This project explores the relationship between artificial intelligence (AI), identity, and misgendering. Many AI systems identify gender based on appearance, names, voices, or patterns in data. Because these systems rely on assumptions, they can reinforce stereotypes and sometimes misgender people. Instead of asking AI to identify or classify someone, this project asks a different question: What happens if AI refuses to assign gender? The project uses a Research through Design approach. The AI is designed to have a conversation with the participant instead of analyzing them. It does not classify, judge, or label the person. Instead, it listens to their responses and transforms them into an abstract visual interpretation. The conversation happens one question at a time to encourage reflection. The questions focus on imagination, emotions, memories, contradictions, and personal experiences rather than gender or appearance. These answers become the basis for the final generated image.
RESEARCH CONTEXT
RESEARCH METHODS
We began by exploring the relationship between gender and bias in artificial intelligence (AI). Our research included a range of sources and methods, such as articles, blogs, existing AI-biased examples, and social experiments, which helped us understand how AI systems represent, interpret, and reproduce gender-related patterns.
Through this research, we identified recurring issues within AI systems: they often rely on stereotypical representations, repeat existing patterns, lack diversity in their datasets, and require highly specific and direct prompts to produce the intended results. Based on these findings, we narrowed our focus to „MISGENDERING“ as a key issue to investigate further.
We then explored culturally specific gender identities and third-gender datasets to understand how gender has been expressed and recognized across different cultural and historical contexts. This research became the foundation for our attempt to develop a performative AI system that moves beyond stereotypical and binary representations of gender.
Project Urgencies
Our research identified three main urgencies:
Diversity in AI: exploring how AI systems can become more diverse and empathetic in their understanding and representation of gender.
Self-expression: creating a platform where individuals can express aspects of themselves that may not fit conventional gender expectations.
Cultural awareness: addressing the lack of awareness around the diversity of gender identities and expressions across different cultural and historical contexts.
After several rounds of ideation and experimentation around how this research could be translated into a performative machine, we arrived at our final concept: “Gender / Self Expression with the Help of AI.”
The final system uses the cultural and gender-related datasets gathered during our research as a foundation for the AI, allowing the project to explore gender as something more complex than a fixed classification. Instead of asking the AI to determine what gender a person is, the project uses AI as a tool for exploring self-expression, identity, cultural context, and the unexpressed aspects of the self.
PROJECT IMPLEMENTATION
The project uses conversation as the main interaction. The AI asks five or six open-ended questions that encourage participants to think about different parts of themselves, such as the identities they perform, the parts they hide, their imagination, emotions, and personal symbols.
After the conversation, the AI generates an abstract image using visual elements like lines, shapes, colors, textures, materials, light, and movement. The generated image allows identity to be expressed without relying on gender or physical appearance.
The final image is not meant to show the participant's „true identity.“ Instead, it represents a moment of collaboration between the participant and the AI. By avoiding classification and focusing on reflection, the project presents an alternative way for AI to engage with identity and invites viewers to question whether machines should define who we are.
REFLECTIONS + QUESTIONS
This project challenges the assumption that gender can be identified from appearance. Clothing, hairstyle, makeup, body language, or facial features do not reveal a person's gender. Yet many AI systems are trained to make these assumptions, often reinforcing stereotypes and leading to misgendering.
Rather than asking AI to become better at identifying gender, this project asks whether it should identify gender at all. By removing classification from the interaction and replacing it with conversation, the project explores another possibility: AI that listens instead of labels.
For us many questions materialized, for example:
* In a world increasingly shaped by AI and data, why does knowing someone's gender expression matter?
* What would happen if AI stopped trying to identify gender altogether?
* How can we design AI systems that value self-expression over classification?
The project challenged, and continues to challenge, our owns mindsets on how important (and conversely not important) gender is to society. Let us clarify, its important for individuals that want to explore and express their own gender identities to do so but that does not mean governments and society as a whole should label, track, or otherwise regulate that expression.
REFERENCES + RESOURCES
Our datasets for the final project stems almost exclusively from texts written by individuals who identify as non-binary, third gendered, or two-spirit. They represent identity that relates to culture and creation; including drag. By doing this we are able to help combat or counter the effects and biases that traditional white-male-driven datasets create. The outcome of the project, the generated image, should represent something more freeing and less constricted by social norms. The below PDFs represent a selection of these invaluable resources.
