Developing a Global Workspace Implementation of Episodic Memory

<p dir="ltr">Memory is a cornerstone of human cognition, enabling us to learn, reason, and navigate the world. Beyond recalling past experiences, the brain can also remember the temporal order in which these experiences occurred—a function known as serial order memory. Despite over a century of research, our understanding of this capacity remains incomplete. The Reconstructive Theory of Self is a recent computational model that explains serial order memory by extending the previously well-established MINERVA 2 model of episodic memory, or memory for specific instances of past experience, as well as associative memory, or the ability to learn relationships between different experiences or aspects of experience. Although this extended model successfully simulates several key serial order memory phenomena, it makes unrealistic theoretical assumptions about working memory—the system responsible for temporarily holding and manipulating information ‘in mind’. This thesis introduces an alternative model that builds on the same episodic and associative memory principles but incorporates a more realistic set of assumptions to explain serial order memory. Throughout the thesis, we describe the development of this model, beginning with its theoretical foundations and predecessor models, and justifying the additional assumptions made. We then outline the computer programming methodologies used to build the model and test its performance by simulating human experiments. Finally, we interpret the simulation results and discuss the strengths and limitations of the model in terms of its core theoretical assumptions, comparing our approach with other modelling approaches and contextualising our findings within the broader cognitive science and neuroscience literature.</p>