Flexible thinking relies on 'Lego blocks' in the brain

The brain sets the stage for what humans do best. Artificial intelligence can write essays, scan medical images, and beat world champions at complex games, but people still outperform AI in one key area: flexible thinking.

You can start a new job, learn fresh software, and figure out a new device in the same week without starting from scratch. Most AI systems stumble on that kind of quick shift.

A new study helps explain why. Biological brains do not reinvent the wheel whenever they face a new task. They reuse mental pieces they already have and combine them in fresh ways.

That reuse gives us a kind of built-in shortcut for learning, and it is something current AI still finds very hard.

Flexible thinking in the brain

Neuroscientists at the Princeton Neuroscience Institute looked closely at how the brain reshapes itself for new tasks.

“State-of-the-art AI models can reach human, or even super-human, performance on individual tasks. But they struggle to learn and perform many different tasks,” said study senior author Dr. Tim Buschman.

“We found that the brain is flexible because it can reuse components of cognition in many different tasks. By snapping together these ‘cognitive Legos,’ the brain is able to build new tasks.”

Researchers call this idea compositionality. It means using existing skills in new combinations.

Repurposing existing skills

Sina Tafazoli is a postdoctoral researcher in the Buschman lab at Princeton and lead author of the new study.

“If you already know how to bake bread, you can use this ability to bake a cake without relearning how to bake from scratch,” said Tafazoli.

“You repurpose existing skills – using an oven, measuring ingredients, kneading dough – and combine them with new ones, like whipping batter and making frosting, to create something entirely different.”

Flexible thinking in monkeys

To see compositionality in action, the team trained two male rhesus macaques on three related categorization tasks.

In each task, the animals looked at a colorful, shifting blob on a screen and had to decide what it was.

Sometimes the choice was about shape, such as deciding whether the blob looked more like a bunny or the letter T.

Other times the choice was about color, judging whether it was more red or more green.

The blobs could be very clear or quite ambiguous, so some decisions were easy and others were far from obvious.

Testing the brain’s use of patterns

The monkeys reported their answers using eye movements. To respond, each animal looked in one of four different directions, and the same physical movement could mean different things in different tasks.

In one task, glancing left meant the blob looked like a bunny, while glancing right meant it looked more like a T.

One of the color tasks used the same eye directions as the shape task, but focused on color instead of shape.

Both color tasks used the same basic red-versus-green judgment but mapped that choice to different eye directions.

By mixing and matching these elements, the researchers could test whether the brain reused internal patterns when tasks shared certain parts.

Cognitive building blocks

When the team recorded activity across the brain, one region stood out.

The prefrontal cortex, a frontal area involved in planning and decision making, showed clear patterns that repeated across tasks that shared a common goal, such as deciding on color.

Groups of neurons formed reusable activity patterns the authors interpret as cognitive building blocks.

“I think about a cognitive block like a function in a computer program. One set of neurons might discriminate color, and its output can be mapped onto another function that drives an action,” said Dr. Buschman.

“That organization allows the brain to perform a task by sequentially performing each component of that task.”

Focusing on the task at hand

In one color task, the animal put together a block that handled color discrimination with another block that converted the decision into a particular eye movement.

In a shape task that used the same directions for responses, the brain instead connected a shape-decoding block to the eye-movement block.

The researchers also found that the prefrontal cortex reduces activity in blocks that are not currently needed, which helps keep attention on the current goal.

“The brain has a limited capacity for cognitive control,“ Tafazoli explained. “You have to compress some of your abilities so that you can focus on those that are currently important.”

“Focusing on shape categorization, for example, momentarily diminishes the ability to encode color because the goal is shape discrimination, not color.”

What this means for AI

The findings speak directly to a big challenge in machine learning. Many artificial neural networks handle a single task well but stumble when they must learn many different ones over time.

“A major issue with machine learning is catastrophic interference. When a machine or a neural network learns something new, they forget and overwrite previous memories,” said Tafazoli.

“If an artificial neural network knows how to bake a cake but then learns to bake cookies, it will forget how to bake a cake.”

Human brains seem to avoid this by separating reusable components from the specific ways they are combined. The color block and the eye-movement block stay available.

New tasks mostly involve new connections between blocks, not a full rewrite.

If engineers can build similar modular structures and selective “quieting” into AI systems, they may eventually handle continuous learning without erasing what they already know.

Clues for treating brain disorders

The study also has important clinical implications. Many neurological and psychiatric conditions make it hard for people to adapt to new rules or apply familiar skills in fresh situations.

The researchers highlight schizophrenia, obsessive-compulsive disorder, and certain kinds of brain injury as examples where flexible thinking often breaks down.

If those difficulties turn out to involve damage or disruption in how cognitive blocks are formed or recombined, that knowledge could guide new therapies.

“Imagine being able to help people regain the ability to shift strategies, learn new routines, or adapt to change,” said Tafazoli.

“In the long run, understanding how the brain reuses and recombines knowledge could help us design therapies that restore that process.”

The full study was published in the journal Nature.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. 

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–