Human memory can store diverse types of information, ranging from basic facts and specific dates to intricate narratives. A key goal in cognitive psychology has been understanding how meaningful stories are retained in our minds.
A collaborative research team from the Institute for Advanced Study, Emory University, and the Weizmann Institute of Science recently constructed a model that uses mathematical objects called "random trees" to explain how humans represent and store narrative memories. Their findings have been published in Physical Review Letters, introducing a novel framework based on mathematics, computer science, and physics for studying human memory processes.
Physical Review Letters"We aimed at addressing the challenge of creating a mathematical theory for how humans remember meaningful material like narratives," Misha Tsodyks, the senior author of the paper, told Medical Xpress. "There's a commonly held belief that the complexity of narratives makes this impossible, but our research suggests there are statistical patterns in how people recall stories that can be predicted by simple principles."
To develop their model using random trees, Tsodyks and his colleagues conducted online experiments on a large group of participants utilizing platforms such as Amazon and Prolific. They presented the subjects with narratives from W. Labov's work and then analyzed the recalled stories to validate their theory.
"We employed a collection of spoken narratives recorded by linguist W. Labov in the 1960s," Tsodyks described. "Modern AI tools were necessary for analyzing this vast amount of data, particularly large language models (LLMs).
"In our study, we found that people often summarize substantial parts of a narrative into single sentences rather than recalling individual events. This led us to the idea that narratives are represented in memory as trees, with nodes closer to the root representing abstract summaries of larger episodes."
The team hypothesized that when individuals first encounter or read a story, they create trees in their minds where each tree has a unique structure reflecting different understandings of the narrative.
"We conceptualized our model as an ensemble of random trees with specific characteristics," said Tsodyks. "The advantage of this model is that it can be solved mathematically, allowing us to directly test its predictions with actual data, which we did successfully. The key innovation of our random tree memory and recall model lies in the assumption that any meaningful material is fundamentally represented in the same manner.
"This research has broader implications for human cognition, as narratives play a crucial role in how we perceive individual experiences and social or historical processes."
The team highlights the potential of mathematical approaches and AI-based techniques to further explore how humans store meaningful information. Their plan is to extend their theory to other narrative forms, such as fictional stories.
"In future research, a more ambitious goal would be to obtain additional evidence for the tree model," added Tsodyks. "This might involve developing new experimental protocols beyond simple recall, including brain imaging while participants engage in narrative comprehension and recall."
ad-free