Gender Inequality in the Artificial Intelligence Workforce: Longitudinal Global Evidence of Pipeline Leakage
Year: 2026
Type: dataset
Abstract: Description Replication Package: Responsible Innovation and Gender Inequality in Artificial Intelligence: Workforce Inclusion and Social Sustainability This replication package accompanies the study “Responsible Innovation and Gender Inequality in Artificial Intelligence: Workforce Inclusion and Social Sustainability.” The study examines gender inequality in the artificial intelligence (AI) and data science workforce using four consecutive waves of the Kaggle Machine Learning and Data Science Survey (2018–2021), comprising 87,945 respondents from more than 50 countries. The analysis examines gender differences across workforce participation, age, educational attainment, occupational roles, coding and machine learning experience, annual compensation, industry sectors, and geographic representation. The study is guided by responsible innovation theory, with inclusion as the principal responsible innovation dimension implicated by the empirical findings. STEM Pipeline Theory is used as an explanatory mechanism for understanding how gender disparities may accumulate across successive stages of workforce participation. The study also considers the implications of these patterns for social sustainability, digital inequality, and the Sustainable Development Goals (SDGs). The replication package contains nine Python 3.9 analytical scripts, figures supporting the empirical analysis (Figures 1–7), a README file providing documentation and reproduction instructions, and a requirements file listing the software dependencies. The underlying survey datasets were obtained from the publicly available Kaggle Machine Learning and Data Science Survey repositories for 2018, 2019, 2020, and 2021 and are not redistributed in this package. Instructions for accessing the original datasets are provided in the accompanying documentation. The study uses publicly available, anonymised secondary survey data. The four survey waves are voluntary, non-probability samples of the Kaggle community and should not be interpreted as representative samples of the global AI workforce. The analyses are descriptive and do not involve individual-level longitudinal matching or causal inference. The package is provided to support transparency, reproducibility, and verification of the analytical procedures and reported findings.
Authors: KAMA, PRISCILLIA ADAKU, GÜLBAY, ALPER
Institution: Girne American University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204896573