Generative Design Methods

Implementing Computational Techniques in Undergraduate Architectural

Education
Asterios Agkathidis1
University of Liverpool
1
[email protected]
1

In continuation to the Deceptive Landscape Installation research project

(Agkathidis, Kocatrk 2014), this paper investigates the implementation of
generative design techniques in undergraduate architectural design education.
After reviewing the main definitions of generative design synoptically, we have
assessed the application of a modified generative method on a final year,
undergraduate design studio, in order to evaluate its potential and its suitability
within the framework of a research led design studio, leading to an RIBA
accredited Part I degree. Our research findings based on analysis of the design
outputs, student performance, external examiners reports as well as student
course evaluation surveys indicate a positive outcome on the studio's design
approach, as well as its suitability for an undergraduate design studio. They
initiate a flourishing debate about accomplishments and failures of a design
methodology, which still remains alien to many undergraduate curricula.
Keywords: CAAD Education, Generative Design, Design Concept

INTRODUCTION: DESIGN METHODS IN ARCHITECTURE

Generating form poses one of the fundamental questions in architectural education and practice. Architectural production is frequently accompanied by debates about the legitimacy of its design approach,
questioning the relationship between function and
form, aesthetics and construction systems, context
and structure, user needs and construction costs, in
all possible congurations. In recent years, computational tools have introduced innovative form-nding
techniques, revolutionizing architectural design and
production. These techniques are often described by

terms such as 'generative design', 'parametric design'

or 'algorithmic design', to name but a few. These offer new design paths to architects by breaking with
predictable relationships between form and representation in favour of computationally generated
complexities, thus enabling the development of new
topologies. They shift the emphasis from 'form making' to 'form nding' (Kolarevic 2003). The critics of
such design approaches claim that they disconnect
architectural output from its context and its users,
and lead to a decrease in spatial quality and a building's integration within the urban environment. Furthermore, some argue that a totally computerized
approach leads to disconnection from physical mod-

CAAD Education - Concepts - Volume 2 - eCAADe 33 | 47

elling and drafting techniques - once essential foundations of architectural education - and so risks the
loss of material qualities, eects and properties. Yet
various generative form-nding techniques existed
in architecture long before the digital revolution. At
the start of the twentieth century, many visionary architects, engineers and designers, such as Frederick
Kiesler and Frei Otto, were applying design methods
that were very similar to today's computational approach. It seems that today's new computational design techniques are not as new as they seem, nor impossible to practise without the use of computational
tools. So is it the tools or the design method that
should be targeted by critics of so-called digital architecture? The following paper will try to cast light
on that perpetual conict.
Why, though, should anyone follow a method
for designing architecture in the rst place? Can architects not simply rely on personal inspiration or
their own sense of beauty? Throughout the evolution of architectural design there have certainly always been schools of thought that have encouraged
a design process based on inspiration or an initial
stimulus. However, others haven promoted adherence to a specic design method, based on rules
rather than intuition, and many now argue that design methods are necessary in order for architects to
deal with today's hyper-complex design briefs, or to
prevent self-indulgence and stylistically driven formal language. Others arm that emerging computational design and fabrication tools are changing the
architect's role, making design methods a necessity.

DEFINITIONS OF GENERATIVE DESIGN

There is no single denition of the term, but many
complementary denitions with common characteristics, which vary according to dierent architectural
theorists. Overall it can be described as a design
method where generation of form is based on rules
or algorithms, often deriving from computational
tools, such as Processing, Rhinoceros, Grasshopper
and other scripting platforms. During the late 1980s
and early '90s, just before the boom of computational

48 | eCAADe 33 - CAAD Education - Concepts - Volume 2

architecture, Peter Eisenman started applying a set

of design techniques, such as scaling, fractals, overlay and superposition, inuenced by Jacques Derrida's Deconstruction theory (1976). Eisenman applied these techniques in relation to rules of order,
developing several projects on this basis, such as the
Biocentrum in Frankfurt and the Nunotani Corporation headquarters in Tokyo (Eisenman 2004). One
could claim that his design method was the rst contemporary generative design attempt. As software
started to oer new possibilities, Eisenman introduced other techniques to his approach, such as morphing images, which was soon followed by UNstudio and their concept of the 'Manimal', a computergenerated icon that merges a lion, snake and human
to represent the hybrid building (van Berkel and Bos
1999).
The technique of folding appeared in Eisenman's
Rebstockbad in 1991 (Eisenman 2004) and, as computational tools advanced further, Greg Lynn started
applying new tools such as animation, splines, nurbs
and isomorphic polysurfaces, inuencing a whole
wave of architectural production, often described as
'blob architecture' (Lynn 1999). As algorithms and
scripting become more accessible to architects and
designers, and digital fabrication more aordable,
parametric and panelization tools, simulation software, optimization and generative algorithms are
dominating today's generative design techniques.
In their book Generative Gestaltung (Lazzeroni,
Bohnacker, Gro and Laub 2009) the authors dene
generative design as a cyclical process based on a
simple abstracted idea, which is applied to a rule
or algorithm (gure 01). It then translates into a
source code, which produces serial output via a computer. The outputs return through a feedback loop,
enabling the designer to re-inform the algorithm and
the source code. It is an iterative operation, relying
on the feedback exchange between the designer and
the design system.
Celestino Soddu (1994) denes generative design as 'a morphogenetic process using algorithms
structured as nonlinear systems for endless unique

Figure 1
Generative design
process diagram, by
Lazzeroni,
Bohnacker, Gro
and Laub

and unrepeatable results performed by an idea-code,

as in nature'. Indeed, the notions of generative design and digital morphogenesis are strongly associated. The term 'morphogenesis' derives from the
Greek words morphe (, meaning 'form') and
genesis (, meaning 'birth'), so could be literally
translated as 'birth of form'. As with the term 'generative design', there is no unique denition for morphogenesis, and it seems that the terminology is changing in relation to emerging technologies and techniques.
Branko Kolarevic (2003) describes digital morphogenesis as follows: 'The predictable relationships
between design and representations are abandoned
in favour of computationally generated complexities.
Models of design capable of consistent, continual
and dynamic transformation are replacing the static
norms of conventional processes. Complex curvilinear geometries are produced with the same ease as
Euclidean geometries of planar shapes and cylindrical, spherical or conical forms. The plan no longer
"generates" the design; sections attain a purely analytical role. Grids, repetitions and symmetries lose
their past raison d'tre, as innite variability becomes
as feasible as modularity, and as mass-customization
presents alternatives to mass-production.' In addition, he sees such methods as unpredictable mechanisms of creation, relying on digital tools, where
traditional architectural values are replaced by complexity, asymmetry, curvilinearity, innite variability
and mass customization. Architectural morphology

is focusing on the emergent and adaptable qualities

of form. Form is no longer being made, but found,
based on a set of rules or algorithms, in association
with mainly digital, but also physical, tools and techniques. They imply the rules; the entire process follows.
Michael Hensel describes digital morphogenesis as a 'self-organization process, underlying the
growth of living organisms, from which architects
can learn' (Hensel, Menges and Weinstock 2006). In
their latest book, Rivka and Robert Oxman (2013)
categorize form generation into six dominant models in relation to its main driver: mathematical, tectonic, material, natural, fabricational and performative. They see digital morphogenesis as 'the exploitation of generative media for the derivation of material form and its evolutionary mutation'. Its key concepts include topological geometries, genetic algorithms, parametric design and performance analysis.
Finally, Toyo Ito compares 'generative order' to
the growth mechanism of trees, whose form derives
from the repetition of simple rules, creating a very
complex order (Turnbull 2012). A tree's shape responds to its surroundings, blurring the boundaries
of interior and exterior spaces - qualities that are easy
to recognize in Ito's Serpentine Gallery Pavilion in
London.
Based on these key denitions of generative design, a modied design method was developed in order to be applied in a year 3 design studio module at
the Liverpool School of Architecture.

CAAD Education - Concepts - Volume 2 - eCAADe 33 | 49

AIMS AND RESEARCH QUESTIONS

Main aim of this research paper is to evaluate the
suitability and implementation of generative design
methodologies into undergraduate design studio. Its
main research questions can be described as follows:
How can generative design methodology be
integrated in an undergrad design studio
module?
Which are its strengths and weaknesses?
Does the applied design methodology
achieve its objectives in producing innovative design solutions and increase students'
design skills and future employability?
Does the integration of generative design
methodology in undergraduate level seem
appropriate?

DESIGN STUDIO MODULE TEACHING

FRAMEWORK AND RESEARCH METHOD
APPLIED
Our research method was based on monitoring the
design path of Analysis - Morphogenesis - Metamorphosis (Figure 02) aiming the production of complete
architectural proposals. It was applied in a research
led design studio unit of 57 students (studio 04). The
studio was led by the author, tutored by four additional tutors and it is part of a 230 student cohort in

the entire year. The design module had a teaching

period of 12 weeks. Students were allowed to choose
among the four studios oered, according to the different studio briefs and teaching methods.
Analysis focused on data collection of various aspects, such as context, programme, material, structure and performance. Morphogenesis, digital or
physical, targeted the generation of abstract prototypes, based on spatial and organizational principles, including unit accumulation, surface continuity, faceting, volume deformation or subtraction, and
algorithmic patterning. The phases of Analysis and
Morphogenesis have certain similarities to those described by Paredes Maldonado (2014), however deriving from a dierent starting point and emphasizing strongly on physical model outputs. In many
cases notions such as genetic algorithms, cellular automata or shape grammars as described by Fischer
and Herr (2001) were applied by using physical modelling only, according to the students individual skills
and preference.
After undergoing a phase of iterations, the resultant not-to-scale 'proto-tectonic' structures (Frampton 2001) were then 'transformed' into building proposals, including oor plans, sections and elevations,
as well as physical and digital models and all their
derivatives (atmospheric images, visualizations, etc.).
The Metamorphosis of an abstract prototype into a
building proposal could occur in two dierent ways.
Figure 2
Generative Design
Method applied in
Studio 04

50 | eCAADe 33 - CAAD Education - Concepts - Volume 2

It could either be literal (a direct transformation into

a building envelope and structure) or operational
(functioning as an apparatus, which could generate
design solutions in various scales and arrangements).
Studio marking consisted out of two components, split between 30% for research and 70% for
design. In Studio 04, the studio presented here, students had to work collaboratively during the rst four
weeks allocated to research (analysis and morphogenesis) and develop their individual projects in the
remaining time of eight weeks. Students were offered introduction workshops in Rhinoceros 5 and
Grasshopper as well as in digital fabrication techniques. Each group was given a precedent to use as a
starting point for their analysis, from which they had
to derive design principles and rules in order to produce iterations of abstract physical prototypes.
The remaining 70% design component was dedicated to the design of a middle sized project (in that
case a regional ferry terminal). Students were asked
to 'transform' and apply their research ndings from
the 30% component into their nal design proposal.
Deliverables were consisting of all typical drawings,
physical models, visualizations and detailing, as required by the RIBA and were identical for all four studios oered.
All projects were reviewed three times by external juries before being marked by the entire group of
studio tutors, moderated internally and examined by
external examiners. All data was collected through
out the entire year from student assessment and
feedback forms, external examiner reports, submission of design project data, as well as an anonymous
student survey. Marking criteria for both studio components (30% and 70%) were co-decided by all studio leaders and were made available to all students
and examiners. In particular the marking rubric for
the 30% research component was composed out of
following marking criteria:
Quality of site investigations / design exercises
Evidence of research into architectural precedents

Evidence of design drivers / process work /

strategy
Clarity of architectural arguments / overall
presentation
Marking criteria for the 70% design components
were as followed:
Development of design process, utilising ndings from component 1 / innovation
Architectural agenda and design approach,
expressed through diagrams, sketches, models.
Quality of layout, drawings, 3d exploration,
models
Quality of tectonic approach and structural
considerations
Quality of the 1:20 detail drawings
Overall body of work and presentation
Analysis of the monitored design output, marking
rubric statistics and student survey, which will be presented in this paper are oering an analytical evaluation overview of generative design methodologies integration in undergraduate education. Obviously a decrease in students' average marking, or
negative comments by external examiners and students would be a strong indication that the teaching method applied is not delivering the expected
results, thus may not be suitable for undergraduate
design studio education and vice versa. External examiners reports are based on the examiners annual
review of the cohort's studio projects as well as a 10
minute interview every student. They do not relay on
a specic list of criteria, but on the examiners' long
year experience in practice and education. Student
surveys were held anonymously through 'vital' the
schools on-line operating system.

PRESENTATION OF STUDIO OUTPUTS AND

COLLECTED DATA
By monitoring students' studio choices for the 4 different studios oered, rst positive signals arise. Studio 04 turned out to be very popular, with 1st choice

CAAD Education - Concepts - Volume 2 - eCAADe 33 | 51

demands exceeding the available 57 slots . In addition the studio's cohort had a high marking average
of 64.5%, considerably higher than the other 3 studios (any mark on 70% or higher, is considered as a
1st class project). Overall there were 12 students with
a 1st class average mark, the highest concentration of
rsts in all 4 studios.
Proceeding with the 30% research component
outputs, which included the phases of analysis and
morphogenesis it proved to be very successful due
to the enormous amount and high quality of design
production, including hundreds of models, diagrams,
sketches and drawings (Figure 03). High variety on
geometrical exploration, typology innovation, scale,
geometry and structure has been documented.
Students managed to tackle the studio's design
approach and requirements quite well, even though
it was the rst time they have operated with such a
design approach. This becomes evident by looking at
their marking statistics, putting studio 04 at the highest range with an average score of 69% (Figure 04).
Continuing with the 70% design component,
student performance drops compared to the 30%
component as expected. In comparison to their studio performance in the previous year, the average
studio mark stays almost unchanged moving from
64.5% into 63.5%. Out of the 57 students, 28 improve
their score while 20 worsen it. 13 manage to move on
a higher grading band while 13 move to a lower one.
Band changes of grades moving up from from 63% to
80% and from 55% to 72% are among the most extreme ones. On the negative site were students who
dropped from 70% to 58% and from 60% to 48%.
By looking at the on-line studio 04 survey, results appear very encouraging. Its overall participation rate reaches 65%. In the question 'How did you
nd studio 04's overall design method?', 40% nd it
very useful allowing them to achieve new outputs,
48.6% useful, neutral 8.5 % and none nds it not useful or would rather design the conventional way. That
shows an overall acceptance of 88.6%.
In reference to the morphogenetic component,
the question 'How important were the 1st phase and

52 | eCAADe 33 - CAAD Education - Concepts - Volume 2

the development of the generative models for the design of your nal building?' receives 37.1 % for very important, 54% for important (has inuenced the building in one or the other way), 2.85% has ignored the
generative models and 5.7% consider it as neutral.
Again an overall acceptance of 91.1% occurs.
Figure 3
Model outputs of
morphogenetic
design component

Figure 4
Marking statistics
on morphogenetic
modelling design
component.
Considering the freedom in design creativity, the majority sees the method as a mean to express their creativity with 51.5% strongly agreeing, 34.3% agreeing,
2.9% disagreeing and 11.4% seeing it as neutral.
Referring to taught skills in 3D modelling and
digital fabrication 88.5+8.5% strongly agree or agree
that they are useful for their future employability
while only 3% think this they are irrelevant.
Finally in the question 'Do you think that the new
skills you have developed in this studio helped you to
produce a better project than last semester?' 54.3 %
nd their current project much better than their previous, 31.4% a little bit better, 8.5% don't like their
current project and 2.8% see it about the same.

By monitoring the external examiner's comments and remarks, based on their annual report, studio 4 is not being considered as problematic or inappropriate. They don't seem to dierentiate it from
all other studios which follow conventional design
methods, except for one complimentary statement.
All examiners 'complained' about the lack of context
engagement in the entire year except for studio 04:
Andrew Peckham stated: 'It was ironic that it was the
studio concerned to 'search for new typologies' that best
registered a contextual urbanism on a constrained site
in the city centre'.

DISCUSSION AND CONCLUSIONS

Figure 5
Market Hall
proposal by Yiqiang
Zhao

Figure 6
Ferry terminal
proposal by Man Jia

By reviewing the design projects and outputs assessed here, a set of conclusions and discussion
points arises. Many architectural educators are very
sceptical about such unconventional design methods, fearing the loss of design control, materiality,
craftsmanship, functionality and relationship to context. They tend to blame generative-based design
for all the negative aspects of contemporary architecture. However, the design method applied here
proves them wrong. It manages to support a high
degree of dierentiation between the schemes, despite the fact that many of the projects were developed using the same techniques (e.g., triangulation).
It also inevitably supports creativity and innovation,
which is why so many of the projects managed to
move beyond standard building typologies and layouts such as the market hall building (Figure 05). Innovation emerged not only in formal design aspects,
but also in terms of building programme and spatial
solutions, oering new building type hybridizations,
such as the ferry terminal proposal (Figure 06).
The mix of dierent design tools and techniques,
switching from traditional physical modelling, such
as plaster casting, to advanced 3D printing and CNC
fabrication in one continuous modelling scheme
(Figure 07), proved to be of great educational value.
It oered students the opportunity to test materials
with their hands, and to experience the advantages,
diculties and opportunities advanced technology

has to oer - a design path that is often excluded,

due to dogmatism, or ignorance of (or lack of respect
for) either handcrafting or computerized techniques.
Neither banning computers nor abandoning traditional craftsmanship oers a solution for the future
of architectural education.
Digital tools can often be seductive for designers. However, while speeding up the design process,
designing with digital tools makes gravity and materiality disappear. Physical modelling helps designers
and students to reconnect with these two key elements, which are so important for architectural production. In addition, the switch between analogue
and digital tools allows students to lter out excess
complexity within a digitalized design process. By
testing digital ndings with physical prototypes, they
can begin to assess whether a complex solution is really oering spatial, aesthetic or programmatic qualities to a project. The issue is not so much whether
CAD and 3D modelling software should be banned
or embraced in undergraduate architectural education, but rather to what extend they should be applied, in which educational year and for what purpose. The same principle applies for the use of digital
fabrication. It is indeed irrational to apply such techniques for cutting out rectangular panels, but more
than appropriate for mastering fabrication of complex geometries.

CAAD Education - Concepts - Volume 2 - eCAADe 33 | 53

Figure 7
Generative models
by Yiqiang Zhao,
Zhenyu Zhu and
Nojan Adami

Alongside the loss of materiality and craftsmanship, many critics of generative design methods argue that the resulting architectural proposals are totally detached from their context. This is a criticism
often applied to modern architecture as well. During our programme, the degree of integration or nonintegration within a context was up to the designer.
Building up a relationship between a building and its
context can be achieved in many dierent ways. It
can rely on form, materiality or programme, or all of
the above. One can choose to harmonize, ignore or
break with a building's context, a decision that does
not depend on the design approach but on the designer's attitude towards the site. Nevertheless, external examiners overviewing the projects expressed
surprise at the high degree of site-specic proposals,
despite the unconventional design approach. All of
the nalized projects managed to comply with standards and requirements dened by the accreditation
body (RIBA) and the module descriptors, as evident
in the drawing and modelling outputs. In that sense,
the generative design method applied proved highly
appropriate for design education, helping students
to develop their skills and self-condence, and enhancing their future employability. This became evident by the student survey, where 97% of all participants expressed their condence about gaining
higher employability perspectives.
Looking at the diculties accompanying such an
approach, ndings varied. Scepticism from other col-

54 | eCAADe 33 - CAAD Education - Concepts - Volume 2

leagues and fellow educators was denitely among

them. This included guidance from tutors involved
in the process as well as criticism from others observing the approach. From a student's point of view, the
shift away from conventional design methods certainly appeared to be very demanding. That became
particularly evident after the completion of the formnding phase; it was the Metamorphosis that presented the biggest challenges. As liberating and exciting as Morphogenesis might have been for some,
abandoning the abstraction of the prototype and
transforming it into a building proposal, overcoming obstacles of structure, urban context, planning
and materiality seemed to be very dicult. Many
tended to start from scratch, leaving everything behind and following the conventional approach they
were most familiar with. Some chose a brutal landing
of their prototype into the site, without developing
an attitude towards the context, while others failed
to use their prototype for something more than simply trendy decoration.
Another issue that arose using this approach was
having to manage the geometrical complexity that
occurred.This was often a problem of representation.
Complex geometries are easily produced in a digital
environment, but controlling and representing their
outputs often requires non-standardized methods as
well. This can be a consequence of using software incorrectly, or of using inappropriate software for the
task required. How can complex geometries be rep-

resented in oor plan, section and elevation? How

can such geometries be built in a physical model?
This is when guidance is needed. Tutors play a critical role here. Educators need to guide students and
enrich academic curricula with new design methods
and tutored skills. But it is the student culture that
plays an even more important role: the intercourse
that occurs between students, either through daily
procedures and presentations or through social media, websites and on-line forums. Today's young designers belong to a generation that has grown up
with smartphones and computer tablets instead of
crayons and paper, granting them familiarity with
digital technologies from a very early age. These
are designers, therefore, who are more than able to
deal with unconventional design methods, and who
will hopefully revolutionize architectural production
in the future.

Kolarevic, B 2003, 'Digital Production', in Kolarevic, B

(eds) 2003, In Architecture in the Digital Age: Design
and Manufacturing, Taylor & Francis, London, pp. 4648
Lazzeroni, C, Bohnacker, H, Gross, B and Laub, J
2012, Generative Design: Visualize, Program, and Create with Processing, Princeton Architectural Press,
Princeton
Paredes Maldonado, M 2014 'Digital Recipes: A diagrammatic approach to digital design methodologies in
undergraduate architecture studios', Proceedings of
eCAADe 2014, Newcastle, pp. 333-342
Oxman, R and Oxman, R (eds) 2013, Theories of the Digital
in Architecture, Routledge, London
Soddu, C 1994, 'The Design of Morphogenesis. An experimental research about the logical procedures in
design processes', Demetra Magazine, 1, pp. 56-64
Turnbull, J (eds) 2012, Toyo Ito: Generative Order (Kassler
Lectures), Princeton Architectural Press, Princeton

ACKNOWLEDGEMENTS
My acknowledgements got to all studio 04 tutors,
Richard Dod, Elif Erdine, Jo Hudson and Jane Moscardini for their hard eort and support.

REFERENCES
Agkathidis, A and Kocatrk, T 2014 'Deceptive Landscape Installation: Algorithmic patterning strategies
for a small pavilion', Proceedings of eCAADe 2014,
Newcastle, pp. p71-79
van Berkel, B and Bos, C 1999, Move (3-vol. set): Imagination/Techniques/Eects, Groose Press, Amsterdam
Derrida, J 1976, De la grammatologie, Les ditions de Minuit, Paris
Eisenman, P 2004, Eisenman Inside Out: Selected Writings 19631988 (Theoretical Perspectives in Architectural History & Criticism), Yale University Press, Yale
Fischer, T and Herr, C M 2001 'Teaching Generative Design', International Conference on Generative Art, Milan
Frampton, K 1995, Studies in Tectonic Culture: The Poetics
of Construction in Nineteenth and Twentieth Century
Architecture, MIT Press, Boston
Hensel, MA, Menges, A and Weinstock, M (eds) 2006,
Techniques and Technologies in Morphogenetic Design, Wiley & Sons, London

CAAD Education - Concepts - Volume 2 - eCAADe 33 | 55