Institution profile

Università Cattolica del Sacro Cuore

Academic institutioneurope · it
Official website
Research library10linked papers
Opportunities0open roles
Selected work

Representative Papers

Variable Selection in the Context of AI Fairness

Aug 04, 2026

This study addresses the tendency of conventional variable selection methods in AI to overlook ethical and societal contexts, often introducing implicit biases that compromise fairness across subpopulations. To counter this, the authors propose a novel variable selection framework that systematically integrates mathematical modeling, ethical analysis, and regulatory compliance. Innovatively embedding philosophical ethics and social awareness into the variable selection process, the approach advocates retaining sensitive and potentially relevant variables—rather than discarding them outright—to enable fine-grained fairness assessments. Empirical results demonstrate that this strategy effectively reduces disparities among subgroups and enhances model trustworthiness and compliance with regulatory frameworks such as the European Union’s Artificial Intelligence Act.

0 citationsRead paper

Structural Concentration in Weighted Networks: A Class of Topology-Aware Indices

Mar 23, 2026

Traditional concentration measures, such as the Herfindahl–Hirschman Index, neglect network topology and thus fail to capture the joint effect of weight distribution and interaction structure in weighted systems. This work proposes the Network Concentration Index (NCI)—a family of topology-aware concentration metrics based on normalized quadratic forms—that explicitly incorporates network structure into concentration measurement. The framework ensures normalization, invariance, and interpretability. Through graph-theoretic modeling, null-model comparisons, and multilayer network extensions, both theoretical analysis and simulations demonstrate that even with identical weight distributions, differing topologies can yield markedly distinct concentration levels. These findings confirm that the proposed framework effectively captures structural information overlooked by conventional indices.

0 citationsRead paper

Empathic Prompting: Non-Verbal Context Integration for Multimodal LLM Conversations

Oct 23, 2025

This study addresses the challenge of implicit nonverbal emotion perception by multimodal large language models (MLLMs) in high-emotion-sensitivity domains such as healthcare and education. We propose an empathy-aware prompting framework that operates without explicit emotion labels. Methodologically, real-time facial expression recognition—using a commercial API—is employed to extract user affective features, which are then integrated into the prompting pipeline of a locally deployed DeepSeek-LLM via lightweight contextual embeddings, requiring no architectural or training modifications. Key contributions include: (1) the first seamless integration of label-free emotion perception with LLM-based dialogue generation; (2) a modular design enabling straightforward extension to other nonverbal modalities (e.g., prosody, gesture); and (3) preliminary user evaluation (N=5) demonstrating consistent affective alignment in model responses and significant improvements in conversational naturalness and fluency.

0 citationsRead paper
Recent publications

Latest Papers

Variable Selection in the Context of AI Fairness

Aug 04, 2026

This study addresses the tendency of conventional variable selection methods in AI to overlook ethical and societal contexts, often introducing implicit biases that compromise fairness across subpopulations. To counter this, the authors propose a novel variable selection framework that systematically integrates mathematical modeling, ethical analysis, and regulatory compliance. Innovatively embedding philosophical ethics and social awareness into the variable selection process, the approach advocates retaining sensitive and potentially relevant variables—rather than discarding them outright—to enable fine-grained fairness assessments. Empirical results demonstrate that this strategy effectively reduces disparities among subgroups and enhances model trustworthiness and compliance with regulatory frameworks such as the European Union’s Artificial Intelligence Act.

0 citationsRead paper

Structural Concentration in Weighted Networks: A Class of Topology-Aware Indices

Mar 23, 2026

Traditional concentration measures, such as the Herfindahl–Hirschman Index, neglect network topology and thus fail to capture the joint effect of weight distribution and interaction structure in weighted systems. This work proposes the Network Concentration Index (NCI)—a family of topology-aware concentration metrics based on normalized quadratic forms—that explicitly incorporates network structure into concentration measurement. The framework ensures normalization, invariance, and interpretability. Through graph-theoretic modeling, null-model comparisons, and multilayer network extensions, both theoretical analysis and simulations demonstrate that even with identical weight distributions, differing topologies can yield markedly distinct concentration levels. These findings confirm that the proposed framework effectively captures structural information overlooked by conventional indices.

0 citationsRead paper

Empathic Prompting: Non-Verbal Context Integration for Multimodal LLM Conversations

Oct 23, 2025

This study addresses the challenge of implicit nonverbal emotion perception by multimodal large language models (MLLMs) in high-emotion-sensitivity domains such as healthcare and education. We propose an empathy-aware prompting framework that operates without explicit emotion labels. Methodologically, real-time facial expression recognition—using a commercial API—is employed to extract user affective features, which are then integrated into the prompting pipeline of a locally deployed DeepSeek-LLM via lightweight contextual embeddings, requiring no architectural or training modifications. Key contributions include: (1) the first seamless integration of label-free emotion perception with LLM-based dialogue generation; (2) a modular design enabling straightforward extension to other nonverbal modalities (e.g., prosody, gesture); and (3) preliminary user evaluation (N=5) demonstrating consistent affective alignment in model responses and significant improvements in conversational naturalness and fluency.

0 citationsRead paper