Contributed Session 7: Statistical Approaches to Gender & Education Unbalance

Violent or non-violent conflict poses significant concerns in various contexts, including public administration, where disputes can undermine governance and organizational efficiency. This study builds upon existing conflict prediction research focusing on non-violent conflicts. By incorporating gender as a critical yet under-explored variable, this research develops a methodological framework for analyzing and forecasting non-violent conflicts in public administration. Using natural language processing techniques and socio-economic datasets, we propose integrating gender-related variables, such as representation in leadership roles and equality indices, to explore their impact on conflict dynamics. Like logistic regression, traditional statistical models are complemented by ensemble methods – such as Random Forests – to capture complex relationships and feature importance. Although addressing data and computational challenges, this research aims to improve our understanding of the interaction between gender dynamics and conflict, contributing to more effective conflict management strategies and governance reforms.

Artificial intelligence (AI) is increasingly used in medical applications, from diagnosis to treatment recommendations. However, biases in AI models can increase existing disparities. We explore fairness measures to mitigate biases due to gender and sex in AI-driven medicine. We define non-discrimination criteria and introduce statistical fairness metrics applicable to medical AI at different stages of the machine learning cycle to monitor and reduce discriminatory effects.

Women’s participation in academia still faces significant disparities. In Italian universities, only about 26% of women reach the position of full professor, despite equal (or even higher in certain fields) graduation and PhD rates. This study investigates the gender gap in the average time spent in academic positions, in the time required for career advancement, and in the probability of achieving higher roles in the context of Italian academia and taking scientific productivity and competition into account. To the best of our knowledge, this is among the first contributions using Markov models to quantify the gender gap in career progression.

This study explores the relationship between students’ digital competencies, as measured by the ICILS 2018 Computer and Information Literacy (CIL) scale, and their academic performance in Italian, Mathematics, and English, assessed through the INVALSI 2018 tests. Using data from grade 8 students, a precise merge was conducted using common identification codes. The analysis included exploratory data examination, correlation analyses, and regression models. Results indicate a significant positive association between CIL scores and achievements in all three subjects, suggesting that students with higher digital skills tend to perform better academically. However, the models used are preliminary and descriptive. Future research should incorporate multilevel models to account for the hierarchical structure of educational data and better capture the impact of school-level factors on student performance.