arXiv subjects

Dino Pedreschi

Publications and source records attributed to Dino Pedreschi.

At least 19 recordsLinked to original sources

arXiv2026

The traffic concentration effects of urban navigation services

The collective impact of navigation services remains unclear: while often beneficial to individual drivers, they can unintentionally reshape urban traffic patterns. We simulate their impact in Florence, Milan, and Rome (Italy), integrating GPS data, road networks, and route recommendations from leading providers. We identify a concentration effect: as adoption increases, route diversity declines, and traffic and emissions converge onto fewer roads. At full adoption, route diversity decreases by up to 14% compared to a baseline where recommendations are ignored. Moreover, navigation services reduce CO2 emissions at low adoption levels, but these benefits diminish, disappear, or even reverse beyond a city- and service-specific threshold. We replicate our experiments in an abstract setting, obtaining results consistent with those observed in real-world cities.

cs.MA↗

arXiv2026

Explainable AI for Cancer Drug Response Prediction: Beyond Univariate Feature Attributions

Predicting cancer drug response from transcriptomic profiles is a cornerstone of precision oncology, yet the scientific value of machine learning models hinges not solely on predictive accuracy, but also on their capacity to generate reliable biological insights. Current explainability approaches in this setting are computationally costly, lack robustness, and reduce complex drug response to univariate gene importance scores, overlooking the coordinated gene activity that drives sensitivity and resistance. In this work, we present ILLUME+, a scalable post-hoc explainability framework that moves beyond single-gene assessments to capture multiple, complementary forms of explanation. Integrated into our end-to-end pipeline, ILLUME+ produces more stable gene importance scores than existing baselines, recovers established drug-gene associations and mechanisms of action, and enables AI-assisted hypothesis generation to uncover novel interaction-driven molecular signals in cancer biology.

cs.LG↗

arXiv2026

Learning by Surprise: Adaptive Mitigation of Model Collapse in Large Language Models

As AI-generated content increasingly populates the web, generative AI models are at growing risk of being trained on their own outputs, a process known as AI autophagy. This feedback loop has been shown to induce model collapse, typically characterized by a loss of diversity in generated content. However, existing work offers a limited understanding of this phenomenon and relies on mitigation strategies that assume access to human-authored data. In this paper, we conduct extensive simulations across multiple datasets and LLMs to address key gaps in the study of model collapse. First, we introduce model-intrinsic measures based on next-token probability distributions, showing that model collapse corresponds to an increasing concentration of probability mass on a small set of tokens. Second, we demonstrate that model collapse is also associated with a loss of common sense, as measured by a decline in commonsense inference accuracy. Third, we identify perplexity (a measure of model "surprise") as a key driver of collapse: fine-tuning on the least "surprising" documents leads to more severe degeneration. Building on this insight, we propose a perplexity-based filtering strategy that prioritizes high-surprise documents during fine-tuning. Unlike existing approaches, our method does not require distinguishing between human-authored and AI-generated content. Across datasets and LLM families, this strategy consistently mitigates model collapse, achieving performance comparable to, and in some cases better than, human-data baselines, while substantially reducing the concentration of next-token probabilities. Overall, our results provide a unified, model-centric understanding of model collapse and suggest practical, scalable strategies for training generative AI systems in increasingly synthetic environments.

cs.CL↗

arXiv2026

Joint Effects of Recommender Systems and Network Structure on the Visibility of Content and Creators

Social media algorithms allocate users' visibility by ranking content within their social networks. Yet, how recommendation logic and network structure jointly shape visibility across content and creators remains largely understudied. In this work, we tackle this question through agent-based simulations using YSocial, a social media virtual twin, in which agents interact under 7 recommendation strategies and 2 network topologies. We find that recommender logic sets the visibility regime: popularity creates a reinforcement loop in which early reactions increase later exposure, concentrating visibility on a small subset of content and limiting creator visibility to those whose content enters this loop, while collaborative filtering distributes visibility broadly across the active catalogue and user base. When the follower graph shapes candidate selection, network structure changes the direction of inequality: under popularity ranking, creator-level concentration becomes comparable to global popularity, but visibility is systematically redirected toward creators who are already socially popular. Network topology modulates the magnitude of these effects without changing their qualitative ordering. These results show that visibility allocation should be evaluated across content, creators, network position, and temporal reinforcement, and that controlled simulations can help test how feed design distributes visibility before deployment.

cs.SI↗

arXiv2026

Explanations Go Linear: Post-hoc Explainability for Tabular Data with Interpretable Meta-Encoding

Post-hoc explainability is essential for understanding black-box machine learning models. Surrogate-based techniques are widely used for local and global model-agnostic explanations but have significant limitations. Local surrogates capture non-linearities but are computationally expensive and sensitive to parameters, while global surrogates are more efficient but struggle with complex local behaviors. In this paper, we present ILLUME, a flexible and interpretable framework grounded in representation learning, that can be integrated with various surrogate models to provide explanations for any black-box classifier. Specifically, our approach combines a globally trained surrogate with instance-specific linear transformations learned with a meta-encoder to generate both local and global explanations. Through extensive empirical evaluations, we demonstrate the effectiveness of ILLUME in producing feature attributions and decision rules that are not only accurate but also robust and computationally efficient, thus providing a unified explanation framework that effectively addresses the limitations of traditional surrogate methods.

cs.LG↗

arXiv2026

The Diversity Paradox revisited: Systemic Effects of Feedback Loops in Recommender Systems

Recommender systems shape individual choices through feedback loops in which user behavior and algorithmic recommendations coevolve over time. The systemic effects of these loops remain poorly understood, in part due to unrealistic assumptions in existing simulation studies. We propose a feedback-loop model that captures implicit feedback, periodic retraining, probabilistic adoption of recommendations, and heterogeneous recommender systems. We apply the framework on online retail and music streaming data and analyze systemic effects of the feedback loop. We find that increasing recommender adoption may lead to a progressive diversification of individual consumption, while collective demand is redistributed in model- and domain-dependent ways, often amplifying popularity concentration. Temporal analyses further reveal that apparent increases in individual diversity observed in static evaluations are illusory: when adoption is fixed and time unfolds, individual diversity consistently decreases across all models. Our results highlight the need to move beyond static evaluations and explicitly account for feedback-loop dynamics when designing recommender systems.

cs.IR↗

arXiv2025

A survey on the impacts of recommender systems on users, items, and human-AI ecosystems

Recommendation systems and assistants (in short, recommenders) influence through online platforms most actions of our daily lives, suggesting items or providing solutions based on users' preferences or requests. This survey systematically reviews, categories, and discusses the impact of recommenders in four human-AI ecosystems -- social media, online retail, urban mapping and generative AI ecosystems. Its scope is to systematise a fast-growing field in which terminologies employed to classify methodologies and outcomes are fragmented and unsystematic. This is a crucial contribution to the literature because terminologies vary substantially across disciplines and ecosystems, hindering comparison and accumulation of knowledge in the field. We follow the customary steps of qualitative systematic review, gathering 154 articles from different disciplines to develop a parsimonious taxonomy of methodologies employed (empirical, simulation, observational, controlled), outcomes observed (concentration, content degradation, discrimination, diversity, echo chamber, filter bubble, homogenisation, polarisation, radicalisation, volume), and their level of analysis (individual, item, and ecosystem). We systematically discuss substantive and methodological commonalities across ecosystems, and highlight potential avenues for future research. The survey is addressed to scholars and practitioners interested in different human-AI ecosystems, policymakers and institutional stakeholders who want to understand better the measurable outcomes of recommenders, and tech companies who wish to obtain a systematic view of the impact of their recommenders.

cs.IR↗

arXiv2025

A computational framework for quantifying route diversification in road networks

The structure of road networks impacts various urban dynamics, from traffic congestion to environmental sustainability and access to essential services. Recent studies reveal that most roads are underutilized, faster alternative routes are often overlooked, and traffic is typically concentrated on a few corridors. In this article, we examine how road network structure, and in particular the presence of mobility attractors (e.g., highways and ring roads), shapes the counterpart to traffic concentration: route diversification. To this end, we introduce DiverCity, a measure that quantifies the extent to which traffic can potentially be distributed across multiple, loosely overlapping near-shortest routes. Analyzing 56 diverse global cities, we find that DiverCity is influenced by network characteristics and is associated with traffic efficiency. Within cities, DiverCity increases with distance from the city center before stabilizing in the periphery, but declines in the proximity of mobility attractors. We demonstrate that strategic speed limit adjustments on mobility attractors can increase DiverCity while preserving travel efficiency. We isolate the complex interplay between mobility attractors and DiverCity through simulations in a controlled setting, confirming the patterns observed in real-world cities. DiverCity provides a practical tool for urban planners and policymakers to optimize road network design and balance route diversification, efficiency, and sustainability. We provide an interactive platform (https://divercitymaps.github.io) to visualize the spatial distribution of DiverCity across all considered cities.

physics.soc-ph↗

arXiv2025

Interpretable and Fair Mechanisms for Abstaining Classifiers

Abstaining classifiers have the option to refrain from providing a prediction for instances that are difficult to classify. The abstention mechanism is designed to trade off the classifier's performance on the accepted data while ensuring a minimum number of predictions. In this setting, often fairness concerns arise when the abstention mechanism solely reduces errors for the majority groups of the data, resulting in increased performance differences across demographic groups. While there exist a bunch of methods that aim to reduce discrimination when abstaining, there is no mechanism that can do so in an explainable way. In this paper, we fill this gap by introducing Interpretable and Fair Abstaining Classifier IFAC, an algorithm that can reject predictions both based on their uncertainty and their unfairness. By rejecting possibly unfair predictions, our method reduces error and positive decision rate differences across demographic groups of the non-rejected data. Since the unfairness-based rejections are based on an interpretable-by-design method, i.e., rule-based fairness checks and situation testing, we create a transparent process that can empower human decision-makers to review the unfair predictions and make more just decisions for them. This explainable aspect is especially important in light of recent AI regulations, mandating that any high-risk decision task should be overseen by human experts to reduce discrimination risks.

cs.LG↗

arXiv2025

AI, Meet Human: Learning Paradigms for Hybrid Decision Making Systems

Everyday we increasingly rely on machine learning models to automate and support high-stake tasks and decisions. This growing presence means that humans are now constantly interacting with machine learning-based systems, training and using models everyday. Several different techniques in computer science literature account for the human interaction with machine learning systems, but their classification is sparse and the goals varied. This survey proposes a taxonomy of Hybrid Decision Making Systems, providing both a conceptual and technical framework for understanding how current computer science literature models interaction between humans and machines.

cs.LG↗

arXiv2025

Language-Driven Opinion Dynamics in Agent-Based Simulations with LLMs

Understanding how opinions evolve is crucial for addressing issues such as polarization, radicalization, and consensus in social systems. While much research has focused on identifying factors influencing opinion change, the role of language and argumentative fallacies remains underexplored. This paper aims to fill this gap by investigating how language - along with social dynamics - influences opinion evolution through LODAS, a Language-Driven Opinion Dynamics Model for Agent-Based Simulations. The model simulates debates around the "Ship of Theseus" paradox, in which agents with discrete opinions interact with each other and evolve their opinions by accepting, rejecting, or ignoring the arguments presented. We study three different scenarios: balanced, polarized, and unbalanced opinion distributions. Agreeableness and sycophancy emerge as two main characteristics of LLM agents, and consensus around the presented statement emerges almost in any setting. Moreover, such AI agents are often producers of fallacious arguments in the attempt of persuading their peers and - for their complacency - they are also highly influenced by arguments built on logical fallacies. These results highlight the potential of this framework not only for simulating social dynamics but also for exploring from another perspective biases and shortcomings of LLMs, which may impact their interactions with humans.

cs.SI↗

arXiv2024

Human-AI Coevolution

Human-AI coevolution, defined as a process in which humans and AI algorithms continuously influence each other, increasingly characterises our society, but is understudied in artificial intelligence and complexity science literature. Recommender systems and assistants play a prominent role in human-AI coevolution, as they permeate many facets of daily life and influence human choices on online platforms. The interaction between users and AI results in a potentially endless feedback loop, wherein users' choices generate data to train AI models, which, in turn, shape subsequent user preferences. This human-AI feedback loop has peculiar characteristics compared to traditional human-machine interaction and gives rise to complex and often ``unintended'' social outcomes. This paper introduces Coevolution AI as the cornerstone for a new field of study at the intersection between AI and complexity science focused on the theoretical, empirical, and mathematical investigation of the human-AI feedback loop. In doing so, we: (i) outline the pros and cons of existing methodologies and highlight shortcomings and potential ways for capturing feedback loop mechanisms; (ii) propose a reflection at the intersection between complexity science, AI and society; (iii) provide real-world examples for different human-AI ecosystems; and (iv) illustrate challenges to the creation of such a field of study, conceptualising them at increasing levels of abstraction, i.e., technical, epistemological, legal and socio-political.

cs.AI↗

arXiv2023

Dense Hebbian neural networks: a replica symmetric picture of unsupervised learning

We consider dense, associative neural-networks trained with no supervision and we investigate their computational capabilities analytically, via a statistical-mechanics approach, and numerically, via Monte Carlo simulations. In particular, we obtain a phase diagram summarizing their performance as a function of the control parameters such as the quality and quantity of the training dataset and the network storage, valid in the limit of large network size and structureless datasets. Moreover, we establish a bridge between macroscopic observables standardly used in statistical mechanics and loss functions typically used in the machine learning. As technical remarks, from the analytic side, we implement large deviations and stability analysis within Guerra's interpolation to tackle the not-Gaussian distributions involved in the post-synaptic potentials while, from the computational counterpart, we insert Plefka approximation in the Monte Carlo scheme, to speed up the evaluation of the synaptic tensors, overall obtaining a novel and broad approach to investigate neural networks in general.

cond-mat.dis-nn↗

arXiv2023

Dense Hebbian neural networks: a replica symmetric picture of supervised learning

We consider dense, associative neural-networks trained by a teacher (i.e., with supervision) and we investigate their computational capabilities analytically, via statistical-mechanics of spin glasses, and numerically, via Monte Carlo simulations. In particular, we obtain a phase diagram summarizing their performance as a function of the control parameters such as quality and quantity of the training dataset, network storage and noise, that is valid in the limit of large network size and structureless datasets: these networks may work in a ultra-storage regime (where they can handle a huge amount of patterns, if compared with shallow neural networks) or in a ultra-detection regime (where they can perform pattern recognition at prohibitive signal-to-noise ratios, if compared with shallow neural networks). Guided by the random theory as a reference framework, we also test numerically learning, storing and retrieval capabilities shown by these networks on structured datasets as MNist and Fashion MNist. As technical remarks, from the analytic side, we implement large deviations and stability analysis within Guerra's interpolation to tackle the not-Gaussian distributions involved in the post-synaptic potentials while, from the computational counterpart, we insert Plefka approximation in the Monte Carlo scheme, to speed up the evaluation of the synaptic tensors, overall obtaining a novel and broad approach to investigate supervised learning in neural networks, beyond the shallow limit, in general.

cond-mat.dis-nn↗

arXiv2022

How Routing Strategies Impact Urban Emissions

Navigation apps use routing algorithms to suggest the best path to reach a user's desired destination. Although undoubtedly useful, navigation apps' impact on the urban environment (e.g., carbon dioxide emissions and population exposure to pollution) is still largely unclear. In this work, we design a simulation framework to assess the impact of routing algorithms on carbon dioxide emissions within an urban environment. Using APIs from TomTom and OpenStreetMap, we find that settings in which either all vehicles or none of them follow a navigation app's suggestion lead to the worst impact in terms of CO2 emissions. In contrast, when just a portion (around half) of vehicles follow these suggestions, and some degree of randomness is added to the remaining vehicles' paths, we observe a reduction in the overall CO2 emissions over the road network. Our work is a first step towards designing next-generation routing principles that may increase urban well-being while satisfying individual needs.

cs.CY↗

arXiv2022

Measuring the Salad Bowl: Superdiversity on Twitter

Superdiversity refers to large cultural diversity in a population due to immigration. In this paper, we introduce a superdiversity index based on the changes in the emotional content of words used by a multi-cultural community, compared to the standard language. To compute our index we use Twitter data and we develop an algorithm to extend a dictionary for lexicon-based sentiment analysis. We validate our index by comparing it with official immigration statistics available from the European Commission's Joint Research Center, through the D4I data challenge. We show that, in general, our measure correlates with immigration rates, at various geographical resolutions. Our method produces very good results across languages, being tested here both on English and Italian tweets. We argue that our index has predictive power in regions where exact data on immigration is not available, paving the way for a nowcasting model of immigration rates.

cs.SI↗

arXiv2021

The relationship between human mobility and viral transmissibility during the COVID-19 epidemics in Italy

In 2020, countries affected by the COVID-19 pandemic implemented various non-pharmaceutical interventions to contrast the spread of the virus and its impact on their healthcare systems and economies. Using Italian data at different geographic scales, we investigate the relationship between human mobility, which subsumes many facets of the population's response to the changing situation, and the spread of COVID-19. Leveraging mobile phone data from February through September 2020, we find a striking relationship between the decrease in mobility flows and the net reproduction number. We find that the time needed to switch off mobility and bring the net reproduction number below the critical threshold of 1 is about one week. Moreover, we observe a strong relationship between the number of days spent above such threshold before the lockdown-induced drop in mobility flows and the total number of infections per 100k inhabitants. Estimating the statistical effect of mobility flows on the net reproduction number over time, we document a 2-week lag positive association, strong in March and April, and weaker but still significant in June. Our study demonstrates the value of big mobility data to monitor the epidemic and inform control interventions during its unfolding.

cs.SI↗

arXiv2021

Benchmarking and Survey of Explanation Methods for Black Box Models

The widespread adoption of black-box models in Artificial Intelligence has enhanced the need for explanation methods to reveal how these obscure models reach specific decisions. Retrieving explanations is fundamental to unveil possible biases and to resolve practical or ethical issues. Nowadays, the literature is full of methods with different explanations. We provide a categorization of explanation methods based on the type of explanation returned. We present the most recent and widely used explainers, and we show a visual comparison among explanations and a quantitative benchmarking.

cs.AI↗