Victoria University
Can Bias Embedded in Image-Generative AI Systems Influence Public Perception?
Abstract
dc:description.abstractIn a time of fast and unprecedent technological change, this research critically examines how image generative systems such as Midjourney reproduce and amplify racial bias. Based in Australia and centring Critical Race Theory (CRT), this thesis interrogates the intersections of race, representation, and technology, highlighting how visual outputs sustain Eurocentric ideals and structural inequalities. Rather than focusing on the mechanics of AI, this thesis foregrounds the permanence of racism, antiblackness, and evolving notions of Australianness, examining how these are encoded into datasets and reflected in generated imagery. Through semi-structured interviews with six Australian university students, it explores lived experiences that reveal tensions between dominant visual narratives and personal realities, from depictions of white femininity as ideal beauty to portrayals of Aboriginal men shaped by colonial and racist stereotypes. This work does not merely highlight the failings of artificial intelligence as we have it today; it is an invitation to engage with critical race theory as well as a call upon developers, policymakers, and educators to interrogate the structures that allow these biases to persist.
Degree
thesis:*- Name dc:type.qualificationname
- other
- Level dc:type.qualificationlevel
- rmaster
- Grantor dc:publisher.institution
- Victoria University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Oliveira Andrade de Melo, Guido
Subjects
dc:subject × 3Rights
- Language dc:language
- en