Princeton Neuroscience Institute
Memory is organized in relational structures. When we remember something due to perceptual exposure or via replay, activation of the retrieved item itself activates other memories that are directly related to it, as well as memories that are multiple relations away. This cascade of activation renders the relations between memories malleable (known as plasticity). I will present a computational model, which demonstrates how the spread of activation in a memory network together with nonmonotonic plasticity can restructure memory. Using simulations, I will show that selective remembering, i.e. repeated re-activation of some but not other memories, can gradually change the structure of memory networks. This model can account for a range of memory phenomena, from retrieval-induced forgetting to conversational mnemonic convergence and consolidation. I will then discuss how these models, together with my previous work on prioritized offline replay can: (1) explain psychiatric phenomena in PTSD, (2) account for selective replay of negative memories and pruning of positive trajectories in rumination, and (3) provide a computational account of how cognitive therapy can leverage selective activation and forgetting to reorganize memory structures and alleviate symptoms. Finally, I will briefly discuss how this computational account might inform future interactive therapy apps that would interactively help users map and restructure the relational structure of their memories.