generativesystemspapers.cumincad.org/data/works/att/ecaadesigradi2019_100.pdf · creativity:...

Generative Systems:

Intertwining Physical, Digital and Biological Processes, a case study

Gonçalo Castro Henriques1, Ernesto Bueno2, Daniel Lenz3,Victor Sardenberg41,3Federal University Rio Janeiro, LAMO, PROURB - Brazil 2Mackenzie Univer-sity / Positivo University 4Leibniz Universitaet Hannover, Germany1,3{gch|daniel.lenz}@fau.ufrj.br 2,4{ernestobueno|vsardenberg}@gmail.com

The fourth Industrial Revolution is characterised by the computational fusion ofphysical, digital and biological systems. Increasing information in terms of size,speed and scope exponentially. This fusion requires improved, if not new, toolsand methods to deal with complexity and information processing. By openingGenerative Systems to interact with the context, we believe that they can developsolutions that are more adequate for our time. This research began with aliterature review about generative systems and their application to solveproblems. We then selected the tools, Cellular Automata, L-Systems, GeneticAlgorithms and Shape Grammar, and thought about how to translate theseoriginal mathematical tools to specific design situations. We tested theapplication of these tools and methods in a workshop, implementing recursiveloops to open these techniques to interference. Analysing the empirical resultsmade us revise our design thinking, relying on the study of complexity tounderstand how these techniques can be more context-aware, so we can makedesign evolve. Finally, we present a comparative framework analyses thatinterlaces techniques and methods, so in the future we can merge physical, digitaland biological information.

Keywords: generative systems, design thinking, complexity, context interaction,recursion

RESEARCH APPROACHTo promote Generative Systems in architectural de-sign, we developed a 5-stage process. First, we re-searched about generative systems and how to ap-ply them. Secondly, we thought about the transla-tion of abstract mathematical techniques to designcontext. Thirdly, we tested these techniques empiri-

cally to gain a better understanding of their capabil-ities and limitations in a workshop. Fourth, we anal-ysed the tools and methods used and the results foreach technique. Results recommended deepeningour knowledge about design cognition and complexsystems theory. Finally, we synthesised and com-pared the techniques and design methods.

Design - GENERATIVE SYSTEMS - Volume 1 - eCAADe 37 / SIGraDi 23 | 25

GENERATIVE SYSTEMS AND COMPLEXITYThe use of generative systems in design, unlike tradi-tional methodology, implies an indirect relation withthefinal product. Production, rather thanbeingdonedirectly with the “designer’s own hands,” is mediatedby a “generative system” (Fisher and Herr, 2000). Todefine a generative system, it is necessary to definethe abstract set of rules andproceedings thatwill cre-ate a set of objects. If wewant this object tomake anysense in a certain environment, these rules must beabout the object-environment relation. Otherwise,the meaning of the object in a specific context willbe a matter of chance.

Generative systems, exploring computationalpower, can provide unimaginable solutions, expand-ing creativity. According to Gero (1996), the mostcommon concept of creativity does not consider theability to develop possibilities during the genera-tion of solutions; only the ability to improve end-product quality. To create multiple solutions, it isnecessary to work with methods that explore andrecord the possibilities of solving a problem, aggre-gating unexpected and less familiar results usuallydiscarded. Taylor (1972) categorises five types ofcreativity: expressive, productive, inventive, inno-vative and emerging. In this context, it is impor-tant to emphasise beyond the “emergent creativity”,the “productive creativity”, that is, the creativity con-textualised in the domain of a technique, which al-lows controlling the project in the environment gen-erated. Architects rely on Design Thinking (Simon,1969) tohelp solve ill-definedproblems, usingmainlyimplicit design processes instead of linear thinkingand with strict criteria. They value the hands-on ex-perience with the tools and their application to findsolutions, evaluate and develop them, seldom mak-ing any explicit algorithms. We often leave optimi-sation till the final stages, after defining the form,to improve its performance. However, mathematicaloptimisation and processes, such as generative sys-tems, have a precise definition and strict proceduresto solve problems. So, to understand generative sys-tems, architects must enter the algorithmic process.

This requires opening the algorithm“blackbox”,mak-ing the mechanisms to produce results explicit, un-like the traditional “hands-on processes”, where thefinal results are felt more directly. In a sense, archi-tects need to cope with extensive processes to ex-pand their abilities.

Composition and Complex SystemsBy their education, architects are able to deal withcomplex problems related with the nature of theirwork. Beyond “Complexity and Contradiction” (Ven-turi, 1966), the study of complexity in science hasdeveloped to consider complex systems, supportedby the evolution of mathematics and computation, aconceptual and technological evolution, intensifiednow with the fourth industrial revolution disruptivelogic (Schwab, 2017).

Complexity is not new in the Design Thinkingprocess. Formerly, architecture was an art in the inte-grative sense, but, after the Renaissance, it separatedinto different branches of knowledge in a continu-ous specialisation process. In architecture, the segre-gation was not only among disciplines, but betweencomposition and materialisation. Like the architect,the sculptor, the painter, the musician, the chemistand the physicist, all compose (Figure 1). What is therelationship between composition and complexity?

The relationship is closer than one might think.The architect, like other composers, invents the prob-lem - there is no scientific law to describe the con-text of how an object emerges (Buchanan, 2000 andStolterman, 2008). Thenhe formulates circumstantiallaws to adapt a methodology that deals with prob-lems beyond linear logic, dealing with implicit sub-jective techniques, where he intervenes directly. Theartist, like the observer in second-order cybernetics,interferes and changes his environment (Dubberly,Pangaro, 2015). When we speak of composition, weuse notions such as rhythm, weight, dynamics andstatic equilibrium, entailing a transition fromapart toa whole relationship and vice versa. Therefore, in ad-dition to quantities, the artistic process uses qualitiesthat have implicit measure subjected to an individ-

26 | eCAADe 37 / SIGraDi 23 - Design - GENERATIVE SYSTEMS - Volume 1

Figure 1Composition indifferent activities:composition asmaterial search insculpture (Chillida),expressive search inpainting (Tapiés1990), search inmathematics (Fuller1948), and researchin the study of thethree-dimensionalcomposition of thedouble helix(Watson and Crick1954). Photos in thepublic domain.

ual learning process. Colquhoun (1989) refers aboutcomposition:

“Composition came to mean a creative proce-dure in which the artist created ‘out of nothing’ andarranged his material according to laws generatedwithin the work itself ... Form was no longer thoughtof as a means of expressing a certain idea, but asindissoluble from, and coextensive with, the idea.Composition therefore was able to stand for an aes-thetic of immanence in which art became an inde-pendent kind of knowledge of the world.”

This article does not intend to argue about thedichotomy between art and science, but rather referto a time when these two areas of knowledge werebound, forming a wholeness. This return to an in-tegrated conception of different branches of knowl-edgewas anecessity in the50s, a reaction to scientificreductionism, as formulated in the “General SystemsTheory”. Thus, Integrating knowledge and tools in ar-chitecture, promised so much by the introduction ofthe digital in Architecture, is still to be fulfilled.

Observing the introductionof digital intodesign,the separation of materialisation and compositionor the tool and process dichotomy, seems to per-sist (Terzidis, 2003). To bridge tooling and the pro-cessual use of digital, computerisation and computa-tion, Terzidis proposed Algorithmic Design. The ar-chitect’s process seems somehow incomplete whenhe intends tomove into the unknown realm ofmath-ematical tools, new “black boxes” he wants to opento incorporate technology andmore information. Forthis, architects rely on exploratory search, a knowl-edge thathas similaritieswith composition searchus-ing analogue or pre-digital generative systems. Inthis aspect, architects have the ability to invent newrealities in other domains (Carpo, 2011). It is precisely

because the architect can invent the new that he cancontribute to refounding “the science of the artificial”(Simon, 1969). Simon refers to design as indispens-able: “The proper study of mankind is the science ofdesign, not only as the professional component of atechnical education, but as a core discipline for everyliberally educated man.”

So, it is not a question of how to apply designto solve complex problems (Herr, 2002). Rather, itis a question of how Design Thinking can (re)inventapproaches to Complex Systems. For this, we mightremember the characteristics and nature of designproblems (Buchanan, 2000). Also, how does artis-tic composition deal with these new tools? To com-pose is to establish an order among the parts - even ifwe do it according to top-down processes (architec-ture treatises) or bottom-up (rules of composition) -representing an understanding of how things cometo be, that is, describing thing also as a process, ormethod, of design, unifying a set of elements basedon the senses and previous experiences. Parallel toart, when science passed from deterministic logic tomulti-causal and probabilistic reasoning, it requirednew tools. In both cases, one can say experimen-tation is a heuristic process to narrow the computa-tional search space.

Complexity addresses relations among the partsand multiple levels of organisation as well as the re-lations (and behaviour) of these systems and theirenvironment. Due to its comprehensiveness, DesignThinking canbeuseful in different areas: a) Visual andsymbolic communication; b) Material objects suchas products, tools, instruments and machines; c) Or-ganisation of activities and services; d) Complex sys-temsor environments for living,working, playingandlearning. (Buchannan, 2000). Applying a synthetic


approach to design thinking, in concatenating parts,the whole context is essential for complex systems.

MelanieMitchell (2009) points out that to under-stand complex systems, it is necessary to address in-formation, computation, dynamics and chaos, andevolution. In her book, she introduces these sub-jects in each one of the chapters to set a conceptualframework to define and measure complexity. Shemeasures complexity as: Size, Algorithmic Informa-tion Content, Logical Depth, Thermodynamic Depth,Computational Capacity, Statistical Complexity, Frac-tal Dimension and Degree of Hierarchy. Finally, sheconcludes, “The diversity ofmeasures that have beenproposed indicates that the notions of complexitythat we’re trying to get at have many different inter-acting dimensions andprobably can’t be capturedbya single measurement scale.”

Complex Systems and Problem TypesAccording to Weaver (1948), science began by solv-ing simple linear problemswith few variables, duringthe XVII, XVIII and XIX centuries. Then it solved prob-lems of disorganised complexity, with many vari-ables, using statistics. Thus, it solved problems fromthe atoms to the stars resorting to computation us-ing probability and statistical data. The next kinds ofproblems to grasp are those of organised complex-ity, of the mesoscale of everyday life, with more vari-ables.

TRANSLATING GENERATIVE TECHNIQUESWe researched generative systems literature describ-ing the applicationof these techniques to solveprob-lems, identifying four significant techniques:

Cellular Automata (Neumann, 1951; Wolfram2002), L-Systems (Lindenmayer, 1968), Genetic Algo-rithms (Holland, 1975) and Shape Grammar (Stiny,1980). Each tutor studied one of these techniquesalong with a group of students, looking for exam-ples and understanding of the type of problems eachtechnique can solve.

We studied the technique’s application in prac-tical cases, resorting to the algorithms and descrip-

tions available. For 3 months, a group of researchersidentified types of problems, variables and applica-tions, considering the potential and limitations ofeach technique. This helped to formulate the work-shop design problem. References differed in eachtechnique, but we focused on how to implementthem in visual programming, testing available appli-cations. Relying on this experience, we prepared aproblem and tested it in the Workshop Form Findingand Generative Systems, LAMO, Rio de Janeiro 2017.

A Problem in Design ContextPreliminary research identified simple solutions todesign problems on the mesoscale. We set thespace limits between 3×3×3mand 10×10×10m, andthought about how to interact with the (few) vari-ables during multiple generation through feedback.We outlined an interaction between the algorithmand the environment to open the “black box”. Pre-liminary research identified similar situations amongtechniques, variables andcontexts. We looked for ap-plications of these techniques to public spaces, shel-ters, pavilions and constructive systems. By setting aboundary in the methods concerning the tools andthe computational processes, we identified a searchspace and factors that interfered with this search,from mere generic to applied search.

We set up the workshop for undergraduate andpostgraduate students, practitioners and professors,of different origins, in eight groups of three. Theyall learned to apply the different techniques, led byeach tutor, and then we carefully selected 2 groupsper technique. The participants’ diversity of back-grounds and knowledge gave us the certainty to usevisual programming with Grasshopper. This allowedimplementation of generative algorithms withoutthe drawbacks of learning a textual language, whileits open policy enabled access to a great number ofsoftware add-ons for experimentation (Bueno, 2016).

CELLULAR AUTOMATA. Cellular Automata (CA) is asystem that operates locally, as the state of the neigh-bours’ cells, in each interaction, defines the state ofeach cell. John von Neumann and Stanislaw Ulam


Figure 2Cellular Automata(CA) proposals,Estranging TheContext,architecturalelements thatinhabit the moviesfilms, group:Thatilane Loureiro,David Mendonça,Eugênio Moreira(Left), and TheVirtual Cocoon athree-dimensionalCA that produces afour-dimensionalobject to beexplored in VirtualReality, group:Nicolle Prado,Isadora Tebaldi,Emilio Marostega(right)

developed CA in the 40s, and, since then, have beenapplied in a wide range of fields, such as ComputerGraphics or Cryptography. Despite CA’s ability to cre-ate form, it has few applications in architecture. Ac-cording to intuition, a system with simple inputs us-ing simple rules would produce simple behaviour.CA defies this conception, as it can produce acrossthe range from simplicity to emergent complexity, asdemonstrated by Wolfram (2002). CA can produceemergent and complex results.

In the workshop, CA’s high degree of unpre-dictability required introduction of the idea of Form-Making, instead of Form-Finding. This is an alterna-tive to contemporary rational trends that argue that

simple quantitative criteria - such as environmen-tal performance, material use, and cost reduction- should determine the architectural form. (Carpo,2012). We invited the CA participants to analyse thespatial and aesthetic qualities of each configuration,in each recursion, and how they interacted with thesystem. There is no feasible way to foresee exactlyhow to change a rule to affect the product. There-fore, designers interact in a ludicwaywith the systemdiffering from Albertian total control of the drawing.Participants approach the form-making algorithm ina playful horizontal way, supported by previous re-search about CA and computational morphogenesisto handle complex phenomena (Sardenberg, 2013).


CAResults: The approachdiverged from simulat-ingphysical reality to focusondesignas cultural prac-tice. The first team used CA form-making to producearchitectural elements to inhabit cinematographiccontexts. Each project was conceived according toa specific genre (i.e. science fiction, comedy and hor-ror), as an installation that responds to the film andchanges it simultaneously. The second team, “Vir-tual Cocoon”, used three-dimensional CA to producea four-dimensional object that interacts with VirtualReality. Usually we use a series of cubes side-by-sideto represent 3D-CA. However, VR enables the userto interact with virtual objects according to his/hermovement, in an immersive cocoon.

L-SYSTEMS. LS research focuses on literature, but es-pecially on its technical implementation. LS are sym-bolic systems capable of generating growth struc-tures, based on the ability to rewrite rules recur-sively. The biologist, Aristid Lindenmayer (1925-89)invented LS to describe the growth of simple species,such as bacteria and algae. Three concepts are em-bodied in this technique: (i) the initial axiom or seed,which represents the initial state of the system; (ii)the production rules applied to transform the seed,through (iii) recursion, a repetitive computationalmethod that, in each generation, recalls the previousone.

Originally, LS are formal deterministic systems,as the rules are context-independent. Being deter-ministic, there is one, and only one, substitution rulefor each letter of the alphabet. In order to extendthis technique, we searched for methods in visualprogramming to develop context-sensitive LS, allow-ing them to evolve from generating a single artificialplant to different species over time. The referencefor LS design is the book by Prusinkiewicz and Lin-denmayer (1990), and the articles by Fisher & Herr(2000) and Agkathidis (2015). As practical experi-ments, we sought diverse research, but especially thepractical experiments by Coates (2001). While work-ingwith Grasshopper, we evaluated different LS add-ons like Rabbit (Dimitrova, 2016), which has an in-terpreter that translates the code into turtle graph-

ics. We found that this interpreter accepts only lettersymbols as input and is deterministic. Therefore, welooked for recursive loop applications that enablednon-deterministic or stochastic LS. Hoopsnake, citedby Sedrez (2013), was not very intuitive and had lim-itations. We also tested Loop (Turiello, 2013), and, fi-nally, Anemone (Zwierzycki, 2015). The latter is themost intuitive; and during loops maintains the datastructure and has an expandable list of data paths,enabling parameter value change in each genera-tion. Anemone enabled response to external factorsin each generation, implemented with turtle move-ment.

LS Results: The first proposal developed in theworkshop was “Menger Revisited”, and the second,“Hugging Trees”. The possibilities explored are com-plementary. The first revisits the Menger sponge,proposing interactive, recursive transformations: thefractal changes according to the user, in a virtualspace. It went beyond the classical Menger spongeby breaking the symmetry in unpredictable, yet rela-tional, ways, without losing self-similarity. The sec-ond proposal develops tree-like structures that re-configure themselves according to human position,changing the branch angles and planes of rotationaccordingly. The percentage of randomness that af-fects tree movement is associated with external fac-tors. When the user reaches the centre, the trees em-brace him/her. The first solution has potential for VRand can output data for augmented reality. The sec-ond, despite following tree stereotypes, possesses afeasible mechanism for a physical articulated struc-ture.

GENETIC ALGORITHMS. A finite set of rules andoperations defines GA that simulates the combina-tion of individual characteristics of the same species,to select those with the best environmental fitness.The algorithm’s structure considers the main mecha-nisms present in the evolution of species as geneticinheritance, random variation (crossover and muta-tion) and natural selection.

We relate GA with the broad use of engineeringto optimise structures and components. The use of


Figure 3Menger Revisited,showing recursivesystem variationstriggered by userproximity. Group:Fernando Lima,Aurélio Wijnands,Maria Eloisa.Hugging Trees,showing the fractalmovement of treesas the userapproaches. Group:Núbia Gremion,ErickBromerschenckel,Daniel Wyllie.

GA in design is frequent in spatial planning optimisa-tion and form generation. Although it is an optimi-sation method, we can use it as a generator mecha-nism to assist the designer in exploring the solutionspaces, to obtain creative, emergent and unexpectedresults. In this context, the algorithm is set to obtainfavourable solutions independently of the optimi-sation level attained, guaranteeing a flexible choiceamong the solutions generated. Themain referencesareHolland (1995) Bentley (1999) andMitchell (1999).Grasshopper has a genetic solver, Galapagos. How-ever, the number of design problems it can tackle islimited (Rutten, 2013). We used the add-on, Biomor-pher (Harding, Olsen, 2018) that allows the designerto interact with the algorithm during its execution.The designer selects the “parent individuals” fromquantitative (optimal) and qualitative (aesthetic) cri-teria, directing the evolutionary process, as the “indi-vidual parents” will be crossed to generate “individ-ual offspring”.

GA results: The projects developed were “Evolu-tionary Aggregation” (EA) and “Dwell Debris” (DD).

They adopted as criteria: shade, contact with theground and formal arrangement. As a strategy forstructuring theGA, EAdistributed the components ina regular 3D grid, ensuring modularity and orthogo-nality. The DD authors opted to anchor componentsin a random cloud of points, creating an irregular ar-rangement. In both situations, the teams subtractedthe original grids with additional voids, but using dif-ferent components. EA defined a single componentwith freedom of rotation on the central axis. DD de-fined four components that rotate on their axis, allgenerating a diversity of solutions. EA programmedthe GA to find solutions that had the largest con-tact among components and the largest projectedshadow. TheDD teamsearched for thegreatest num-ber of intersections amongcomponents, thegreatestvolume on the ground and the least shade. The par-ticipants used the add-on, Biomorpher to cause evo-lution of formal solutions with human intervention.

SHAPE GRAMMARS. SG uses encoded abstract for-malisms that limit its use to specialists. These for-malisms also limit free interaction in the form of de-


Figure 4Group DD strategyand solutions.Group: Anael Alves,Loan Tammela,Felipe Lannes.Group EA strategyand solutions.Group: LucianaGronda, IgorMachado, MoniqueCunha, WellidaCoelho.

velopment. SGs have different natures as descriptiveand generative grammars (Garcia, 2016). To bridgethis gap, we developed a more intuitive interface invisual programming, relying on earlier Stiny texts tostructure the definition. We identified a recurrent de-scription that we organised into the triple ordinate:

G = (v, r, d)(1) (1)

where G is grammar, v vocabulary, r rules set, and dderivation. The algorithm concatenates three sets,maintaining each variable’s internal independenceand the set’s co-ordination. The loop uses v as input,applying the r rule indicated in d, accumulating theresult in v. While v and d are simple lists, r is morecomplex. The initial axiom has two shapes, settingthe initial and final form, memorising the Euclideantransformations, and then applying them as inputshape. As holdbacks, the definition operation can-not recognise the shapes used. The next step wouldbe recognising emergent forms. The algorithm’s sim-plicity encouraged participants to develop their owndefinitions.

SG Results: The first team used SG that relates afaçade composition with music notation, while theother designed a shelter. They adapted the initial al-gorithm, exploring the generation of diverse designs.If we know the result of a rule application, it is easierto predict the emerging phenomena. Thus, in the SG

definition, we knew the initial and final form, whichfacilitated the interaction. The derivation played afundamental role in SG. However, SG teaching fo-cuses on shapes and rules, how to change the deriva-tion that needs to incorporate uncertainty that re-mains obscure. Full randomness probably leads tomeaningless design, and some randomness can stillgenerate meaningful results. This requires encodingthe derivation data, organisation (data tree) and therandom derivation paths.

RECURSIVE LOOPSBeings evolve interacting with the environment ina continuous process in time. For generative sys-tems to evolve, in addition to interaction and selec-tionmechanisms, we need recursion. The creation ofsuch repetitive cycles was limited to thosemasteringformal computation. By introducing recursion in dif-ferent design situations, using visual programming,we pointed out how systemic interaction can gener-ate evolution.

TECHNIQUES AND INTERACTIONThe interaction between CA rules and the contexthappens after the form generation, when the usercan interact in virtual reality with the CA. In LS, weopen the system boundaries to accept interference


Table 1GenerativeTechniques inDesign (CA,LS,GA,SG): Purpose,Scope, Challenges,Outcomes andInterference.Interference isbased on the resultsof the empiricaltests of theworkshop,envisioning futurepossibilities.

between the user and the context, in every gener-ation, using scholastic algorithms. In GA, we over-came the limitation of a blind system of blind optimi-sation by allowing the user to interfere in each gen-eration, introducing qualitative criteria that changeevolution. In SG, we broke the closed system, allow-ing the user to affect, not only initial and final form,but also interfere with the derivation rules.

Visual programming contributions

• A cellular automaton including a library ofcontext-specific objects;

• 4D-CA with VR visualisation;• Responsive stochastic L-systems;• Multi-objective evolutionary optimisations

(geometricmax shade optimisation), possibil-ity of qualitative decisions during iterations;

• Automation shape grammars (Grasshopper);• A 3D graphical music grammar.

However scarce, the workshop results reveal the gapbetween tools and processes. Singh Gu (2012) com-pared generative tools to integrate them in a “Com-putational Design Framework”, a valuable contribu-tion to research. However, besides the literature sur-vey, is necessary to test them in context. We usedempirical knowledge to fuel design thinking aboutboth tools and design processes. Abovewe present asynthesis, explaining howwe interfere with the toolsrules, envisioning future possibilities.

GENERATIVE TECHNIQUES ANDDESIGNRESULTS DISCUSSIONThe techniques described in this article have existedsince the 80s, but their implementation is slow. Pre-vious formal approaches failed to translate the toolsinto design methods. We proposed a frameworkgrounded onDesign Thinking and complexity, to im-plement contextual interference with recursive cy-cles. Finding an alternative to formal computation,


using loops in visual programming, we show howto interact with tools using recursion. Anemoneprovedadequate for recursiveprocesses, inbothdatareplacement and incremental growth, allowing de-signers to interfere between the cycles. Biomorpherdemonstrated the value of human interaction withan evolutionary process, working as input of qualita-tive criteria. We found recursion and selectionmech-anisms have potential to develop systems that mayevolve in the desired direction using Visual Program-ming. However, this requires technical improvementof the loopprocesses and further development of thesolution selection mechanisms.

With the increase in Information. the problemspace augments. With the introductionof interactionin time, with a selection mechanism, and enablingthe system to learn in the future,weexpect to enlargethe solution space. This would follow the Von Neu-mann self-replicating machines that can extend AI.Finally, the generative tools are several algorithmicmethods,with their specificities. Wemust tame thesealgorithms and go beyond the dichotomy betweentechniques and methods (tools and processes). Onlythen can these algorithms be absorbed, both by thedesign thinking and by the maker culture.

Facing the fourth Industrial Revolution, this fu-sion of techniques reinterprets old methods: by in-troducing direct interaction and an increasing num-ber of context-related variables, generative systemscan deliver responsive designs. Generative methodscan offer a symbiotic partnership between the de-signer and the system to expand architectural abili-ties skills. In the light of the fourth revolution, theycan intertwine information, whether digital, physicalor, in the future, biological.

Form Finding andGenerative SystemsLAMO Seminar/Workshop Form Finding and Gen-erative systems, Rio de Janeiro Federal University,28 Aug to 7 Set 2017, Coordination Gonçalo CastroHenriques, Andres Passaro and Elisa Vianna. Tutorsgenerative tools: Victor Sardenberg, Ernesto Bueno,Gonçalo Castro Henriques, Jarryer de Martino and

Daniel Lenz. Our gratitude to PROURB for support-ing the event and to all participants and collabora-tors, we cannot name here due to space limitations.

REFERENCESAgkathidis, A 2015 ’Generative design methods imple-

menting computational techniques in undergradu-ate architectural education’, eCAADe 2015, pp. 47-55

Buchanan, R 2006, ’Wicked Problems inDesign Thinking’,Design Issues, 8, pp. 5-21

Carpo, M 2011, The Alphabet and the Algorithm, Cam-bridge, Massachusetts, MIT Press

Fischer, T and Herr, CM 2001 ’Teaching Generative De-sign’, Conference on Generative Art, Milan, pp. 1–11

Garcia, s 2017, ’Classifications of Shape Grammars’, inGero, J (eds) 2017, DCC, Springer, pp. 229-248

Gero, JS 1996, ’Creativity, emergence and evolution indesign’, Knowledge-Based Systems, 9, p. 435 448

Harding, J and Brandt-Olsen, C 2018, ’Biomorpher: Inter-active evolution for parametric design’, InternationalJournal Architectural Computing, 16(2), pp. 144-163

Holland, JH 1995, Hidden Order: How Adaptation BuildsComplexity, Basic Books

Martino, JA 2015, Algoritmos evolutivos como métodopara desenvolvimento de projetos de arquitetura,Ph.D. Thesis, Universidade Estadual de Campinas

Mitchell, M 2009, Complexity: A Guided Tour, Oxford Uni-versity Press

Rutten, D 2013, ’On the Logic and Limitations of GenericSolvers’, Architectural Design, 83, pp. 132-135

Sardenberg, V 2013, Processos Emergentes em TerritóriosInformais, Master’s Thesis, Mackenzie University

Schwab, K 2017, The Fourth Industrial Revolution, Pen-guin Random House.

Sedrez, M and Martino, J 2018, ’Sistemas generativos’,in Celani, G (eds) 2018, Arquitetura contemporâneae automação: Prática e reflexão, Probooks, pp. 25-28

Simon, H 1969, The sciences of the artificial, MIT PressSingh, V and Gu, N 2012, ’Towards an integrated gen-

erative design framework’, Design Studies, 33, pp.185–207

Stiny, G 1980, ’Introduction to shape and shape gram-mars’, Environment and Planning B, 7, pp. 343-351

Terzidis, K 2006, Algorithmic Architecture, RoutledgeVenturi, R 1996, Complexity andContradiction inArchitec-

ture, The Museum of Modern Art, New YorkWeaver, W 1948, ’Science and complexity.’, American sci-

entist, 35, pp. 1-12Wolfram, S 2002, A New Kind of Science., Wolfram Media


generativesystemspapers.cumincad.org/data/works/att/ecaadesigradi2019_100.pdf · creativity:...

Documents