AI & Research Data: Making Sense of Complexity

this illustration shows how ZENN, a new kind of AI model, helps computers make sense of messy, real-world information. The flowing, multicolored surface represents the many possible patterns hiding inside complex data, even when that data comes in very different forms. Images, written text and location data-shown by the icons at the bottom-are all combined into one shared picture. By bringing these different sources together, ZENN can spot meaningful patterns and make better predictions than traditional models that struggle with inconsistent or imperfect data. Credit: Jennifer M. McCann

Researchers at Penn State have developed a new artificial intelligence framework, called ZENN, that can analyze inconsistent and imperfect data-a breakthrough that could accelerate discoveries in fields from alzheimer’s disease research to materials science. The findings were featured as a showcase in the Proceedings of the National Academy of Sciences (2026). DOI: 10.1073/pnas.2511227122

Provided by
Pennsylvania State University

Why Traditional AI Struggles with Real-World Data

What makes ZENN different from other AI systems? Traditional machine-learning models frequently enough assume data is consistent, but real-world measurements vary widely in quality, resolution, and reliability. Imagine trying to piece together a puzzle where some pieces are crystal clear,while others are faded or missing chunks.That’s the challenge many AI systems face when analyzing data from different sources.

ZENN, short for Zentropy-Embedded Neural Networks, addresses this issue by teaching AI to recognize and adapt to hidden differences in data quality. The framework was created by Shun Wang, postdoctoral scholar of materials science and engineering; Wenrui Hao, professor of mathematics; Zi-kui Liu, professor of materials science and engineering; and Shunli Shang, research professor of materials science and engineering.

The Science Behind ZENN’s Approach

The foundation of ZENN lies in a concept called Zentropy, an advanced theory of entropy developed by Zi-Kui Liu. Zentropy posits that systems naturally move toward disorder unless energy is applied to maintain order. Researchers embedded principles from thermodynamics directly into neural networks-AI systems modeled after the human brain-allowing the model to distinguish meaningful signals from noise and uncertaint