A1521
Title: Lost in the supermarket: Individual and collective dynamics through the lens of innovation processes
Authors: Margherita Lalli - Scuola Normale Superiore di Pisa (Italy) [presenting]
Francesca Tria - Sapienza University of Rome (Italy)
Luca Pappalardo - ISTI-CNR (Italy)
Abstract: The focus is on an urn-based model aimed at characterizing the interplay between individual and collective exploration in a population of interacting agents. The model is inspired by the feedback dynamics of recommender systems, where users' choices and algorithmic retraining form a continuous loop that can generate unintended emergent effects. A central question, widely debated in computer science, is whether and under which conditions increasing individual-level diversity in choices may paradoxically reduce diversity at the collective scale. Conclusive results are particularly hindered by the opacity of real recommendation algorithms, which motivates the need for generative models capable of disentangling the underlying mechanisms. Our framework is empirically guided by a rich dataset of supermarket transactions featuring statistical regularities typical of innovation processes such as Zipf's, Heaps', and Taylor's laws, alongside systematic discrepancies between individual and aggregate purchases. To capture this, we build upon the Urn Model with Triggering, an extension of Polya's urn capable of reproducing such phenomenologies at a single scale. Unlike previous models of interacting urns, our formulation assumes neither a predefined network structure nor a nested user-product organization, offering a complementary perspective on the mechanisms driving the tension between individual exploration and collective regularities in consumption.
