To develop human-level AGI systems, we need a cognitive modeling approach that prioritizes functional similarities with human cognition over merely structural ones. We have realized this through our Economic Attention Networks (ECAN), which serve as the attention-allocation and resource-regulation subsystem of the Hyperon architecture, originally implemented and tested within the OpenCog AGI framework.

Inside ECAN, working memory—as implemented in the PRIMUS cognitive architecture—is stored in the form of Atoms that maintain sufficiently high Short-Term Importance (STI) values. A novel enhancement to ECAN within Hyperon reframes this attention allocation as incompressible fluid dynamics with optimal control semantics. Instead of treating attention as something that diffuses randomly through the knowledge graph, this approach models it as a conserved fluid, where the flow is optimally steered toward goal-relevant regions.