1. Introduction
Search Engines:
Google historically mirrored stereotypes, showing white males for "doctor" and "professor," white females for "nurse," and black males for "labor." Searches like "black girl" led to inappropriate results, highlighting algorithmic biases in perpetuating societal stereotypes.
Generative A.I.:
A.I. art generators like Midjourney and Dall-E, akin to Google, often reinforce stereotypes. Despite Google's adjustments, these platforms, such as Midjourney's biased portrayal of "professor" as male, showcase the challenge of unbiased representation in A.I. art.
Stereotyping:
Google and A.I. platforms like Midjourney struggle with inclusive representation of Native Americans, often resorting to stereotypical, historical imagery. This perpetuates the notion of Native Americans as a "vanished race," underscoring the fine line between accurate depiction and reinforcing stereotypes.
2. The Algorithm
The Prompt:
Using "American Robot" and "Native American Robot" as prompts, Midjourney's AI generated distinct artworks, highlighting the stereotypical characteristics embedded in these terms. The difference, driven solely by the addition of "native," showcases AI's interpretation of cultural identities, using robots as neutral canvases to explore biases.
Visual Analysis:
The "Native American Robot" artwork features a sunset background, dried landscapes, and a Wild West setting, aligning with stereotypes of Native American environments. The robot, depicted with human-like proportions but exaggerated features and tribal decorations, lacks a head and has a single central eye, reinforcing stereotypes. Its attire and decorations, including feathers and tribal patterns, suggest a narrow, stereotypical view of Native American identity, reflecting AI's reliance on clichéd imagery to interpret cultural prompts.
3. The Process
The Process:
For better understanding, I have divided Midjourney into two components, the "A.I. Researcher" and the "A.I. Artist." The whole process then consists of 3 participants (the user, the A.I. Researcher, and the A.I. Artist) and eight steps.
The Keyword/Prompt: The prompt "Native American Robot" is interpreted by Midjourney based on historical and discriminatory associations found in cyberspace, reflecting biases embedded in language and digital content.
Database Scanning: Midjourney searches cyberspace for images tagged with "Native American" and related terms, encountering a mix of stereotypical and diverse representations.
Popularity Ranking: The algorithm ranks images based on frequency of reference and clicks, often prioritizing stereotypical representations due to their prevalence in digital and cultural narratives.
Search Results: The A.I. Researcher generates search results based on popularity, without filtering for stereotypes, providing raw material for the A.I. Artist influenced by prevailing biases.
Condensing Stereotypes: The A.I. Artist distills common characteristics from the search results, often reducing complex identities to simplified, stereotypical visual elements.
Reproducing Stereotypes: Midjourney combines stereotypes of "Native American," "American," and "Robot" based on cyberspace's associative use of these terms, reinforcing existing biases.
Producing Artifacts: Midjourney iteratively generates art, cycling through data analysis, stereotype condensation, and visual creation, producing variations that further populate its database.
Choosing the Desired Artifact: The user selects from four generated artifacts, influenced by both conscious and unconscious biases, contributing to the cycle of stereotype reinforcement.
4. Conclusion
Why Care?
The internet offers a platform to correct real-world mistakes, challenging stereotypes and enabling individual representation. Stereotyping not only affects misrepresented communities but also influences future generations, embedding unfair expectations and limiting self-expression.
How to Proceed?
We must leverage our position to influence A.I. ethics, promoting diverse participation in technology development. This ensures technologies recognize all human perspectives, counteracting biases. Mindful interaction with digital content is crucial for fostering inclusivity.
Conclusion:
A.I. art generators like Midjourney mirror societal biases due to complex interactions among social practices, consumer behavior, and content creation, highlighting our collective role in guiding technology towards more equitable representation.
Tools Practiced...
Midjourney
Credits...
Harshal Pilania, Researcher
Anne Ronan, Art Theory & History Professor
.jpeg)
