Speaker: Yun Fei Liu

Abstract: As artificial intelligence becomes increasingly capable of generating working code from everyday language, the human role shifts toward structuring problems and evaluating solutions. To enable productive human-AI collaboration, it is now time to understand how programming is implemented in the human mind and brain behind the machines.

Programming is a recent cultural invention that our brains did not evolve for. So, how do we learn to think in code? One possibility is that programming adapts the brain’s language network, much like programming ability emerges in large “language” models through intensive training on text. Alternatively, programming may build upon pre-existing neural systems for logical reasoning.

In this talk, I present evidence for the latter. Brain imaging studies show that programming code primarily engages the reasoning network, where algorithmic structure is represented. Such representations exist even before formal programming instruction, when people read plain-English descriptions of algorithms. After instruction, the same neural patterns are reused to represent programming code. Furthermore, behavioral evidence shows that that reasoning ability, rather than language ability, predicts programming learning outcomes. Meanwhile, the language network plays a complementary role. It produces an initial interpretation of the code, which the reasoning network then elaborates into a structured mental model.

Together, these findings provide a foundation for understanding how neural representations of algorithms develop over time, both within a naturalistic programming session and throughout the development of programming expertise. This work has implications for programming education and for designing tools that expand creative expression through technology.

Bio: Yun-Fei Liu received his PhD in cognitive neuroscience from Johns Hopkins University, where he is currently a postdoctoral researcher in the lab of Dr. Marina Bedny. He began with a broad interest in the neuroscience of reading, exploring the neural bases of natural Chinese reading and braille literacy in blind individuals during his graduate training. His work on programming grew out of a simple question, partly inspired by his undergraduate background in electrical engineering: how does the brain “read” code? That curiosity gradually developed into a broader research program on how the brain engages in computational thinking.

He uses functional MRI, behavioral experiments, and computational analyses to investigate these questions and is continually interested in new methodologies and interdisciplinary collaborations. In parallel with his academic research, Liu works as a data analyst for the Medical Evidence Project at the Center for Scientific Integrity, where he develops large-scale databases of medical reviews and analytical tools to identify potential errors in the scientific literature.

Yun Fei Liu's Website