Proposal Description

Gender disparities continue to influence educational pathways, scientific careers, and the credibility assigned to experts. As data-driven decision-making expands across education, public communication, and technological systems, statistical methods play a crucial role in identifying and addressing these inequalities. This session brings together four contributions that apply innovative statistical approaches to study gender dynamics in STEM education, evaluation, public trust, and AI-driven environments.
The first presentation investigates gender as a mediator in the academic trajectories of students at a major university. Using graphical models to capture complex dependency structures, the study shows how the relationships among performance indicators, time-to-degree, and other variables differ between male and female students. By proposing tailored methods to detect gender-specific mediation effects, the contribution offers insights into structural factors that shape disparities in education.
The second contribution focuses on hidden gender bias in STEM evaluations using Differential Item Functioning (DIF) techniques. By comparing item responses from men and women with equivalent latent abilities or attitudes, the study identifies items that systematically favor one gender. This analysis highlights the importance of psychometric fairness and provides methodological guidance for improving the design and interpretation of surveys and assessments in STEM fields.
A third presentation expands the field of inquiry beyond educational settings to examine cultural and intergenerational determinants of trust in female scientists across five countries. Using a large-scale randomized survey experiment involving adults and children, the study reveals cross-national differences in how respondents update their beliefs depending on the gender of the information provider. While trust in science remains high overall, attitudes toward female scientists are far from uniform, even in countries with strong gender-egalitarian norms. The inclusion of parent–child dyads provides rare evidence of the intergenerational transmission of scientific trust and highlights how gendered perceptions are learned, reproduced, and potentially transformed across generations.
Finally, the session includes a contribution turns to the domain of financial decision-making, investigating how gender interacts with behavioral data in credit assessment systems. Using detailed information on building-savings contracts, the study applies fairness-aware statistical methods to examine whether behavioral indicators such as saving intensity or discipline function differently across genders in predicting creditworthiness. By identifying where gendered patterns may enter algorithmic models, the work contributes to broader debates on ethics, fairness, and the responsible use of AI in financial governance. used to evaluate and mitigate gender bias in data-driven systems. As AI increasingly influences decisions in education, research, and public life, ensuring fairness and transparency becomes a central methodological and ethical challenge.
Together, these presentations demonstrate how statistical reasoning can illuminate gender inequalities across different settings and support the development of more equitable scientific, educational, and technological environments.