Primary Supervisor
- Position
- Associate Professor
- Division / Faculty
- Faculty of Science
Overview
Machine Learning and Artificial Intelligence models require large-scale datasets to learn their behaviours. Many of these systems are trained as a decision or recommendation systems, predicated on the belief that they will be free from human bias and provide a rational, explainable decision-making process. There are two issues with these assumptions: first, many of these systems are "black-box" systems, and their behaviour and "reasoning" are opaque. Secondly, the focus of this project is that if these systems are trained using human data, then the existing human biases are "learned", defeating the purpose of a decision support system as an unbiased, transparent system. In this project, students will explore the issue of bias in training data and how it affects systems, whether it is possible to measure in trained systems or assess a priori, and what can be done to mitigate or address this issue.
Research engagement
This project will involve a review of the current literature on bias and bias in training data. It will require them to critically review the existing literature, assess the effectiveness of current methods and proposed methods, and synthesise potential solutions or strategies.
Research activities
They will work out of the space provided at CDS on the Gardens Point Campus. They will engagte in literature review a formulation of research questions and a plan of activities concluding with a report on their findings.
Research skills
Critical thinking, a knowledge of ML and AI training methods, and writing
Outcomes
A report of the students findings and a set of solutions or strategies for addressing bias from training data.
Skills and experience
A knowledge of basic statistical modelling and ML and AI models, as well as good oral and written communication skills,
Start date
2 November, 2026End date
19 February, 2027Location
GP Y Block
Keywords
Contact
A/Prof Gentry White
3138 1658
gentry.white@qut.edu.au