MOME
2020
Facescan – Graduation project
An online educational platform about the risks of facial recognition technology, where visitors experience through their own face how machines see us
Special thanks for the development to
Bank Benjámin
Overview
I graduated in 2020 from the Graphic Design MA program at Moholy-Nagy University of Art and Design (MOME), where my graduation project was the Facescan platform.
Facial recognition systems are being used more and more widely today, making it increasingly important for the conversation about their benefits and risks to reach a broader public. Facescan is an online educational platform that focuses on the dangers hidden within today's facial recognition systems, creating a space to learn about and discuss the technology. Its goal is to raise awareness of these risks through visual, hands-on, experience-based methods.
My role
I worked on this project from concept through UX/UI design to final execution. The platform's development was handled by Benjámin Bank, with professional guidance from my MOME instructors, Balázs Vargha and Béla Hegyi. I was responsible for developing the platform's visual identity, functionality, and content structure.
I also had the opportunity to present the project at the Goethe-Institut Budapest as part of the MI2040 interactive exhibition, where visitors could try the platform for themselves. During the exhibition design process, I received mentorship from Zsófia Ruttkay and Judit Bényei.
Challenges
Making machine vision tangible
It's not enough to simply explain how facial recognition gets things wrong, visitors needed to experience it firsthand. This meant designing an interface that shows in real time and on the visitor's own face, how the system works and where it fails, for example, when it misidentifies gender or age.
A graphic solution only humans can see
One of my research questions was whether it's possible to design an illustration system in which faces are recognizable to the human eye but not to machine vision.
Reaching a younger audience
Facial recognition systems identify children too, so the content and experience needed to be understandable and relevant for visitors under 12 as well.
Structuring a broad topic
Facial recognition can be approached from many angles, so the content had to be structured in a way that let visitors easily orient themselves and explore the topics that interested them most.
Design solution
Facial recognition technology built into the interface
Facescan analyzes the visitor's face in real time, letting them experience firsthand how machines see us and what problems can arise during identification. The platform shows both how the system can categorize us based on our faces, and the kinds of discrimination this can lead to.
Illustrations AI can't recognize
When developing the visual direction, the goal was for the visual language itself to reflect the nature of machine vision, point-by-point encoding and breakdown into pixels. The research question was whether a graphic style could be created in which the depicted faces are recognizable to the human eye but not to machine vision. The illustrations were built around this idea: the human eye identifies the faces, while automatic machine vision does not, sharply illustrating the gap between the two.
A dynamically changing visual identity
The platform's identity includes a dynamically changing face symbol and a matching logotype whose form also reflects machine vision, the lettering breaks apart into pixels and dots, evoking the scanning of a face. The illustrations are built from different raster structures that change dynamically depending on the mood the system detects. If the system reads a happy mood from a visitor's face, the image's raster pattern is made up of happy faces too.
Children's version
If the system detects that a visitor is under 12, an age-appropriate interactive page appears instead. In a face-puzzle game, the user has to assemble the opposite of the detected mood using movable facial features, eyes, mouth, nose, and eyebrows. Players can choose between light- and dark-skinned faces, pointing to the fact that facial recognition systems unfortunately identify people of different skin tones with varying accuracy.
Content structure
The content's primary organizing principle was the direction of action between humans and machines. I grouped the content into three categories: human actions against machine facial recognition, human–machine "collaboration," and machine actions directed against humans resulting from the misuse of facial recognition technology.
Content selection was also guided by another question: in which sphere does the face function as a means of communication? Following Edward T. Hall, I distinguished four spheres: intimate, personal, social, and public. Tags above each piece of content indicate its type, giving visitors a quick overview while also functioning as filters, so related topics are easy to find.
