Foundations of human spatial problem solving.

Area of Science:

• Neuroscience
• Machine Learning
• Cognitive Science
• Computational Neuroscience

Background:

• The precise mechanisms by which the human brain achieves general problem-solving remain largely unknown.
• Existing research in machine learning and neuroscience has yet to fully elucidate this capability.

Purpose of the Study:

• To investigate the human brain's general problem-solving abilities through the lens of engineering control theory.
• To develop and validate a computational neural model capable of solving arbitrary problems.

Main Methods:

• Development of a computational neural model employing localist learning laws.
• Experimental design involving a multi-step task with dynamic start and end states for both the model and human participants.
• Utilizing human functional magnetic resonance imaging (fMRI) and representational similarity analysis (RSA).

Main Results:

• The computational model successfully found solutions to arbitrary problems, mirroring human task performance.
• fMRI and RSA data revealed that specific brain regions function as interacting components of a control theoretic system.
• Reinterpretation of the functional roles of the orbitofrontal cortex, hippocampus, basal ganglia, anterior temporal lobe, lateral prefrontal cortex, and visual cortex.

Conclusions:

• The findings suggest that engineering control theory provides a viable framework for understanding general intelligence in the brain.
• The study proposes a novel perspective on the neural basis of problem-solving and cognitive control.
• This research opens new avenues for developing artificial general intelligence (AGI).