← Back to Publications
Authors:
Argues that a cognitive architecture for cumulative learning must provide the information and control structures that let agents learn incrementally and autonomously, continuously relating sensory information to their goals across a perception–cognition stack. It examines the role of attention in managing this process across levels of abstraction, drawing on dual-process (System-1 / System-2) framings for neurosymbolic perception and cognition.
An attention-centric neurosymbolic source for the ECAN Deep Dive — it treats attention as the control mechanism relating perception to goals in a resource-bounded cognitive architecture, complementing ECAN's economic attention-allocation account. (NARS/ONA-adjacent; authors include Pei Wang and Patrick Hammer.)