Generative Design.  A swimmer in a natural sea frame


Celestino Soddu

Professor of Architectural Design at Master in Building Engineering

Director of Generative Design Lab, Dept. of Architecture and Planning

Politecnico di Milano University.  website:



Generative Design is a logical synthesis of a creative process using transformation rules (algorithms). It can be realized designing a program able to simulate this process and to generate outputs as 3dmodels of architecture, cities, objects. As all creative processes it involves subjectivity in the definition of how the process runs and how the transforming rules are created and organized into a system.

As in all creative processes two main factors are involved: the unpredictability of external factors linked to each design occasion, like the environmental context and client’s requests, and the subjectivity of designers when they interpret these external factors.

Being a logical synthesis of this complex dynamical system, Generative Approach can be successful used in teaching architectural design as subjectivity-oriented approach.


Morphogenetic Meta-Project versus Project

Generative Design could be represented like a morphogenetic meta-project, an organized idea of “how to run” a design process. In sixties of last century meta-projects were the structure of organization of incoming projects. They were constructed with the aim of identifying the best structure to answer to “objective” functional needs. It was not possible to full develop this kind of approach because functional needs, extended to practice functions but also to symbolic and aesthetical functions, are strongly related to the subjectivity of customers and to the subjectivity of designers.

Generative Design is a meta-project with two fundamental extension:

1. it involves subjectivities going more in deep into complexity of (architectural, town environment, industrial objects…) designed artificial systems.

Perspective versus Axonometric

It’s like moving from axonometric to perspective view.  Adding subjectivity you can move from the axonometric representation, “objective” because free from subjective views but limited by the dimension of the sheet, to the perspective view that, using subjective points of view, can represent the infinite in one sheet and, following that, the increasing complexity of represented systems.

2. it can run the design process a lot of times, being sensible to little variations of inputs (similar to the different feel of the designer in different moments) and it can generate a sequence of endless results, all different but all related to the designer idea.

Anamorphic versus Axiomatic

It’s like moving from normal perspectives view into anamorphic perspective views. Each different point of view transform the anamorphic representation into unpredictable scenarios. Looking at each of these outputs we can discover one of the possible representation of the idea.


Process versus Output

Generative Design, as subjective operative meta-project, can be used to design a kind of artificial objects, an artificial DNA of a species of objects because is oriented to set up a process and not only to reach one result. More, it defines and renders explicit all the steps of a “normal” design process, from the first sketch to the final executive project. And, in this way, it’s a wonderful support for teaching (architectural and industrial) design.


Idea versus Solution

If you are a designer and someone ask you: “which is your idea, which kind of forms are you thinking to” or “which character do you love for your architectures” or “ which kind of impact do you prefer to have when you enter in a unknown city”, you can explain it using words but you cannot show it if not using a generative approach because it represents an idea and not peculiar solution of problems. Each idea can be developed with multiple solution but each solution is not exhaustive of the idea.


Synthesis versus Simplification

Interpretation versus Analysis

The reason is that you cannot use drawings, forms or images able to explain your thought in an exhaustive way because you have to perform a synthesis of all your beloved forms, including possible unknown forms that could fit your needs. You cannot simplify.

More, answering to all these questions, you have to explain the complex system of relationships and possible interfaces that a town environment must have for linking your needs and your interpretation of unpredictable user’s needs.

Generative design is not an analytical process but a synthesis process. The core of each generative project is the synthesis, using algorithms, of own subjective approach to context and to own subjective cultural, technological and functional references. In brief, generative design performs own modus of approaching the transformation of existent worlds into possible environments more closed to own idea of quality.

Following that, In generative design processes, but also in all creative processes, the subjective interpretations of existing world are the main creative acts.


Subjectivity versus Objectivity

Imitation versus Copy

The role of subjectivity is really important in design activity. Without subjectivity we loose the main stimulus for evolving our functional scheme into a project, and we lose the possibility to use our cultural, symbolical, aesthetical, technological references because these references can be used only if we have our interpretation of them. If not we can do only copies.


Variations versus Optimisation

In last century we experienced a design approach focused on optimisation. It came by the need to identify “the best” and realize it with industrial assemblage chains. This approach belongs to the concept that all people are equal, all people need the same equal product. This approach is not more accepted and it’s known that we can realize a product with the best performances but with different form. These differences fit the need of personalization of products, fit the need of customers to find his own product, fit the need of each person to find out a product, a house, a car, a square, a city, an environment that fit his needs and that is the mirror of his identity and uniqueness. The subjectivity of designers fit the subjectivity of customers.

The possibility to manage variations is inside the quality of a design process.  Generative projects face directly this need generating unpredictable, but recognizable outputs.


Identification versus Homologation

Because the recognizability of outputs is the explication of the architect’s (or artist’s) imprinting. But it’s also a function that each customer appreciates when looking for something facing his own subjectivity. We can recognize a print of  Piranesi, also if we never have seen it before, because we recognize the style or, we can say, the DNA of his drawing process, his stratified interpretations of surrounding environment that makes unique his drawings.


Recognisable versus Anonymous

In the some way we need to identify our home when we go home, to identify our city and to love the link between its unique character and our way to look at the future.


Generation versus Cloning

Generative Design realize a species and not only single outputs. It’s like in Nature. Generation is strongly different from cloning, like art craft is different from assemblage chain. But now an industrial production of all different objects is technically possible using the existing industrial equipment and generative projects. More, variations, as in nature, enhance the recognizability and peculiarity of each idea facing, in the meantime, the need of personalization of each customer. In this era, when someone  tries  to clone, like in the last century industries, the natural events, we like to rediscover, in the artificial world, the uniqueness of generation, like in Nature.


Unpredictability versus Repetition

The unpredictability of variations of natural objects, like a rose or a cat, enhance the rose and cat concept, the identity of these species, like the unpredictability of a each variation of Bach enhance the identity of his music and our ability to identify and appreciate it. On the contrary, repetition destroys identity. A compound  of all equal houses has less identity (and is less fascinating) of a compound where all the houses are different but each one follows a recognizable common idea of quality.  Variations as mirror of the subjectivity of each inhabitant.


Random of requests versus Random of outputs

The use of random factors is important in constructing this process as software. It follows two different approaches. Random as possibility to create unpredictable requests, constrains, needs as occasions to improve the complexity of the process and testing the recognizability of the idea in all generated events, or random as generator of casual forms. The difference is, substantially, the recognizability of the artist/architect/designer imprinting.


Teaching How versus Teaching What

I am teaching architectural design from 30 years and, after having developed my first generative software in 1987 and published the related book in 1989 (C.Soddu, Citta’ Aleatorie  (Random Cities), Masson Publisher), It was me clear that, as my Argenia Generative Software worked simulating the design processes and generating architectural outputs able to fit different occasions and client’s needs, so I would have been able to use Argenia as model for investigating about designing process. It is, in fact, an effective logical synthesis of normal design processes.

So I used it as base of an operative and effective teaching structure for Design Studio Labs in architecture, environmental and industrial design.

Outputs of students were soon really encouraging because based on discussions about “how” develop their work and not about “what” they are designing, giving tools for managing their work without discussing about their partial and temporary results but looking at them as a first step of a transforming process. This teaching approach enhanced their subjectivity and the possibility to use at the best their own cultural references. In other words they succeed in using all their previous learning work by interpreting these references as transforming codes. But this aspect created also some difficulties. We can synthesize these difficulties as belonging to the student’s denial to render explicit his subjectivity, his cultural peculiarity, also if only in the field of design approach. The common request of some students (but only in the first steps of the learning path) was the request of a more “objective” teaching process were each student can work “analytically” for reaching a surely acceptable result. But design process is not an analytical process. If teacher don’t talk about subjective approach (as done by design teaching in the last century for ideological problems) he misses the possibility to enter in deep into discussion on design processes.

It was also difficult to clear to the students that the assessment of their projects developed in the Design Studio Labs is done valuating:

1. the congruence between the aims indicated by each student at the beginning of their work and the structure of the transforming rules that they designed during this experience.

2. the progressive transformation path in terms of increasing-complexity difference between the initial sketch and the final drawings.

3. The possibility to manage again, in different design occasions, the same reached quality. This point is the main point because it demonstrates the increasing professionalism of the student.


Teaching structure. A subjective oriented design approach for teaching design.

Transforming versus Forming

The steps for running a generative approach in teaching design are:

1. Each student is required to identify the character of his idea by interpreting his references into transforming rules. First the student can list a sequence of characters using words, like adjectives. Each student is required to identify 3 adjectives for describing his design aim. (3 and not less or more because it’s interesting to make synthesis and to be not too much axiomatic. The choice of three adjective was used, the first time, by E,Colabella that experienced in her courses the same logical process). Students have to identify some references as representation of each character.

2. Second step is identifying the different design moments, when the designer have to choose how to fit the incoming functional, aesthetical and symbolical requests. These moments can be identified like:

How to fold an element

How to divide an element into different parts

How the element ends

How the element lean on

How make holes into elements

How …

3. the third step is the more creative one and each student is required to explain and use his design subjectivity.  Each personal reference can be interpreted as transforming rule applied to each different design moment. The request is: how I can transform the previous step of my project into an incoming one fitting a concrete request and reaching, in the meantime, the identified character? For example, if the student have to open a door into a wall, to divide a wall into parts, to shape how his building can end, and so on, how he will manage these incoming transformations using his own references? He will identify previously a sequence of transforming rules that could be applied to the project in progress for fitting different requests. The result of this creative work is a set of rules that could be considered like designer’s subjective DNA. The reference to DNA is correct because these rules, when used, bring the project to be transformed in progress fitting two main questions: the increasing functionality of the project referring to the client’s requests and the increasing identity and recognizability of the project by fitting the characters  representing the designer’s imprinting.

4. the subsequent step, that is the fist one directly linked to the starting a particular project, is to set up a paradigm of organization able to support the transforming path increasing its peculiarity and its functionality. The main difficulty regarding this stap is to design an open system, really adaptive but, in the same time, really characterised.

5. Now we have the two engines: the paradigm of organization and the set of transforming rules. It’s time to run the generative process using the rules in front of each incoming request, This action will generate a scenario. That is only one of the possible results. This work is exactly the same of normal design activity. But the structure is really clear: these scenarios are generated using predefined transforming rules (focused on defined characters) into a peculiar field of relations designed for representing the project functions. The interesting question is that this design path is clear and understandable by each student that can open a discussion with the teacher referring to his peculiar interest and references. Also if the students, as normally happens, don’t use a software to manage this path but run the process with “normal” tools.

6. The last moment is the possibility to generate variations. Using the defined rules in different moments of the evolving design history, each student discovers that he can reach his own aim with different results. More, students discover that variations are a good representation of their idea that cannot be represented by only one result. And they discover that they can manage the reached quality of their project also in other incoming projects, because the transforming rules that they have designed are useful to be used again reaching the same character and imprinting. They have discovered how to manage their professional identity and recognizability when managing their incoming projects.

7. In practice, each student have done his subjective meta-project that represent his identity as architect.

Using this approach, finally, teachers and students can discuss about design process, about how each designer can go ahead with his ideas, following his specific means, needs and aims and not only discussing on the (final) output.


Layering versus Permutation

One of the interesting quality of generative approach is that students learn how to manage the complexity using layering of different transforming rules. As it’s known, complexity cannot be reached in only one step. The quality of architectures and cities spring up from history, from the personal history of a design path when the designer fill the project with different feelings of different moments (the more interesting example is that architects, when they need to go ahead with a project that don’t seems to grow, turn the drawings to the opposite site for having a different view of their work, for finding out a unpredictable point of view) and, regarding cities, from the history of different cultural moments.

More, the generative approach using transformations instead of solutions give a further possibility: a good work for a team of different people with different field of interest. If each partner of a design team gives his contribution with a solution, it’s really difficult to put together all the contributions into a final output. Forms/solutions cannot be stratified but only permutated. But if each partner gives his contribution with a transforming rule it’s easy to run, one after the other, each rule. At the end each partner will find, in the final result, the representation and attainment of his own idea.

This teaching approach was experienced by around two thousands of students of Politecnico di Milano, Faculty of Architecture, Industrial Design and Engineering in the courses held by me and by Enrica Colabella starting from 1989. In 1992 E.Colabella and me wrote the book “ Il progetto ambientale di morfogenesi” (the Morphogenetic Environmental Project) published by Esculapio Progetto-Leonardo Editor, Bologna, explaining this teaching approach. In the meantime more then one hundred of master thesis with my and E.C. supervision were made using this approach. All the student’s experience were, staring from 1995, posted in the website

The main results, as confirmed by students that are now running their professional activity, are:

1. they are facilitated to increase in progress the quality of their projects because they know how to reach, in each subsequent project, the quality already reached in the previous ones.

2. they succeed in realizing their projects in less time, because they use the experienced transforming codes. In this way they can find time for increasing their transforming rules taking advantage from the peculiar requests of each project.


Philosophy versus technology

The generative design approach is not a technology but a philosophy. It identifies a particular approach to understand, design and manage the incoming complexity of artificial systems, cities, architectures, environment, objects. It can be easily transformed in technological tools because it uses transforming rules that can be easily written in algorithms.


Rules versus Forms

Organizing versus Choosing

The creativity is focused on logical processes and not on results, on organizing the system and not on choosing solutions. Choosing the emergent event using random forms could bring to shape a good result but, in this case, the quality is not repeatable.


Impervious versus Flat

Occasion versus Obstacle

The client requests, the constrains, the difficulties of a project are welcome. Each new request open the possibility to increase the final quality. Complexity is considered as the ability to answer to the different, sometimes contradictory, unpredictable needs of users.


Organic versus Minimalist

The design process is similar to processes in Nature. It uses something like artificial DNA performed like a set of transformation rules. The aim is a natural organic architecture able to answer to unpredictable requests through complexity, as natural events. On the opposite side we could find minimalism if we  consider it as an attempt to fit different needs with something that looks like an optimisation.


The design approach is focused on transforming and not on forming. This approach  is considered related to the real approach experienced by designers. The generative philosophy of design can be synthesized by:

Proportions versus Grids

Dynamic versus Static

Parameter versus Measure

The design approach focused on transforming and not on forming have a long history in Renaissance. The attention, experienced in the past industrial era, to modules and to grids change into a new, but traditional, attention to proportions and parametric definitions because they are more closed to the management of dynamical evolutions. Modules and fixed measures are not more useful for managing the industrial production that uses numerical control equipments and robots.


Permeable versus Waterproof

Improving versus Substituting

Facilitation versus Hindering

And versus Or

Transformed versus New

Memory versus Lost

Palpable versus Untouchable

Contaminate versus Pure

Perfectible versus Perfect

The quality of a design process can be valuated through complexity, that is not complication but synthesis of different and contradictory abilities. This complexity can be reached through layering of different inputs, needs, references, feelings. And the project in progress have to be read, to be improved, transformed, contaminated without loosing his character and identity. But enlarging its own memory. This is possible only if the identity is managed by the modus of running the process and not by the used forms.

Works of my students at their Master Thesis (

Interpretation of a tree from Van Gogh and generation of 3D trees. Emilio Molinaro 1996


Interpretation of “Le Carceri” by GB Piranesi and construction of subjective 3D model. Enrico Mazzei 1996


Transformation of 3D car models following identified codes. Luigi Martinetti 1997


Interpretation of sections and geometries and design of Gaudi’ codes for generating endless “Gaudi’” buildings. Matteo Codignola 2000.












Construction of generative transformation rules of Manhattan identity. Four steps of evolution. Mariateresa Capodici, Marco Melino  1996





(for more references



C.Soddu, ‘L’immagine non euclidea’ (the not Euclidean image), Gangemi publisher, 1987

C.Soddu, ‘Citta’ Aleatorie’, (unpredictable cities), Masson Publisher. 1989

E.Colabella, C.Soddu, ‘Il progetto ambientale di morfogenesi’, (the environmental design of morphogenesis), Progetto Leonardo Publisher 1992

C. Soddu, “Milan, Visionary Variations”, Gangemi publisher, Rome 2005



C.Soddu, ‘Simulation tools for the learning approach to dynamic evolution of town shape, architecture and industrial design’, in ‘Proceedings. International Conference on computer aided learning”, published by Press Polytecniques et Universitaries Romandes. Lausanne 1991.

C. Soddu, "From Forming to Transforming", proceedings of Generative Art Conference, Milan, Alea Design Publisher, December 2000

C.Soddu, “Recognizability of the idea: the evolutionary process of Argenia” in P.Bentley & D. Corne (edited by), "Creative Evolutionary Systems", Morgan Kaufmann Publisher, San Francisco US, 2001

C.Soddu, “New Naturality: a Generative approach to Art and Design”, Leonardo Magazine 35, MIT press, July 2002

C. Soddu, "Generative Design / Visionary Variations - Morphogenetic processes for Complex Future Identities" in the book Organic Aesthetics and generative methods in Architectural design" edited by P. Van Looke & Y. Joye in Communication&Cognition, Vol 36, Number 3/4, Ghent, Belgium 2004

C.Soddu, E.Colabella, "A Univesal Mother Tongue", Leonardo Electronic  Almanac Vol.13, Number 8, August 2005