@conference{evaluation-gsi-conference-2026,
author = "Mu{\~n}oz L{\'o}pez, Sergio and Iglesias, Carlos A.",
abstract = "The proliferation of digital news platforms and sophisticated recommendation algorithms has fundamentally transformed information consumption patterns. These systems often prioritize engagement through preference reinforcement, creating echo chambers that limit exposure to diverse viewpoints and contribute to social polarization. While efforts to incorporate diversity into news recommendation systems have shown promise, existing solutions face several limitations, including data sparsity, limited semantic understanding, and difficulty in balancing relevance and diversity. These shortcomings highlight the need for innovative approaches that integrate relevance, diversity, and fairness in content selection. In this work, we explore the use of Large Language Models (LLMs) to address these challenges in digital news discovery. Specifically, we evaluate LLMs' capabilities for retrieving and selecting news content from the web, with a focus on identifying relevant articles that represent a range of perspectives. Our study proposes an agent-based system that systematically compares different LLMs through a set of quantitative and normative metrics, evaluating their effectiveness in content retrieval and perspective diversity. The results offer practical insights into the strengths and limitations of current LLM-based approaches for digital news discovery.",
address = "Marbella, Spain",
booktitle = "Proceedings of 18th nternational Conference on Agents and Artificial Intelligence (ICAART)",
comments = "CORE B",
keywords = "diversity;news recommender",
month = "March",
organization = "ICAART",
publisher = "SciTePress",
title = "{E}valuation of {D}iversity in {LLM}-{B}ased {N}ews {D}iscovery {T}hrough an {A}gent-{B}ased {S}ystem",
year = "2026",
}