Study Shows Single Neurons Integrate Rules and Sensory Signals to Guide Decisions
Researchers from the University of Chinese Academy of Sciences observed mice adapting to changing sound-based rules, finding that different parts of the same neuron process abstract rules and sensory input.
Brain Neurons and Flexible Decision-Making
A Chinese research team has revealed how brain neurons support flexible decision-making through a study published in Science. The study suggests that:
- Rule-related information and sensory information are compared within the same neuron.
- The neuron itself takes part in cognitive computation.
Traditional Theories vs. New Insights
Traditional theory treats neurons as simple point-like units. However, recent research shows that higher cognitive functions, like flexible decision-making, require:
- Integration of massive contextual information and abstract rules.
- Dynamic adjustment of output.
This complexity challenges the simple structure of point-like neurons.
Experimental Setup
Xu Ninglong and his team at the University of Chinese Academy of Sciences conducted experiments where mice heard different sounds and had to infer hidden rules. They then changed the rules and observed the mice's adaptation using two-photon microscopy on neurons in the mice's auditory cortex.
Neuron Structure and Function
A neuron is like a tree with many branches. The cell body is the main part, while dendrites are branch-like extensions.
- Apical dendrites mainly encode abstract rules.
- The cell body represents external auditory sensory input.
These signals undergo nonlinear integration within the neuron, resulting in a decision output that drives behavioral choice.
Learning and Memory
The experiments showed that the brain slowly strengthens neuronal synapses and stores rules permanently after multiple verifications.
AI vs. Brain Mechanisms
Although AI uses context and data to adjust answers, its mechanisms differ significantly from the brain. AI develops capabilities through training on data, while the brain combines experiential learning, internal states, and real-time information processing. Even a single neuron has a complex internal computational structure.