Approach · 1960s computational roots–present · International art, research, and engineering
Generative design
Author a process that produces possibilities.
What is Generative design?
Generative design uses a rule system, algorithm, or responsive process to produce outcomes. The designer chooses the inputs, constraints, evaluation criteria, and degree of variation. It includes much more than contemporary generative AI: deterministic code, controlled randomness, simulations, and biological processes can all participate. A compelling result shows an intentional relationship between the system and the brief. An attractive output selected from many unexplained variations is not, by itself, evidence of a thoughtful design process.
The story
Computer art and rule-based practice established important precedents before today’s tools. Later fabrication and simulation systems connected digital processes to physical materials. Neri Oxman’s Silk Pavilion projects illustrate a bio-digital branch in which environmental and biological behaviour contribute to form.
How do you recognise Generative design?
- Families of related but nonidentical outcomes
- Explicit rules and constraints
- Controlled variation rather than arbitrary noise
- Selection criteria documented alongside the output
Colour, type, layout and material
- #F0E8DA
- #C8A775
- #75846C
- #233A39
Typography. Use calm typography to identify rules, seeds, and versions. Treat labels as part of the work’s explanation rather than pseudo-scientific ornament.
Composition. Show one selected result alongside a small family of alternatives. Make the relationship between input, process, and outcome visible without overwhelming the final composition.
Materials. Code, simulations, procedural geometry, data, biological processes, and digital fabrication. Choose materials and algorithms that relate to the problem rather than to fashion.
Why does Generative design work?
Variation within recognizable order can sustain curiosity. Perceived complexity does not imply intelligence or quality. People need an explanation of the process to distinguish meaningful adaptation from decorative randomness, especially when results affect decisions.
The principle. Design the conditions of emergence. Creativity lives in the rules, the choice of inputs, and the judgment used to evaluate results. A system becomes more meaningful when the maker can explain both what it permits and what it excludes.
Where it was used
Silk Pavilion II, Neri Oxman and The Mediated Matter Group, 2020. A co-fabricated pavilion combines a designed support and manufacturing process with silkworm behaviour. It demonstrates generative thinking beyond image-generation software.
- Variable identity systems
- Material and product exploration
- Procedural motion and exhibition graphics
What to avoid
- Randomness without a governing idea
- Hiding selection bias behind algorithmic language
- Assuming generated form is automatically optimized
Try it: a short exercise
Write a rule that generates twenty compositions from one subject-specific input. Define three evaluation criteria before running it. Select a result, then document why it succeeds and what the rejected versions reveal about the rule.
How do you prompt for Generative design?
These prompts ask for an original piece informed by the style, not a copy of any work. Replace the words in square brackets.
For an image or visual design
Create a generative [FORMAT] for [SUBJECT] and [AUDIENCE]. First define the input, rules, constraints, and selection criteria. Produce a family of six related compositions with controlled variation in density and orientation. Explain how the variation reflects the subject. Select one result for refinement and preserve its distinctive irregularities. Avoid arbitrary particles, meaningless networks, and claims of optimization without evidence. Create an original composition, not a replica of a named artwork or brand. Explain three decisions that connect the result to this brief. Use supplied wording accurately; leave space for manually typeset text if text rendering is unreliable.
For an interface
Implement a generative visual tool for [SUBJECT] and [AUDIENCE]. Expose a small set of meaningful parameters, provide reproducible seeds and a reset action, and explain how inputs affect output. Keep the rendering lightweight and provide a static fallback. Separate experimental visualization from any real decision support. Let users compare saved variants and understand why a selected result differs. Use semantic HTML, responsive layouts, visible keyboard focus, labelled controls, and accessible contrast. Include loading, empty, error, and success states where relevant. Respect reduced-motion preferences. Use realistic supplied content, not generic placeholder marketing. Explain how the visual choices support the user’s task. Deliverable: [FORMAT].
Lineage
Draws on: Kinetic Art, Parametric design, Abstract Expressionism.
Influenced: Data-driven visual design, Glitch aesthetics, Immersive design.