Created by: Group 24
Overview: This project investigates how generative AI imagines attractiveness on social media by comparing images generated by ChatGPT and Doubao across three categories: attractive influencer, ordinary person, and unpopular person. The same prompts were given to both platforms and the outputs were compared in terms of race, gender, skin texture, lighting and visual style. The final output is a short video presenting AI-generated images through a simulated social media interface.
The project found that ChatGPT tended to generate young, white, smooth-skinned women conforming to Western aesthetics, while Doubao produced images closer to Chinese internet celebrity style. Both platforms presented highly idealised and homogeneous beauty standards despite their stylistic differences. The project argues that online attractiveness is not neutral but shaped by platform logic, and that AI repeats and reinforces existing aesthetic norms and racial biases already present in society.
Created with: ChatGPT, DouBao
Concept
Poster Description: The poster displays six pairs of AI-generated faces arranged in a visibility hierarchy from high to medium to low, comparing outputs from two fictional algorithmic systems given the same prompt, with polished and idealised faces at the top and less styled, rougher images at the bottom. It argues that online attractiveness is not neutral but ranked by platform logic, and that AI systems turn beauty into measurable, sortable data shaped by visibility metrics and algorithmic culture rather than personal judgement.