Description
Speaker 1: Dengdeng Yu
Assistant Professor of Statistics, Department of Management Science and Statistics
Carlos Alvarez College of Business, UTSA
“Word Embeddings via Causal Inference: Gender Bias Reducing and Semantic Information Preserving”
With widening deployments of natural language processing (NLP) in daily life, inherited social biases from NLP models have become more severe and problematic. Previous studies have shown that word embeddings trained on human-generated corpora have strong gender biases that can produce discriminative results in downstream tasks. Previous debiasing methods focus mainly on modeling bias and only implicitly consider semantic information while completely overlooking the complex underlying causal structure among bias and semantic components. To address these issues, we propose a novel methodology that leverages a causal inference framework to effectively remove gender bias. The proposed method allows us to construct and analyze the complex causal mechanisms facilitating gender information flow while retaining oracle semantic information within word embeddings. Our comprehensive experiments show that the proposed method achieves state-of-the-art results in gender-debiasing tasks. In addition, our methods yield better performance in word similarity evaluation and various extrinsic downstream NLP tasks.
Biography
Dengdeng Yu is an assistant professor in the Department of Management Science and Statistics at the Carlos Alvarez College of Business at The University of Texas at San Antonio (UTSA). His research focuses on high-dimensional data analysis, functional data analysis, statistical machine learning, causal inference, quantile regression, neuroimaging data analysis and imaging genetics. Yu earned his Ph.D. and M.Sc. in Statistics from the University of Alberta and holds an M.Sc. in Finance from the University of Ulm and a B.Sc. in Information and Computing Science from Southeast University. Prior to joining UTSA, Yu was an Assistant Professor at the University of Texas at Arlington and held postdoctoral positions at the University of Toronto and the Canadian Statistical Sciences Institute (CANSSI).
Speaker 2: Isil Koyuncu
Assistant Professor of Management Science, Department of Management Science and Statistics
Carlos Alvarez College of Business, UTSA
“A Risk-Averse Stochastic Optimization Framework for Fuel-tax Increase and Choice of Indexing for Revenue Adequacy”
Transportation funding based on fuel taxes is becoming unviable due to fuel-efficient vehicles, inflation, rising construction costs, and alternative fuels. Indexing fuel taxes offers a potential solution but involves challenges in choosing methods that account for risks and uncertainties in future revenue streams. This study presents a framework to evaluate fuel tax indexing methods with risk considerations and their impact on one-time tax rate increases to achieve targeted revenues over time. The problem is modeled as a two-stage stochastic optimization, where the first stage determines the index choice, and the second stage calculates the optimal initial tax increase. Using Sample Average Approximation, a Monte Carlo method, the study demonstrates convergence to expected values with large samples. A risk-averse version is also analyzed with chance constraints, showing that increased risk aversion leads to higher initial tax increases and shifts in preferred indexing methods. Sensitivity analyses on crude oil prices, alternative fuel adoption rates, and gas price elasticity provide robust insights. Findings highlight how risk acceptance impacts tax policies, assisting policymakers in selecting effective fuel tax indexing methods to ensure revenue adequacy while managing user burdens. This study offers a practical tool for policymakers to design adaptive and sustainable transportation funding strategies.
Biography
Isil Koyuncu is an assistant professor in the Department of Management Science and Statistics at The University of Texas at San Antonio (UTSA). Her research interests lie at the intersection of transportation operations and sustainability. Koyuncu particularly focuses on optimizing decisions surrounding the adoption of alternative fuel vehicles, the routing of alternative fuel vehicles and electric vehicle charging station capacity design. She formulates problems using discrete and stochastic models and develops metaheuristics and matheuristics to solve large combinatorial optimization problems. Additionally, Koyuncu is interested in other application areas of transportation operations, such as airline operations and flight scheduling, emergency medical service routing and revenue generation strategies for transportation infrastructure under uncertainty. Prior to joining the department in Fall 2020, she obtained her Ph.D. in Operations Management from the University of Alabama. She holds M.Sc. and B.Sc. degrees in Industrial Engineering
Featuring: Dengdeng Yu, assistant professor of statistics & Isil Koyuncu, assistant professor of management science
Location
BB 1.01.20L, Multipurpose Room
Category:
Campus Events