Detecting Bias and Fairness Trade-offs in Machine Learning Models Using Real-World Datasets
Research Field
Artificial Intelligence | Machine Learning | Data Science | Ethics in AI
Location
Remote Only
Project Overview
As machine learning systems are increasingly deployed in high-stakes settings, ensuring fairness and minimizing algorithmic bias has become a critical challenge. In this project, a student will work with real-world datasets to examine how common machine learning models can unintentionally encode or amplify bias across demographic groups.
The project introduces students to applied AI research at the intersection of technology, ethics, and policy, reflecting challenges actively explored in industry research environments.
Student Responsibilities
The student will:
Analyze real-world structured datasets with demographic attributes
Train and evaluate baseline machine learning models (e.g., logistic regression, decision trees)
Apply standard fairness metrics to assess model performance across groups
Explore trade-offs between accuracy, interpretability, and fairness
Summarize findings in a formal research paper suitable for competition and presentation
Prior coding experience is helpful but not required; motivated students will be supported in learning foundational tools.
Skills Gained
Practical machine learning model development
Data preprocessing and exploratory data analysis
Introduction to algorithmic fairness metrics and evaluation
Research documentation and technical communication
Exposure to industry-relevant AI research practices
Time Commitment
Approximately 6–10 hours per week, depending on depth of exploration and project scope.
Ideal Student Profile
Strong interest in computer science, AI, or data science
Comfortable with mathematics and logical reasoning
Curious about ethical and societal implications of technology
Able to work independently with structured mentorship
