In this section, I describe the reflective practice that I engaged in during the development of this dissertation. The way I describe and organise the material of this dissertation and the evidence provided by the design projects is based on object-oriented thinking. Object- oriented thinking is closely related to the systems design methodology that governed the development of Typogenetic Design. My thinking process, therefore, is strongly focused on the “instance or manifestation” [Stamm 2013] of singularities and their assembly into larger systems in a modular fashion. The emerging theoretical framework is determined by both - i.e. instances and their relationships - and as a result, it can be described as singularity, because the reflective evaluation of the subject matter resembles an unique
expression of knowledge [Stamm 2013]. This expression was gathered from “key inten- tions”, “ascension moments” and “key incidents and explicating them to a wider audience” [Stamm and Blythe 2017] during Practice Research Symposia and academic conferences. The project sequence and the experimental sequence advanced my personal experi- ence in a material mode 6 during my PhD process were captured and mapped to gain a deeper understanding of the progression and sequence of the different activities. Through reflective devices built on the observation of, and interaction with, my PhD community, I was able to generate insights from a variety of angles and perspectives with respect to the generated body of work.
My interpretation of design as a communicative process 7 and as a complex adap- tive system 8 led to the exploration of an interactive conversation with the computer-as- machine. The collaborative, iterative creation of solutions that are intended to increase the efficiency of this conversation in the open-minded and conscious exchange of informa- tion between human and machine led to the development of a “reflective practive” [Stamm 2013] to capture the substance9and integral meaning of “conceptual abstraction” [Stamm 2013]10 11
In the context of this research, there are multiple levels of reflective practice:
• project work as case study research in a material mode: reflection leading to new potential for inquiry
• parallel development of computational design systems: reflection unfolded the con- ceptual and theoretical context
• development of a systematic knowledge collection in an adaptive framework for cre- ative design: reflection enabled conceptual mapping with other disciplines 12
6In the material mode of research, creation and designing are used to generate knowledge about the research matter through interaction with the expressive self. The choice of media determines the results of such an investigation. During this PhD research, the material medium used for reflection was the actual software prototype implementing the adaptive framework for creative design, as well as case studies that manifest some of the insights encountered during the PhD process.
7The communication with another person, community or medium to gain additional insights in the nature of a design problem and in succession reducing the problem space continuously until the design problem is resolved.
8
I argue this point in reference to the constant readjustment of requirements, constraints and param- eters during exploration of design alternatives, often undertaken in an agile fashion.
9
Referring to the material outputs and creative explorations at the heart of the research matter. 10
The abstraction process during the research establishes the necessary distance to the subject matter necessary for reflective practice. [Stamm 2013]
11
As discussed by Hegel in his seminal work ‘The Logic’ [Hegel 1969], the process of conscious reflec- tion correlates the observation of specific phenomena with their essence. This change from implicit to explicit knowledge as the thought process moves toward the stage of essence allows one to recognise the identification of the object. The means for this process is the postulation of terms and the correlation of pairs.
12
Simultaneous development of myself as a creator expanded my knowledge and skills based on the nature of the enquiry. Besides the development of digital tools, constant reflection on the creative pro- cess revealed new methods for the development of interactive computational systems in the context of architectural design.
Figure 3.5: The hybrid research process in application
Reflection, in this context, has a variety of characteristics - one of them being the organic nature 13 of the “essence” [Stamm 2013] 14. The emerging understanding of the research object during the reproduction of a certain mode of design under the influence of different circumstances or influences15was regularly reassessed during periods of reflection
16. The reflective process, in the context of the practice-based research programme at
RMIT University, includes the continuous development of projects - in this PhD study, case studies - to uncover coherent themes and repeating motives in architectural practice. A diagram of the PhD process, based on the framework of Marcelo Stamm and Richard Blythe [Stamm and Blythe 2017] and adapted to the hybrid research methodology used in this study, is shown in Figure3.5.
Uncertainty of outcomes is a feature shared by architecture and scientific research in information-based societies. Creative design in architecture, being the dominant mode of expanding the body of knowledge in architectural design research, is based on product- oriented thinking independent of applied research methods. Framing design research in projective thinking reveals the real nature of architectural research, which is concerned with the materialisation of design in the built environment [Hight 2004]. Architectural design addresses the natural, cultural and historical progression of the built environment. Those aspects vary depending on the personal background of the designer, and lead to a subjective understanding of instances and their relationship in a conceptual framework. This conceptual framework is more like a “grammar book” than a revealing of hidden aspects [Stamm 2015].
13
Organic nature in this context refers to the growing, changing and transforming description of the phe- nomena encountered during the research. The concepts and ideas that want to express the understanding of those phenomena evolved during the research.
14
The essence of the research amounts to the aggregate of the descriptions of phenomena. This term also refers to a distilling of knowledge during a period of inquiry. Here, different modes of practice lead to different perspectives on the research, and therefore allow one to gain an understanding of concepts and models that describe aspects of the essence. The expression of essence is something that must be communicated verbally, since it is incapable of being simply visually expressed [Stamm 2013]
15The phenomenology talks about other things in this context. In the case of this PhD research, the object of enquiry is the mode of design in the context of human-computer-interaction for designing interactive computational design systems.
16During the active process of reflection, the adaptation of the observed object to the circumstances of different environments reveals different characteristics or properties of the object.
Understanding Instances and Materialisation The personal experiences during working with the literature and the case studies led to the realisation of creative artefacts and materialisation 17, which I captured by means of language to provide an explicit understanding for the design of aesthetic decision support systems 18 As all knowledge creation is a cumulative process, the definition of source terms is critical to build explicit knowledge. In the case of this PhD study, the encountered terminology consists of terms and concepts rooted in the discourses of architecture, computer science, engineering and systems theory.
Source Terms By identifying relevant linguistic expressions and a strict exploration of the range of definitions in the multi-disciplinary research space, I developed the conceptual basis as a “cognitive scaffold” [Stamm 2013] for this PhD study. Those terms made it possible to reason about the theoretical implications of the phenomena encountered during the present research, and they were strategically exploited during phases of reflective thinking about the activities conducted during the research. At certain stages, repetitive utterance 19 of research content at the Practice Research Symposium (PRS) provided a means to identify the meaning of particular phenomena in the context of the research projects. At other points, the communication process engaged during the PRS helped me to extract meanings from terminology that were only present in my mind as implicit knowledge beforehand.
Contrasting and Differentiation In some cases, understanding the meaning of a particular term was facilitated by contrasting it with other similar or distant concepts
20 The subsequent branching of conceptual understanding during reflective practice as an
“ordering technique” can be seen as a “mapping exercise” in reference to Marcelo Stamm’s use of this term in [Stamm 2013]. The self-similarity of systems on different levels allowed for frequent reinterpretation of the conceptual branching derived from contrasting and differentiation by mapping them from one scale to another 21.
Interpreting Key Incidents and Ascension Moments The interpretation of the key moments of the interrogative practice engaged in during this PhD research was crucial
17These selected aspects of the subject matter might be referred to as “phenomena” in the phenomeno- logical sense.
18Language in this context is used as both a tool for abstraction to gain distance to the subject matter and to adequately describe the core of the research (its essence).
19This is sometimes considered a tautology. 20
This process is called ‘dichotomy’. It allows one to continuously differentiate terms in a network or tree, identifying fine nuances between ideas. Polar or similar phenomena sometimes needed to be established as dichotomies to allow me to position a set of ideas on either a binary or continuous spectrum of notions. Therefore, dichotomies were sometimes used to explore intermediate phenomena that would have otherwise been left undiscovered.
21
And therefore from one set of phenomena to another. As a result, the meaning of those dichotomies depends on the spectrum of their application to generate meaning.
to identifying the meaning and relationships between different objects and instances of objects22 A fruitful approach for the visual understanding of the reality of a designer’s
mind was the mapping practice conducted continuously during the development of the research 23. The chosen approaches ranged from mapping phenomena in system and control diagrams to mapping knowledge in a community of practice review 24. Another interpretative approach involved reviewing my previous creative work and the successive extraction of different motives, urges and fascinations in response to my research progress
25. As a central aspect of my thinking, interpreting my previous actions (in designing),
acting (in presentations) and reasoning (about my research matter) allowed me to position the research project within the relevant community of architectural practice.
22
Different hermeneutic methods were used to interpret the encountered phenomena, their relationships and their association to different scientific discourses.
23
Reflection relates to a visual phenomenon in nature [Stamm 2013]. Thus, a generation of visual representations supports the search for a verbal expression of the subject matter and its constituent parts.
24
This community of practice review can be seen as a conceptual and relational diagram that tries to understand the connections and impacts of individuals in the context of the subject matter.
25
I developed a fascination with computational and parametric design and the relationship between biomimetic principles and design tools during my formal studies. The use of optimisation tools was an urge I felt in response to discussions about sustainability that I encountered in architectural practice. Structure, as defining element of architecture, and tectonic articulation were identified as core elements of my emerging practice.
CHAPTER
4
Theoretical Framework
“As an architect, you design for the present, with an awareness of the past, for a future which is essentially unknown.”
–Norman Foster
In this chapter, a theoretical framework 1 for aesthetic decision support is reported and its modules described in detail. This descriptive account of the theoretical framework allows the reader to understand the background of Typogenetic Design. Typogenetic Design could be applied to some fields of practice - for example structural engineering or architectural design - to facilitate performance-based computation. The algorithms used for the evolutionary search could also be exchanged so that the adaptive framework for creative design might be translated to other artificial intelligence (AI) technology. I present this Typogenetic Design as a contribution to architectural design. Algorithmic variants are beyond the scope of this dissertation.
Intelligent user interfaces combine the use of AI and human-computer-interaction (HCI) usually in software systems that support users, decision-makers and designers. Intelligent user interfaces can be used to build extended DSS. Active tools are extended DSS. “Extended support involves an explicit effort to influence and guide decision making, while respecting the primacy of judgement [...].” [Keen 1987] This category of DSS aims to provide consultancy by supporting selection activity and highlighting well-performing solutions in contrast to limiting the activity of the computational agent to the analysis of solutions chosen by the user. [Keen 1987] The theoretical framework provides the terms, models and methodologies for the design, development and construction of extended DSS.
1
I understand the word “framework”, as used in this dissertation, as a scaffolding continuously built over the research process. As extensively discussed in [Heidegger 1954] in connection with the term “Ge- Stell”, describing a challenging claim that drives the emergence of a collection or assemblage that contains the richness of conceptual relationships of reality observed in research. This framework is a result of the revealing challenge that I promote the reader to accept by unfolding the inventory present in this chapter.
Adaptive user interfaces go full circle from monitoring of user interaction to adapt- ing the decision-making component to this interaction. By adopting this architecture in contrast to adaptable interfaces in Figure4.1, the system behaviour of the intelligent user interfaces developed during this PhD study exhibits the adaptive behaviour necessary to support interactive Generative Design (GD). A designer’s decision paths are not always linear, but tend to be iterative and sometimes tangential, so that an adaptive user interface is critical to providing the flexibility, adaptivity and usability required for the application in architectural design.
Figure 4.1: Adaptable and adaptive user interfaces adopted from [Stephanidis et al. 1998] The creative DSS in Typogenetic Design augments genetic programming as adaptive process for shape generation [Muehlbauer et al. 2017a] by adding capabilities for aes- thetic decision support. Genetic programming as a population-based algorithm uses the mechanisms of mutation and recombination to modify sets of solutions over time during advancing generations. An inital population is evaluated for its fitness (e.g. performance measures like building sustainability and costs). The requirement for assessment of the fitness function, which can consists of multiple design criteria is grammar translation. In Typogenetic Design grammar translation is facilitated using shape grammar. This trans- lations of an emergent representation to a set of geometries allows the simulation and calculation of a variety of fitness measures. The best-performing part of the population is set aside for preservation in the next generation by a process called elitism. A sepa- rate selection process defines the parts of the population modified in mutation (random changes in the program tree) and recombination (a crossover of two program trees, which exchanges a part of the tree in corresponding positions in the program tree). Next, the generation cycle is closed by assembly of a new population consisting of the three parts created by elitism, recombination and mutation.
come up with new ideas or design technical artefacts. “Although there have been numerous studies on creativity over the past three decades, they have not addressed the essence of organisational creativity support systems and their design. They have been mainly focused on creative problem solving, creative processes, and individual creativity support.” [Olszak et al. 2018] “[...] It is clear that DSS’s oriented toward individual users are only a special case of the much broader problem of aiding organisational decision processes” [Lee 1982] Therefore, investigating the mechanisms of interactivity that are necessary to build creative DSS contributes to a wider problem space at the same time. Before I describe some of the attributes and properties of DSS, I provide a list of support activities suitable for creative decision support compiled by Niamh O. Riordan and Philip O’Reilly [Riordan and O’Reilly 2012]: (1) identify problem/opportunity; (2) gather information; (3) analyse information; (4) determine measure of performance; (5) identify alternatives; (6) evaluate alternatives (incl. simulation); (7) model or simulate; (8) make selection; (9) implementation; (10) review.
For an overview of DSS, I refer to [Shim et al. 2002]. “A DSS is an interactive system that helps decision-makers utilise data and models to solve unstructured or semi-structured problems.” [Ford 1985] Hence, DSS are capable of supporting architectural design pro- cesses with usually wicked or untamed problem propositions. In contrast to DSS, “an ex- pert system is a problem-solving program that achieves good performance in a specialised problem domain that generally requires specialised knowledge and skill.”[Ford 1985] Ex- pert systems aim to provide domain-specific knowledge to increase the knowledge-base available to the user, while DSS enhance and augment the user’s ability to make deci- sions. The architecture of a DSS consists of a dialog-component, a knowledge-component and a model-component for processing the user input. Figure 4.2 presents the respective DSS Framework adopted from [Ford 1985] in reference to [Sprague Jr 1980].
“Knowledge-based decision support applications differ from those typical of AI ex- pert systems in their open-ended, evolutionary character and need to coordinate with other systems resources, such as organisational databases and quantitative analysis routines.” [Lee 1982] In an adaptive DSS, the knowledge-base for decision support is extracted from observation of the decision-making process. The acquired knowledge-base is then used in an inference engine that intercedes as a support mechanism for the decision-making process. [Fazlollahi et al. 1997] In Typogenetic Design, the computational system ob- serves the decision-making process of the designer and extracts knowledge about aesthetic preferences.
Using “AI type knowledge representations” [Lee 1982] DSS build on two key tech- nologies: databases containing facts and models performing inference. [Sprague Jr 1980] Explicit knowledge is encoded in data bases, often as if-then-rules that capture domain knowledge. Tacit knowledge can be captured using machine learning models that re- move the need for semantic permeation by a purely data-driven modelling process. “A
Figure 4.2: The Dialog-Data-Models DSS Framework adopted from [Ford 1985] knowledge-based DSS must be adaptable and extendable to meet the evolving needs of the user and changing conditions in the environment.” [Lee 1982] Therefore, an Online Clas- sification is the most suitable approach for updating the machine learning model during the decision-making task.
As a result of the interactive nature of the choice-based decision-making process investigated during this PhD research and the chosen decision-tree model, the knowledge base consists of a declarative representation of (mainly) qualitative knowledge [Lee 1982]. Thereupon, declarative knowledge could be extracted based on training of the machine learning model in an intuitive approach that emphasises tacit knowledge.
“The objective of a DSS is to support the user in the decision making process by pro-