Pursuing Artificial General Intelligence by Leveraging the Knowledge Capabilities of ACT-R

Intelligence is a multifaceted phenomenon which makes trying to capture its very essence a slippery task. In this paper, we commit to a hybrid notion of intelligence, conceived as the combination of cognitive operations and knowledge resources that leads to purposeful behavior. Accordingly, this paper describes an artificial system that benefits from both mechanism–centered and knowledge–centered approaches. In particular, the system integrates the ACT-R cognitive architecture with SCONE, a knowledge-based system for ontological reasoning, to combine ACT-R’s subsymbolic cognitive mechanisms with SCONE’s knowledge representation and inference capabilities. We apply the hybrid system to computationally approximate human intelligent behavior in a task of visual recognition.