Shawki The Engineer

3

min

Performative Male Detector with LLM Training

A website that detects if you are a "Performative Male"

Summary

Quick Facts

Team Members: Shawki Izzat, Joana Alves, Michelle Li
My Role(s): UI/UX Designer

Tools Used: Visual Studio Code, Loveable, Huggingface, Python, HTML/CSS/JS

Release: May 2024
Timeline: 4 months

Description

The Performative Male Detector is a playful web project created as the final submission for my Materiality in Machine Learning class. The site uses a custom-trained model (built with Python and Hugging Face image classification) to determine whether real-time camera footage aligns with the internet archetype of the “Performative Male.”

Quick Facts

Team Members: Shawki Izzat, Joana Alves, Michelle Li
My Role(s): UI/UX Designer

Tools Used: Visual Studio Code, Loveable, Huggingface, Python, HTML/CSS/JS

Release: May 2024
Timeline: 4 months

Description

The Performative Male Detector is a playful web project created as the final submission for my Materiality in Machine Learning class. The site uses a custom-trained model (built with Python and Hugging Face image classification) to determine whether real-time camera footage aligns with the internet archetype of the “Performative Male.”

Quick Facts

Team Members: Shawki Izzat, Joana Alves, Michelle Li
My Role(s): UI/UX Designer

Tools Used: Visual Studio Code, Loveable, Huggingface, Python, HTML/CSS/JS

Release: May 2024
Timeline: 4 months

Description

The Performative Male Detector is a playful web project created as the final submission for my Materiality in Machine Learning class. The site uses a custom-trained model (built with Python and Hugging Face image classification) to determine whether real-time camera footage aligns with the internet archetype of the “Performative Male.”

Inspiration

I wanted to do something fun and slightly unhinged for my final, which led me to the viral “Performative Male” discourse online. According to internet lingo, Performative Male is a slang term for a man who insincerely adopts specific, trendy, and often "soft" or progressive behaviors, aesthetics, and interests to appear more likable, safe, or sophisticated to women. This persona is considered a curated, superficial performance designed for social validation or dating success.

Common tropes include matcha, cardigans, wired earphones, and Sally Rooney novels (yes, that’s a real thing). I leaned into these clichés for my project.

I wanted to do something fun and slightly unhinged for my final, which led me to the viral “Performative Male” discourse online. According to internet lingo, Performative Male is a slang term for a man who insincerely adopts specific, trendy, and often "soft" or progressive behaviors, aesthetics, and interests to appear more likable, safe, or sophisticated to women. This persona is considered a curated, superficial performance designed for social validation or dating success.

Common tropes include matcha, cardigans, wired earphones, and Sally Rooney novels (yes, that’s a real thing). I leaned into these clichés for my project.

I wanted to do something fun and slightly unhinged for my final, which led me to the viral “Performative Male” discourse online. According to internet lingo, Performative Male is a slang term for a man who insincerely adopts specific, trendy, and often "soft" or progressive behaviors, aesthetics, and interests to appear more likable, safe, or sophisticated to women. This persona is considered a curated, superficial performance designed for social validation or dating success.

Common tropes include matcha, cardigans, wired earphones, and Sally Rooney novels (yes, that’s a real thing). I leaned into these clichés for my project.

Design Journey

Training Model

The first step was training a model to visually differentiate between “Performative Male” and “Non-Performative Male” archetypes. I collected and organized approximately 25–30 images per category by scraping the web, then fine-tuned a model using Hugging Face in Python. The goal was not accuracy in any scientific sense, but to explore how subjective cultural categories are translated into datasets.

Testing

Once trained, the model outputs a percentage score indicating how closely an image aligns with the performative male archetype. This numeric abstraction became the foundation for the project’s interaction and tone.

The Fun Part

Rather than presenting the result as a neutral metric, I mapped different percentage ranges to humorous, exaggerated responses. These outcomes were presented as a downloadable “Performative Male Certificate,” designed for sharing and social play. The certificate format intentionally mirrors institutional authority, highlighting the absurdity of algorithmic judgment applied to identity.


Training Model

The first step was training a model to visually differentiate between “Performative Male” and “Non-Performative Male” archetypes. I collected and organized approximately 25–30 images per category by scraping the web, then fine-tuned a model using Hugging Face in Python. The goal was not accuracy in any scientific sense, but to explore how subjective cultural categories are translated into datasets.

Testing

Once trained, the model outputs a percentage score indicating how closely an image aligns with the performative male archetype. This numeric abstraction became the foundation for the project’s interaction and tone.

The Fun Part

Rather than presenting the result as a neutral metric, I mapped different percentage ranges to humorous, exaggerated responses. These outcomes were presented as a downloadable “Performative Male Certificate,” designed for sharing and social play. The certificate format intentionally mirrors institutional authority, highlighting the absurdity of algorithmic judgment applied to identity.


Training Model

The first step was training a model to visually differentiate between “Performative Male” and “Non-Performative Male” archetypes. I collected and organized approximately 25–30 images per category by scraping the web, then fine-tuned a model using Hugging Face in Python. The goal was not accuracy in any scientific sense, but to explore how subjective cultural categories are translated into datasets.

Testing

Once trained, the model outputs a percentage score indicating how closely an image aligns with the performative male archetype. This numeric abstraction became the foundation for the project’s interaction and tone.

The Fun Part

Rather than presenting the result as a neutral metric, I mapped different percentage ranges to humorous, exaggerated responses. These outcomes were presented as a downloadable “Performative Male Certificate,” designed for sharing and social play. The certificate format intentionally mirrors institutional authority, highlighting the absurdity of algorithmic judgment applied to identity.


Results

Live Demo

Generated Certificates (read at your own risk!)

Live Demo

Generated Certificates (read at your own risk!)

Live Demo

Generated Certificates (read at your own risk!)

Reflection

This project was a fun and accessible entry point into training ML models with Python, while also poking fun at how easily culture gets flattened into data. The site isn’t public yet, but my next steps are refining the UI and releasing it as a party-friendly experience meant to be shared, laughed at, and maybe taken a little too seriously.

This project was a fun and accessible entry point into training ML models with Python, while also poking fun at how easily culture gets flattened into data. The site isn’t public yet, but my next steps are refining the UI and releasing it as a party-friendly experience meant to be shared, laughed at, and maybe taken a little too seriously.

This project was a fun and accessible entry point into training ML models with Python, while also poking fun at how easily culture gets flattened into data. The site isn’t public yet, but my next steps are refining the UI and releasing it as a party-friendly experience meant to be shared, laughed at, and maybe taken a little too seriously.

Create a free website with Framer, the website builder loved by startups, designers and agencies.