[ Material experiments for probing ]
{ Coded-This-Way }
2 Weeks | 2023 | Interaction, Design, Research | Guidance: Prof. Manuel Beltrán
Design research project that investigates a black-box system and translates findings into an educational experience.
Context
/**
Generative AI tools are increasingly used in creative practice, often without a clear understanding of how they work or the biases embedded within them. This project investigated AI image generation systems through research and experimentation, and explored ways of making their hidden processes more visible and understandable.
*/
Problem STATEMENT
Design Challenge
How might we make the hidden workings of AI image generation tools more visible to everyday users?
The challenge was to investigate systems that operate as black boxes, with limited access to source code, training datasets, and internal decision-making processes.
Process
// Step_01: Secondary Research
- AI image generation
- Data annotation
- Data labour
- AI ethics & law
// Step_02: Inquiry
- Research objective
- Investigation framework
// Step_03: Primary Research
- Prompt testing
- Image generation
- Reverse engineering
// Step_04: Analysis
- Visual patterns
- Training datasets
- Biases & limitations
// Step_05: Experimentation & Ideation
- Communicating findings
- Interactive experiences
// Step_06: Prototype
- Interactive website
- Wireframes
- User journey
Deep_Dive Research
Primary Research & Analysis
/**Using prompts, generated outputs, and image annotations as research material, I investigated the relationships between training data, human labour, and AI-generated imagery. Mapping these connections helped uncover some of the hidden processes, dependencies, and biases embedded within image generation systems.*/
[ What I Designed ]
Coded-This-Way
/**
A prototype website designed to help users understand how AI image generation tools work through interactive exploration.
*/
// The experience introduced:
> training datasets
> image annotation
> generated outputs
> ethical implications of AI-generated content
Website Hi-Fi Prototype
// Surf through to view website prototype screens.
Key Design Decisions
Strategy & Principles
// Decision_01
// Decision_02
Interaction over explanation.
// Decision_03
Visibility over abstraction.
// Decision_04
Exploration over instruction.
Reflection over consumption.
Why I Designed It This Way
Design Principles
> Interactive learning: users learn by exploring and doing, not by reading long explanations.
> Making hidden systems visible: exposing the invisible layers of AI image generation so people can see how outputs are constructed.
> Connecting data and outputs: revealing how prompts, datasets, and data annotations shape what users see on the screen.
> Critical engagement: encouraging reflection, questions, and discussion instead of passive consumption of AI-generated images.
Why this Project
