Tech

Researchers find out which part of the brain is active when a person evaluates a computer program


onyx

Credit: Pixabay/CC0 Public Domain

Functional magnetic resonance imaging (fMRI), which measures changes in blood flow in the brain, has been used over the past few decades for many applications, including “functional anatomy”—a way of identifying areas Which brain is activated when a person performs a particular task. fMRI has been used to examine people’s brains while they are doing all sorts of things—solving math, learning a foreign language, playing chess, improvising on the piano, solving crosswords, and even watching TV shows like “Curb Your Enthusiasm.”

A less-noticed pursuit is computer programmer—both the work of coding and the equally confusing task of trying to understand a piece of prewritten code. Shashank Srikant, Ph.D. student at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), “that’s definitely worth a look. A lot of people are dealing with code these days—reading, writing, designing, editing. error—but no one really knows what’s going on in their head when it happens.”

Fortunately, he’s made some “steps” in that direction in a paper—written with colleagues at MIT, Benjamin Lipkin (the other lead author of the paper, along with Srikant), Anna Ivanova, Evelina Fedorenko and Una-May O’Reilly—that’s presented earlier this month at the Neural Information Processing Systems Conference held in New Orleans.

New article built on a 2020 study, written by many of the same authors, uses fMRI to monitor programmers’ brains as they “understand” small pieces of code or code. (Understood, in this case, means looking at a piece of code and determining the exact result of the computation performed by that piece of code.)

The 2020 work shows that code comprehension doesn’t trigger the language system consistently, brain region language processing, explains Fedorenko, a Brain and professor of cognitive science (BCS) and co-author of the previous study. “Instead, the multi-needs network—a brain system linked to general reasoning and supporting areas like mathematical and logical thinking—was at work.” The current work, which also uses programmers’ MRI scans, will “go deeper,” she said, seeking to gain more insights.

While previous research looked at 20 to 30 people to determine on average which brain systems are relied upon to understand code, the new study looks at individual programmers’ brain activity as they process specific elements. of computer programs. For example, suppose there is a single-line piece of code that involves word manipulation and a separate piece of code that requires a math operation.

“Can I go from the activity that we see in the brain, the actual signal of the brain, to trying to reverse engineer and figure out what, specifically, what the programmer is looking at? ” Srikant asked. “This will reveal information regarding the unique coded programs in our brains.” For neuroscientists, he notes, a physical property is considered “encoded” if they can infer it by looking at someone’s brain signals.

For example, a loop—an instruction in a program to repeat a particular operation until a desired result is achieved—or a branch, another type of programming instruction that can cause the computer to switch from the from one activity to another. Based on observed patterns of brain activity, the team was able to tell if someone was evaluating a piece of code that involved a loop or a branch. Researchers can also see if the code is related to a word or mathematical symbol, and whether someone is reading the actual code or just a written description of it.

That addresses a first question an investigator might ask as to whether something is, in fact, encrypted. If the answer is yes, the next question might be: where is it encoded? In the cases cited above—loops or branches, words or operations, codes, or descriptions of them—the levels of brain activation are thought to be comparable in both linguistic and network of multiple needs.

However, a notable difference has been observed when it comes to code properties regarding what is known as dynamic analysis.

Programs can have “static” properties—such as the number of digits in a string—that do not change over time. “But programs can also have dynamic aspects, such as the number of times a loop runs,” says Srikant. “It’s not always possible for me to read a piece of code and know how long it will run.” MIT researchers have found that for dynamic analysis, information is encoded in a multi-demand network much better than in a language processing center. That finding is a clue in the quest to understand how code comprehension is distributed in the brain—which is involved and which takes on a larger role in certain aspects of the task. there.

The team performed a second series of tests, incorporating machine learning models called neural networks that were specially trained on computer programs. These models have been successful, in recent years, in helping programmers complete pieces of code. What the team wanted to find out was whether the brain signals seen in their study when participants were examining the codes resembled the activation patterns observed when the neural networks analyzed the same code. code or not. And the answer they get is qualified.

“If you put a piece of code into the neural network, it generates a list of numbers that tell you, in a way, what the program is about,” says Srikant. Brain scans of people studying computer programs also produced a list of numbers. For example, when a program is dominated by branching, “you’ll see a distinct pattern of brain activity,” he adds, “and you’ll see a similar pattern when the machine learning model tries to try to understand that same piece of code.”

Mariya Toneva of the Max Planck Institute for Software Systems considers findings like these “particularly exciting. They advance the ability to use computational code models to better understand what happens in our brains. when we read the shows,” she said.

The MIT scientists are certainly intrigued by the connections they have discovered, which shed light on how the discrete parts of the universe are. Computer Programs encoded in the brain. But they still don’t know what these new insights can tell us about how people execute more complex schemes in the real world.

Completing tasks of this type—such as going to the movies, requesting showtimes, arranging transportation, buying tickets, etc.—cannot be handled by a single unit of code and only one algorithm. single math. Instead, the successful implementation of such a plan would require “composition”—stringing together different pieces of code and algorithms into a logical sequence that leads to something new, like a collection of individual bars to form a song or even a symphony. Create model of code composition, O’Reilly, a principal research scientist at CSAIL, said, “we can’t capture it at the moment.”

Lipkin, PhD in BCS. students, see this as the next logical step—finding a way to “combine simple operations to build complex programs and use those strategies to effectively solve general reasoning tasks. ” He also credits some of the progress towards that goal that the group has made so far thanks to its interdisciplinary structure.

“We can draw on personal experience in program analysis and neural signal processing, as well as combined work on machine learning and natural language processing,” says Lipkin. “These kinds of collaborations are becoming increasingly common as neuroscientists and computer scientists join forces in their quest to understand and build shared intelligence.”

More information:
Paper: Convergent representation of computer programs in artificial and human neural networks

This story is reprinted courtesy of MIT News (web.mit.edu/newsoffice/), a popular website covering MIT research, innovation, and teaching.

quote: Researchers find out which part of the brain is involved when one evaluates a computer program (2022, 22 December) retrieved December 22, 2022 from https://techxplore. com/news/2022-12-brain-engaged-person.html

This document is the subject for the collection of authors. Other than any fair dealing for private learning or research purposes, no part may be reproduced without written permission. The content provided is for informational purposes only.

news7f

News7F: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, Sports...at the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button