Browsing by Author "Morrison, Don"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    An Integrated Computational Framework for the Neurobiology of Memory Based on the ACT-R Declarative Memory System
    (Springer Science and Business Media LLC, 2023-12-28) Stocco, Andrea; Rice, Patrick; Thomson, Robert; Smith, Briana; Morrison, Don; Lebiere, Christian
    Memory is a complex process that spans multiple time-scales and stages, and, as expected, involves multiple brain regions. Traditionally, computational models of memory are either too abstract (Shiffrin & Steyvers, 1997) to be meaningfully connected to a biological substrate, or, when explicitly connected, are narrowly focused on one specific region and process (Blum & Abbott, 1996; Weber et al., 2017). By contrast, a comprehensive model of memory with a plausible neural interpretation would be extremely valuable to drive further research in memory function and dysfunction. In this paper, we attempt to fill in this gap by providing a detailed biological analysis of ACT-R’s declarative memory system. This system, developed over four decades, has evolved into a consistent framework that describes how memories are formed, retrieved, forgotten, mistaken, and merged. Building on existing mappings between some components and their biological counterpart, as well as the existing literature, this paper provides a comprehensive view of how the framework’s various computations map onto different brain regions, their network dynamics and functional connectivity, and biological structure. We also show that these mappings provide further insights and explanations for puzzling findings in the memory disorders literature. Finally, we outline the remaining gaps (such as the transition from episodic to semantic memory) and how they could be addressed by future research and modeling efforts.
  • Thumbnail Image
    ItemOpen Access
    Strong CogSci: Guidance from Cognitive Science On the Design of a Test of Artificial Intelligence
    (AAAI, 2015) Lebiere, Christian; Bothell, Dan; Morrison, Don; Oltramari, Alessandro; Martin, Michael; Romera, Oscar; Thomson, Robert; Vinokurov, Jerry
    We propose a test of human-like intelligence inspired by constraints from cognitive science on unified theories of cognition. The most salient characteristics of this test include the capacity for open-ended, generative behavior, the ability to learn and adapt in real time in complex environments, and the ability to interact productively with other human and artificial agents. The proposed test environment is Minecraft, a virtual world reflecting many of the key characteristics of real world embodiments. The core task is to build assemblies of varying complexity and purposes in cooperation with human or artificial entities.