Research published in Nature has revealed that neural computations in different individuals can be implemented to solve the same decision-making tasks, even when the behavioral outcomes appear identical.
Cognitive flexibility is the ability of a brain to adapt its response to the same external stimulus, like light or sound, based on different contexts. For example, if someone calls your name in a crowded room, you must focus on the sound's location or the voice characteristics to identify the person. This flexibility in selecting and processing relevant information while ignoring irrelevant information is crucial for survival and effective interaction with our environment.
While previously studied, the individual variability in neural computations yielding the same outcomes is poorly understood and lacks a comprehensive framework. The researchers in the Nature study aimed to understand these mechanisms.
Phys.org spoke to the first author of the paper, Dr. Marino Pagan, who performed the experiments described in the paper as a postdoctoral researcher under the supervision of Carlos Brody at Princeton University.
"As a computer engineer, I have always been deeply fascinated by our brain's ability to perform complex computations. In particular, I am very interested in understanding cognitive flexibility. The ability is crucial in our everyday lives, and its impairment can create issues in neurodevelopmental disorders," said Dr. Pagan.
Investigating cognitive flexibility
The researchers aimed to develop a framework to explain how neural networks compute context-dependent selection, and link neural and behavioral variability. They began by training rats to perform decision-making tasks based on external auditory cues. Their decision-making was based on a set of two alternating rules.
The location rule required the rats to respond to the location of a series of auditory clicks. On the other hand, the frequency rule required them to respond to the frequency of the clicks while ignoring their location.
A context cue before each rule informed the rats of which rule to follow. The rules switched rapidly, requiring the rats to adjust their decision-making process quickly.
"Rats can learn to solve this task with very high accuracy, and analysis of their behavior and neural activity over many trials allows us to precisely characterize the mechanisms they are using to select the context-relevant stimuli and make the right decisions," said Dr. Pagan.
To understand how the rats processed each task, the researchers measured the rats' neural activity. The measurements were recorded from the frontal orienting fields (FOF), a region of the brain involved in decision-making and orienting responses to external stimuli, especially in terms of adjusting behavior based on context. This would later help to understand the mechanisms at play in content-dependent decision-making.
Developing a framework
The researchers developed a theoretical framework to explain how the brain computes context-dependent decision-making. This was based on three possible dynamic solutions for how the brain might process information.
Next, the researchers developed RNNs to simulate how each solution could be used to solve the task presented to the rats. RNNs are a type of artificial neural network used in machine learning, designed to handle sequential data -- like time series or patterns that change over time.
"RNNs can be trained to solve the same task as the rats using different mechanisms. We characterized the variations in their neural responses, which served as signatures for these mechanisms. We then applied the same analyses to the rats' neural activity and found different neural signatures across individual animals," explained Dr. Pagan.
This approach allowed them to map out the full range of possible strategies that could solve the task and match them to the actual neural signatures observed in the rats.
Do all brains use the same mechanism to solve these types of complex tasks?
No. The researchers found that not all brains use the same mechanism to solve a task, even if the same outcome is achieved.
"Measurements of brain dynamics differed between individual animals, suggesting that different brains use different mechanisms to solve the same task, even though on the surface it might look like their behavior is very similar. This result is important because it has been very hard to study this kind of individual variability before," noted Dr. Pagan.
Additionally, the team found a strong correlation between variability in neural responses and behavioral outcomes, identifying neural signatures for these correlations. The results from the RNN models matched the observed brain activity in the rats, confirming their finding of a high degree of individual variability in handling the same task.
Variability and disorders
This study provides the first comprehensive frame for identifying individual variability in context-based flexible decision-making, bridging the gap between neural activity and behavior. The evidence found is concrete and provides a tool for studying this variability.
Speaking of developing this research further, Dr. Pagan mentioned two directions currently being pursued at the University of Edinburgh, where he is the Group Leader at the Simons Initiative for the Developing Brain (SIDB).
"The first focuses on understanding the source of the substantial variability observed across different brains. We want to know if this variability is innate and predetermined from birth or whether it develops during learning," he said.
"The second direction investigates how cognitive flexibility and decision-making are affected in neurodevelopmental disorders. While significant strides have been made in identifying the genetic underpinnings of these disorders, the mechanisms that connect genetic mutations to cognitive deficits remain poorly understood," concluded Dr. Pagan.