MIT researchers study which parts of brain are engaged when person evaluates computer program

An MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) study has delved into the world of computer programming using functional magnetic resonance imaging (fMRI) in hopes of gaining a better handle on the field as a whole.

Presented at the Neural Information Processing Systems Conference, researchers sought to ascertain which brain systems are involved as part of the process, an MIT news release said. The research picked up where the same team of researchers left off in a 2020 study, which utilized fMRI to observe the brains of programmers as they comprehended small snippets of code.

While the initial study found that code comprehension did not consistently activate the brain's language system, researchers in the recent study sought to determine whether they could reverse-engineer and decipher the code elements that programmers were looking at based on the observed brain signals.

The study revealed that both the language system and the multiple demand network were involved in encoding code-related information. But when it came to dynamic analysis, including analyzing aspects of code that change over time, the multiple demand network encoded that information more effectively than the language processing center. As part of their study, the researchers incorporated machine learning models called neural networks that were trained on computer programs, enhancing their ability to study connections between brain activity and code comprehension. In the end, they noted that brain signals observed when participants examined code resembled the patterns of activation observed when the neural networks analyzed the same code, raising the possibility of using computational models of code to better understand how our brains process programs.

The interdisciplinary team of researchers concluded that the insights gained from the study provide valuable information on how discrete pieces of computer programs are encoded in the brain. At the same time, the researchers acknowledged that understanding more complex coding tasks, such as code composition, remains a mystery. Code composition involves combining simple operations to build complex programs, similar to assembling individual bars of music to create a symphony.

The study was funded by grants from institutions that included the MIT-IBM Watson AI Lab, the MIT Quest for Intelligence, the National Institutes of Health, the National Science Foundation, the McGovern Institute for Brain Research, the Simons Center for the Social Brain, and the MIT Department of Brain and Cognitive Sciences.