AI Ethics and Human Agency
Designing with judgment
Ai + Emerging Tools · RESOURCES
Generative tools raise questions that sit squarely in the designer's seat. Whose work were these models trained on, and did those creators consent or get paid? Who is represented in the output, and who is quietly left out? What gets claimed as original when a machine assembled it from everything it has seen? None of these questions has a pop-up answer — but every one of them now passes through design work.
Human agency is the other half of the equation. When a tool suggests, generates, and refines at machine speed, the risk is not that it takes over — it is that designers stop deciding. Taste, brief, and judgment are the job; automation should compress the boring distance between an idea and a first draft, never replace the act of choosing. The designers who thrive with these tools are the ones who stay the editor, not the approver.
This page is not a rulebook — the field is moving faster than any code of conduct. It is a working set of questions and habits: know your tools' training stories, disclose AI use honestly to clients, audit output for bias and sameness, and keep the decisions that define the work in human hands.
The step-by-step guide
- 01
Ask where the training data came from
Before adopting a tool, read up on what it was trained on. Prefer tools built on licensed or opt-in datasets when the work is commercial, and know which tools carry legal or ethical exposure before they touch client work.
- 02
Keep the brief and the judgment yours
Let AI generate options, but never let it decide. Write the brief yourself, set the constraints, and make every selection, rejection, and revision a deliberate act — that is where design actually happens.
- 03
Audit output for bias and sameness
Generated imagery skews toward the averages in its training data: narrow ideas of faces, bodies, cultures, and 'good design'. Check every output for who is missing, then push past the generic look the model keeps offering.
- 04
Disclose AI use honestly
Tell clients where AI entered the process, and check agreements for clauses on generated content. Honest disclosure builds trust and protects you when questions of ownership or originality surface later.
- 05
Protect the people behind the work
Watch for tools that imitate living artists without consent, and be deliberate about what you feed them. Reference ethically, credit where credit is possible, and support the human creators whose work makes these tools possible.
- 06
Stay the editor, not the approver
The trap of fast generation is rubber-stamping. Slow down where it matters: interrogate the drafts, rework them in your own hand, and only sign work you can explain and defend line by line.
The further reading
Partnership on AI
Visit site ↗A coalition of companies, researchers, and civil-society groups setting guidance on responsible AI — including how synthetic media is disclosed and how creative industries adapt to generative tools.
Ada Lovelace Institute
Visit site ↗An independent research institute examining how AI affects people and society. Its work on data rights, biometric surveillance, and model provenance is essential background for anyone shipping AI-assisted products.
Mozilla Foundation
Visit site ↗Campaigns for an internet that is open, trustworthy, and human-centered. Its AI transparency work and Internet Health reports track who controls these systems — and who they control.
AI Now Institute
Visit site ↗A research center focused on the power concentration around AI. Its reports on labor, accountability, and the industry behind the models connect directly to design's own labor questions.
Design Justice Network
Visit site ↗A community of designers working to redistribute design's benefits fairly. Its principles are a practical compass for who a design serves, who it harms, and who gets to decide — the same questions AI tools now amplify.