NBIA Colloquium: Misha Tsodyks

Title: Random Tree Model of Meaningful Memory

Abstract: How do we recall a story, a long narrative with its many twists and turns? Traditional studies of memory for meaningful narratives focus on specific stories and their semantic structures, but they do not address common quantitative features of memory recall across different narratives. I will introduce a statistical model ensemble of random trees to represent memorized narratives as hierarchies of key points, where each node is a compressed representation of its "leaves", which are the original narrative segments. Then the "branches" of the tree provide a rough outline of the story, on which the attached leaves of individual memories are arranged. Memory recall from this hierarchical representation is constrained by working memory capacity. The analytical solution of this model aligns with observations from large-scale narrative recall experiments. Additionally, the theory predicts that for sufficiently long narratives a universal scale-invariant limit emerges, where the fraction of a narrative summarized by a single recall sentence follows a distribution independent of narrative length.

Speaker: Misha Tsodyks (Weizmann Institute of Science)