Words, Webs, and Time: A Quantitative Analysis of Autobiographical Memory Transformation

The quantification of memory processes is an ever-evolving area in neuroscience. Memories are reconstructive, with the passage of time and the act of recall taking part in their transformation. Emotionality modulates this trajectory, altering a memory's internal organization and its susceptibility to structural reorganization over time. This modulatory role is particularly relevant for autobiographical memories, the ability to recall personal experiences. Yet, because these memories are inherently variable and subjective, they have long posed a challenge to systematic, quantitative analysis. This study aimed to understand the role of the passage of time and recall in autobiographical memory, and whether their effect differs with the emotional profile of the recalled event. To achieve this, we compared autobiographical narratives of contrasting emotional valence, Argentina's loss to Saudi Arabia and its championship win at the 2022 World Cup, each narrated by the same participant at two temporal instances. Interviews were characterized through NLP (POS tagging, sentiment analysis, semantic embeddings) and graph theory metrics. Using this feature set, a support vector machine model was trained to distinguish narratives from the first temporal instance from those of the second. Through the resulting feature importance, we characterized how autobiographical memories are transformed over the passage of time, and whether this transformation differs with emotionality.