CAIO research project
Publications
Recommendation Feedback Loops Shape Urban Income Segregation and Venue Visitation
arXiv:2610.41911 (2026)
Ths articles shows how location-based recommender systems create a human-AI feedback loop that inadvertently increases income segregation in cities and causes over a fifth of local venues to lose all visits. While standard accuracy metrics fail to capture these long-term urban impacts, implementing a simple re-ranking intervention that boosts less popular venues successfully reduces both segregation and venue disappearance without compromising user comfort.
The traffic concentration effects of urban navigation services
Nature Communications 17, 8548 (2026)
This study investigates how the widespread adoption of urban navigation services can concentrate traffic on a limited number of roads, affecting congestion, emissions, and the resilience of urban mobility systems.
Algorithms vs Diversity
Aspenia 1 "The new scramble", (2026)
This article discusses the paradox that characterizes algorithmic recommendation systems: they increase the variety of individual choices but concentrate collective attention on a few popular options.
Learning by Surprise: Adaptive Mitigation of Model Collapse in Large Language Models
ACM Transactions on Intelligent Systems and Technology, (2026)
This study provides a model-centric account of collapse in LLMs repeatedly trained on AI-generated content, linking it to concentrated token probabilities, declining common-sense reasoning, and low-perplexity training data. It proposes a perplexity-based filtering strategy that mitigates collapse without requiring access to human-authored data.