data sets used to train them. Ultimately, the work is driven by the people behind the scenes who identify risk factors and design models intended to anticipate risk before it becomes a reality. “When developing AI models, there is no universal solution to these complex challenges,” Dr. Paheding said.

Naqibahmed Kadri, a graduate student in the MS in Computer Science program, is working with Dr. Paheding to develop AI-driven standards for monitoring geotechnical infrastructure. Their research is supported by the NSF in collaboration with the University of Mississippi (Ole Miss) and the NASA Connecticut Space Grant Consortium.

Using satellite imagery from space, Kadri and Dr. Paheding identify mine tailings impoundments—large holding areas for mining waste product, mostly composed of sand particles mixed with chemical residue. Often built near active or former mines, these impoundments can pose serious environmental risks if they fail.

The danger escalates during extreme weather events such as floods or hurricanes. If an impoundment’s structure can’t withstand the added stress, toxic material can spill into surrounding land and waterways, causing severe environmental damage and threatening human lives. The primary challenge, according to Dr. Paheding, is knowing where these impoundments are located; many are poorly documented or difficult to monitor.

To date, the Fairfield researchers have trained a deep learning model to identify more than 1,000 locations across six continents where mine tailings impoundments exist, marking a critical first step in building tools that can help monitor these sites before disaster strikes. “This dataset provides a robust foundation for advancing AI-based detection and analysis,” Dr. Paheding said.

Once trained, the AI model developed at Fairfield can do more than locate impoundments. It can help predict which sites are most vulnerable—flagging areas at high risk based on factors such as soil moisture, terrain, and environmental variables. Each site is assigned a risk profile, allowing researchers to prioritize monitoring and intervention protocols.

Through this federally funded research, students and faculty are not only advancing artificial intelligence but shaping how it can be used responsibly. By combining deep learning with careful model design, the AI Lab at Fairfield focuses on preventive solutions—building systems that are secure, transparent, and capable of addressing complex environmental challenges without introducing new risks.

For Dr. Paheding, responsibility begins long before a model is deployed. “As AI tools become more powerful, ethics must remain at the forefront of researchers’ minds—especially when it comes to mitigating bias.”

Related Stories