Mathematical Representation of Memory and Schema for Improving Human-Generative AI Interactions
Dimitrios P. Panagoulias, Persephone Papatheodosiou, Anastasios Bonakis, Dimitrios Dikeos, Maria K. Virvou, George A. Tsihrintzis · 2024
In this paper, we explore memory and Schema through the lens of mathematical representation. This approach aims to define more abstract themes about cognitive processes, to simulate and enhance interactions between Generative AI systems and humans. By doing so, humans can leverage their experiences and interactions in ways that improve the performance and personalization of the large language models they interact with. Furthermore, by formalizing these concepts using set theory, we can increase performance and base the accuracy of responses on streamlined historical conversations. This involves organizing data into logical clusters (memory) that utilize information (Schema) in a more comprehensive and structured manner. We define memory as a function of time and size, where the size of memory decays as time passes. The resolution of memory also fades, leading us to add more abstraction to our memory function. Memories with thematic resemblance are then structured into Schemas based on a maturity threshold influenced by criticality, emotional-user interaction, and mass, which is the measure of the summation of related memories.