Academic
Research Record
|
Halil
Erhan, PhD
www.sfu.ca/~herhan Introduction This document summarizes my academic record and
highlights my academic research contributions throughout my career. I begin
with my experiences relevant to my current academic interests and how I
transformed from an architect to a researcher while keeping my designerly
aspirations alive and active and maintaining my passion for design. I will then
present a portfolio of my sample research output, citing relevant papers and
ordering them chronologically and by area of interest to demonstrate the
evolution of my interests in computational design research over time. Following
a summary of the research funding I received, I include a teaching and student
supervision statement that relates to my research[*]. In framing and structuring my academic work, I will consider SIAT's
core values as cited in the Renewal, Tenure, and Promotion document (2022): 1. SIAT values computational, technological, artistic, design and
cultural artifacts (artifacts include processes) (all hereinafter referred
to as research artifacts) as part of scholarly research, teaching, and
learning, or as a form of service. 2. SIAT values benefits to society, such as improvements in health,
environment, social services, equity, and inclusivity. Background Brief
Biography I
have a multidisciplinary academic background in architecture, design cognition,
and human-centred computing. I earned a Bachelor of Architecture from the
Middle East Technical University in Turkey in 1991. During my early career, my
professional training in architecture sparked my interest in using computing to
design built environments. This led me to explore the
integration of 3D geometric forms with non-spatial design information. For my
master's thesis in Construction Science and Management at Clemson University in
1996, I conducted research on 3D modelling and integration as an alternative
communication medium between the Architecture, Engineering and Construction
(AEC) disciplines. Later, during my Ph.D. studies, my interest shifted from
being a tool adaptor to a tool designer and developer. I investigated
interactive computational support for modelling and generating design
requirements. During this time, I developed RaBBiT, one of the earliest
examples of parametric requirements modelling systems in architecture. I earned
my Ph.D. in Computational Design from the School of Architecture at Carnegie
Mellon University in Fall 2003. Following graduation, I began my first academic
position as an Assistant Professor of Software Engineering at the United Arab
Emirates University, where I worked until Fall 2005. I was then hired as an
Assistant Professor at the School of Interactive Arts and Technology, where I
currently hold the position of Associate Professor after receiving tenure in
2014. I co-led the Computational Design Lab with Prof. Robert Woodbury for 15
years until his retirement. My recent research focuses on building on human
problem-solving and other cognitive theories. It explores how data-informed
creative decision-making can be achieved using Design Analytics interfaces that
combine Visual Analytics and human-centred design computation. Transforming
from an architect to a tool designer and researcher My research and teaching are centred on two primary
domains: architectural design and computing. To reflect on my transformation
from an architect to a (tool) designer and researcher, I would like to share my
past experiences before introducing my research statement. In the early days of
computational systems, I was at the forefront of adapting Computer-Aided Design
(CAD) for built-environment projects of varying complexities. After obtaining
my architecture degree in 1991, I began working as a professional architect at Vakif Insaat in Istanbul, where I
was also tasked with restructuring the project office to integrate CAD-driven
workflows. During my time there, I designed and supervised the construction of
various architectural projects, including the Accounting Ministry's Kadikoy Revenues Offices (50,000 sqm) in Istanbul, Turkey.
This project was selected as one of the 12 finalists of 216 for the Fourth
National Architecture Exhibition Awards in 1994. I also co-designed VakifBank's Adana, Kadiköy, Zeytinburnu Branches and the Accounting Ministry's Umraniye Revenues Complex with Architect B. Ömer Dartan. In addition to these projects, I developed several
architectural projects for the Ministry of Accounting, including a project that
involved restoring and repurposing their Ankara branch using laser scanning to
capture point-cloud data and generate 3D models of historical structures, which
was a pioneering approach at the time. Another restoration project involved
converting over 100-year-old army barracks into Accounting Ministry service
facilities. As a recurring building type, I developed modular
projects for the Foundations Offices Higher-Education Student Dormitories and
modelled them computationally. In 1993, I was hired by Architect Cengiz Gencata as the project coordinator for the Selcuk Culture Center, Ahlat,
Turkey, after he won the national project competition proposing to adapt the
computational design as part of the project management and delivery. Between 1998-2000, while pursuing my doctoral studies,
I worked with Architect George Anderson in Pittsburgh as a designer. During my
time there, I transformed his conventional practice to a CAD-centered design,
increasing the office's efficiency in designing research labs for the
University of Pittsburgh and reducing project completion time from several
weeks to just a few days. My research and teaching are deeply rooted in my
real-world experiences as an architect. I continue collaborating with industry
partners from the AEC community and software companies such as Bentley Systems,
Autodesk, Perkins and Will, Stantec, and Dialog. Research Statement Architecture has interesting features: all problems
are ill-defined. The literature presents a wide range of interpretation of this
concept; among them I prefer to use Simon’s explanation to define what it may
mean in my research practice. “…definiteness of problem structure is largely an
illusion that arises when we systematically confound the idealized problem that
is presented to an idealized (and unlimitedly powerful) problem solver with the
actual problem that is to be attacked by a problem solver(s) with limited (even
if large) computational capacities. If formal completeness and decidability are
rare properties in the world of complex formal systems, effective definability
is equally rare in the real world of large problems. In general, the problems
presented to problem solvers by the world are best regarded as ill-structured
(wicked) problems. They become well-structured problems only in the process of
being prepared for (or by) the problem solvers. It is not exaggerating much to
say that there are no well-structured problems. only ill-structured problems
that have been formalized for (or by) problem solvers.” (Simon, 1973)[†] A defining characteristic of ill-structured problems,
such as those encountered in architecture, is that they cannot be defined
beforehand and are addressed within a problem space where the problem and its
solution develop together. These types of problems require solution-based
thinking and involve multiple stakeholders rather than a single end-user or
problem-solver. Additionally, designing complex systems is constrained by legal
contracts and codes with far-reaching consequences. Another key feature is the
presence of tacit knowledge, where individuals cannot explicitly articulate
requirements or define good design but recognize it when they see it. Tacit
knowledge is typically not accessible to introspection. These features are
evident in the research that informs my work, which I summarize in this
document. A second set of considerations suggests that the design artifact
imposes constraints or offers interesting opportunities in the media and
techniques that can be employed in its creation: a design task and its outcome
change when a new medium is adapted, as it creates new opportunities and
obstacles in searching a design space. Architecture, with its multiple disciplines, complex
geometry, explicit codes and requirements and the composition languages
developed by experts poses issues different from other design disciplines. As
an expert, I aim to make better progress in architectural design than in other
domains by seeking human-centered computational tools to support designers’
creative practices, i.e., help designers to change their task environments with
new tools to explore creative solutions in a problem space. I investigate
(architectural) design as a cognitive and collaborative complex-problem solving
and aim at improving it by augmenting the capabilities of designers with
effective and engaging computational tools for creating built-environments.
I conduct iterative and incremental research at three
overlapping layers (Figure
1):
(a) investigating theories explaining design from cognition, collaboration,
computation perspectives, and developing underlying design computation
frameworks through experimentation; (b) applying the knowledge from the
theories and frameworks relevant to design computation, developing prototype
design tools and testing their ecological validity with real world cases and
scenarios; and (c) learning from and improving design and other creative
practices in the wild. Centered on ‘design and computation’, I use empirical
methods to study and observe designers’ performance as they create solutions to
a range of design problems using various tools and in different settings. I use
the outcome from empirical studies as input to systems research methodology to
design and experiment with new types of tools. I often borrow knowledge and
solutions from other domains such as cognitive psychology, computer science,
and engineering. Design
and Computation Designers today face a changing role as design expands
to encompass a broader range of concerns, including social, environmental, and
technological factors that impact the lifecycle of built environments and
artifacts. Computational design plays a crucial role in this expanding
landscape by enabling the creation and use of tools with decision-support
features that facilitate the exploration, generation, and sharing of design
data. These tools also allow for multidisciplinary collaboration, rapid prototyping,
and direct input to fabrication and simulation. However, the increasing amount
of design data generated requires employing transforming methods and tools that
bridge different knowledge domains, such as data analytics and visualization.
In this context, design computation is also concerned with visualizing and
analyzing design data, similar to making sense of vast
and diverse data in other fields. As design continues to evolve, computational
design researchers must question current practices and explore future
characteristics of design task environments to keep up with the ever-expanding
scope and complexity of design challenges. Research
on Data in Design, and Computation My
current research focuses on developing computational design systems with
enhanced features for designing alternative solutions for built environments by
working with associated data. This program, called Design Analytics, is
essentially an extension and application of Visual Analytics to help designers
effectively deal with the large volumes of data associated with multiple
versions of a design and to collaboratively work with various data types, not
just basic geometric data. The
data created, modified, explored, and shared in design is high-dimensional and
complex. Current computational tools are pressed for features to support
working with this data through multidisciplinary collaboration, agile data
creation and management, rapid validation and verification, vertical and
horizontal technology integration, and experimentation. Therefore, such tools
should provide multiple interaction modalities[‡] to be flexible to
accommodate the designers’ goals. Such modalities can be, e.g., manual,
mixed-initiative, or automated, depending on the task. While
searching for new computational design systems, I explore answers to the
following questions: How can designers effectively generate, analyze, manage,
and share diverse and massive design information in a computer-mediated
context? How can these contexts benefit from novel human-computer interfaces
and interaction techniques? What are the possible opportunities to integrate
multidisciplinary perspectives into this process? Building
on my progress, my objectives are as follows: Objective
1: I aim to
develop functional interfaces and algorithms for creating representations of
design alternatives and evaluate their impact on practice. The bottleneck in
achieving representations lies in two significant folds: the tension between
'design' as a creative act and modelling as a formal way of representing design
alternatives. First, user interfaces should form a transparent layer between
the designer and the design tasks. In addition to the technical challenges in
design data structuring, the goal must be how data is presented and how
designers use it. Second, designing built environments is a complex task that
requires collaboration between various disciplines. Data exchange between
stakeholders will remain an issue. I aim to support multidisciplinary
collaboration and design workflows using task-centric representations as part
of this research objective. I explore and use various human-centred computing
approaches for interactively creating and managing such representations. Objective
2: I aim to
enable designing with alternatives. I develop interactive techniques for
accessing, evaluating, sorting, branching, pruning, and cross-pollinating many
ideas in multiple interaction modalities. With today's computational power, a seed
design model can quickly generate thousands (or more) of design alternatives.
Making sense of them requires various visual, logical, and temporal structuring
in different views. These structures can integrate design criteria such as
cost, performance, material complexity, feasibility. I experiment with system
features such as layout control, filtering, labelling, scoping, and grouping to
reduce solutions to a minimum set of maximally differentiated alternatives. I study
the management of alternatives linked with parallel editing and design
narratives through experimentation with various interaction methods. Objective
3:
Designers often defer their decisions and revisit them to edit, refine, and
explore designs. I aim to develop tools for working with multiple alternatives
in various sequences or branches. I also evaluate interfaces for interacting
with design history to provide reversible activity timelines, design decisions,
actions, and states as interactive narratives. These tools can allow recursion
for discovering new alternatives by reversing the development process to an
earlier, more pliable stage. Design narratives and parallel editing are two
closely related activities. The new tools must enable alternatives created
through direct operations on the model and when designers backtrack and revise
their decisions. Existing logging, versioning, and transaction records are
limited for these purposes. Achieving this objective will reveal whether and
how design narratives can be part of the new design environments and how they
may change design decision-making. Tools for
Design: Selected Research Outcome “And the pain and cost of acquiring the new
tools must be far less than the pain and cost of trying to master difficult
problems with inadequate tools." (Simon and Newell, 1971)[§] My research consists of three interconnected threads.
The first thread focuses on the study of design and the methods employed by
designers. The second thread examines design modelling and representation,
which is relevant to design and other creative domains. The third thread
involves developing techniques to facilitate working with multiple design
alternatives. In this text, I provide an overview of these threads by
highlighting some of the publications and prototypes that were developed in my lab. Design Cognition and Tool Effect
One of the areas that I am particularly interested in
is the impact of notation systems (tools) on the design process. Design tools
often employ notation systems to structure and represent information about
different design aspects. Understanding how these notation systems influence
design is essential. Developing mental models of these systems and
representations created using them requires significant cognitive effort, which
increases when working with multiple models simultaneously. For example, what
cognitive aspects do designers face when using notations for complex
associative models in parametric CAD (pCAD), where parametric changes are not easily traceable?
Furthermore, what are the issues with the cognitive and visual perception
systems in using pCAD,
including several notation systems for representing a single design idea? I
explore these questions through various projects. Change control and perception: We discovered that controlling changes in a parametric design is a significant bottleneck, which includes imperceptible changes, zoom clipping, occlusion, change in multiple locations, a chain of changes, and the effect of change (Figure 2).We proposed the Visual Sensitivity Analysis (ViSA) method to augment visuospatial cognition when working on and controlling changes. ViSA uses controllers and visual cue objects to manage unexpected or hard-to-predict changes without tightly coupling them to the parametric model. In a follow-up study, we investigated change detection in pCAD and visual perception and compared two interface prototypes composed of a symbolic graph and a 3D geometric view. Our results suggest that separating the symbolic graph from 3D geometry helps create a model while overlapping views are more effective for model comprehension. These findings indicate a need for further research on interface design for next-generation CAD systems.
This research mainly contributes to understanding the human cognition system when working with parametric design representations. A practical outcome is the identification of expected characteristics from design interfaces to augment visual cognition. In addition, they have guided us to build a framework for backtracking and deferral moves. In turn, we are using the framework for developing novel interfaces in parametric design environments. · [**]Erhan, H., R. Woodbury and N.H. Salmasi (2009). Visual sensitivity analysis of parametric design models: Improving agility in design. Proceedings of CAAD Futures, T. Tidafi and T. Dorta (eds) Joining Languages, Cultures and Visions: CAAD Futures 2009, PUM, 2009, (pp. 815- 829). · Erhan H., N.H. Salmasi and R.F. Woodbury (2010). ViSA: A Parametric Design Modeling Method to Enhance Visual Sensitivity Control and Analysis. International Journal of Architectural Computing, Special Issue, (pp. 461-483) ·
Kolarić, S., H. Erhan, R. Woodbury, and B. E. Riecke (2010). Comprehending parametric CAD models: an
evaluation of two graphical user interfaces. Proceedings of the 6th Nordic
Conference on Human-Computer Interaction: Extending Boundaries (NordiCHI '10). ACM, New York, NY, (pp. 707-710). · Nasirova, D., H. Erhan, R.F. Woodbury, and B. E. Riecke (2011). Change Detection in 3D Parametric Systems: Human-Centered Interfaces for Change Visualization. Proceedings of 14th International Conference on Computer Aided Architectural Design Futures, CAAD Futures, Belgium, (pp. 751–764). Patterns in Design-moves: Designers keep a record of the design process through their sketches and notes. In pCAD, the record of design moves is limited to command histories and can be implicitly found in the elements upon which the model is built. To understand designers’ work patterns in pCAD, we studied how design task and method specificities affect the design outcome. Through an empirical study, my group showed that the quality of the solutions could be more affected by the goal specificity than by any other design metrics, including quantity and novelty. Following this, my MSc student Rodolfo Sanchez and I identified design moves when dealing with past design decisions (Figure 3). We tested two different types of pCAD systems under similar tasks. The findings helped us to define a framework consisting of backtracking and deferral as the main patterns of design moves. We identified these moves not only for design exploration but also for error correction. We proposed an approach to capture the design narrative as an artefact of design that develops the solutions from a set of discrete action-level data and process-centric interactive visualizations to support better decision-making. The results are favourable, with positive feedback and interesting research questions such as why backtracking patterns differ and what solutions can supersede or change the existing interfaces to support these two patterns.
· Shireen, N., H. Erhan, R. Sanchez, J. Popovich, R.F. Woodbury, and B.E. Riecke (2011). Design Space Exploration in Parametric System: Analyzing Effects of Goal Specificity and Method Specificity on Design Solutions. Proceedings of the 8th ACM Conf. on Creativity and Cognition, Atlanta, Georgia. (pp 249-258). · Erhan, H., R. Sanchez, R.F. Woodbury, V. Muller, M. Smith (2012). Visual Narratives of Parametric Design History: Aha! Now I see how you did it! Proceedings of the 30th International Conf. on Education and Research in Computer Aided Architectural Design in Europe, Prague Czech Republic (pp. 259-268). · Shireen, N., S. Kolaric, H. Erhan, D. Botta, and Robert Woodbury (2013). Exploring representations for parallel development of design solutions using parametric systems. GRAND NCE Annual Conference. ·
Sanchez, R. and H. Erhan (2014). Design ReExplorer:
Interactive Design Narratives for Feedback, Analysis and Exploration,
Proceedings of 32nd Education and Research in Computer Aided Architectural
Design in Europe Conference, (10 pages). Working with a large number of alternatives: In our recent empirical study, we sought answers to how designers work with computer-generated large sets of design alternatives. In investigating the participants' design exploration behaviours with 1000 computer-generated alternatives, our findings revealed a typical cyclic pattern among participants in their design exploration process. It also identified three ways participants represent their design exploration processes. We used visual analytics techniques to study the temporal data mapped to event descriptions and spatial task environment configurations. The initial data shows that designers can only focus on one alternative at a time while comfortably handling 2 to 7 options (or alternative groups) in parallel. They selected to explore around 50 to 100 alternatives in this context through rapid scanning (Figure 4). ·
Shireen, N., H. Erhan, R. Sanchez, J. Popovich, R.F. Woodbury, and B.E. Riecke
(2011). Design Space Exploration in Parametric System: Analyzing Effects of
Goal Specificity and Method Specificity on Design Solutions. Proceedings of the
8th ACM Conference on Creativity and Cognition, Atlanta, Georgia. (pp 249-258) ·
Shireen N., H. Erhan, R. Woodbury, and I. Wang, (2017). Abstract: Making sense
of design space: What designers do with large numbers of alternatives? In Gulen
Cagdas, Mine Ozkar, Leman Figen Gul, and Ethem Gurer
(Eds.), Future Trajectories of Computation in Design: CAADFutures
July 2017, page 414. ·
Shireen, N. and H. Erhan (2019). ParaXplore: Uncovering the Criteria-Building Patterns in
Exploring Large Design Spaces, CHI 2019 Late Breaking Work (5 pages). ·
Shireen, N., H. Erhan, and R. Woodbury (2019).
Encoding design process using interactive data visualization. In J.-H. Lee (Ed.), Computer-Aided
Architectural Design. “Hello, Culture”, Communications in Computer and
Information Science, Springer (pp. 231–246.) ·
Shireen, N., H. Erhan, R. Woodbury, and A. Anttle (2020). Spatial Metaphors for Multi-Dimensional
Design Gallery Interfaces. 25th International Conference of the Association for
Computer-Aided Architectural Design Research, Thailand, (10 pages).
Parametric
Modeling Systems and Interfaces
Parametric
models are dynamic—they change with their inputs, as in spreadsheet
applications. The main benefit of building a parametric model is that design
becomes amenable to rapid changes for experimentation, exploration, and reuse.
Most are familiar with representing the spatial aspects of a design using
parametric models. They can also be used to capture non-spatial design
information, such as design requirements. In this research thread, I look at
how we can adapt and enhance parametric modelling techniques to support a range
of design tasks spanning from early decision-making to exploration of
alternatives and possibly seamlessly linking them. Unfortunately, the most
challenging task is to build the parametric structure of these models for
several reasons. Below I present some system and interface solutions we
developed to enable designers to work with diverse design information using a
parametric modelling approach. Parametric models are dynamic—they change with their
inputs, as in spreadsheet applications. The main benefit of building a
parametric model is that design becomes amenable to rapid changes for
experimentation, exploration, and reuse. Most are familiar with representing
the spatial aspects of a design using parametric models. They can also be used
to capture non-spatial design information, such as design requirements. In this
research thread, I look at how we can adapt and enhance parametric modelling
techniques to support a range of design tasks spanning from early
decision-making to exploration of alternatives and possibly seamlessly linking
them. Unfortunately, the most challenging task is to build the parametric
structure of these models for several reasons. Below I present some system and
interface solutions we developed to enable designers to work with diverse
design information using a parametric modelling approach. Parametric modelling of non-spatial design information: Building on my Ph.D. thesis, I developed a framework for design requirements modelling as a Generalized Means and Ends Analysis, and I built a prototype called RaBBiT, adapting the framework (Figure 5). RaBBiT
provides a graph-based interactive interface to capture design requirements through parametric and associative dependencies. The requirements are encapsulated in nodes and conditional links that can dynamically switch on or off a relationship between two nodes. They result in different graph composition under a different set of conditional changes on the nodes. RaBBiT aims to model rules and domain concepts in users’ terms and generate decision compositions using these models. I received a Boeing Canada AeroInfo fund to explore RaBBiT’s parametric capability for testing decision configurations for possible applications in the Visual Analytics domain. I am working on a conceptual task model for RaBBiT that is operational enough to guide support for reusing ‘decisions’ for ‘what-if’ scenarios, impact analysis, and traceability. We also showed how non-spatial information could be linked with spatial information on different projects. RaBBiT has the potential to be linked with other parametric modelling tools to better support the design seamless process. · Erhan, H.I. (2003). Interactive Computational Support for Modeling and Generating Design Requirements. PhD Thesis; Computational Design, School of Architecture and Institute of Complex Engineered Systems, Carnegie Mellon University, Pittsburgh, PA. · Erhan, H. and U. Flemming (2005). User-System Interaction Design for Requirements Modeling, CAADRIA 2005. Proceedings of the 10th International Conference on Computer Aided Architectural Design Research in Asia New Delhi (India) 28-30 April 2005, vol. 2, (pp. 160-170) (Received Best Paper Award among 96 papers). Erhan, H. and F. Djebbar (2007). Pair-Collaborated Usability Study
of RaBBiT in Requirements Modeling and Generation. Proceedings of the 12th
International Conference on Computer Aided Architectural Design Research in
Asia, Nanjing (China) 19-21 April 2007, (pp. 399-409) (Received Best
Presentation Award among 87 papers).
Interfaces for graph-based modelling: Graph-based modelling, as a form of a visual programming technique, is not exclusive to pCAD. The computational systems with graph-based modelling enable representing design intent through visual programming. We identified more than 30 software tools using graph-based modelling for different purposes, e.g., animation, generative art, and testing electrical circuits. However, they are criticized for needing to be more scalable for complex models, and they can quickly grow to have hundreds (or more) of nodes and links. To alleviate this bottleneck in the data-flow graph interface in GenerativeComponents, we collaborated with the developers in Bentley Systems (Figure 6-Top). We created rich-node design and interaction techniques to increase their functionality and reduce the cognitive challenges posed by graph complexity, such as semantic zoom, composition, and scoping. We created a prototype that adopts the rich node design and introduces ‘debugging-like’ interactions to understand and control the parametric dependencies in the graph (Figure 6-Bottom). The outcomes of this research thread are predicted to apply in other domains, such as games, engineering, and product design. · Nasirova, D., H. Erhan, R.F. Woodbury, and B. E. Riecke (2011). Change Detection in 3D Parametric Systems: Human-Centered Interfaces for Change Visualization. Proceedings of 14th International Conference on Computer Aided Architectural Design Futures, CAAD Futures, Belgium, (pp. 751–764).
Visual
Sensitivity Control and Analysis in Design: Parametric
computer-aided design systems allow designers to rapidly generate models and explore
the downstream impacts of changes to key design parameters. However, these
systems can become insufficient when exploring increasingly complex parametric
design models. The main challenges for exploration are controlling and
analyzing changes in the design model, mainly when changes are introduced
continuously. The system interfaces, and the human visual perception system can
alleviate these challenges. To address these challenges, we developed ViSA, a
visual sensitivity analysis method for parametric design modelling that aims to
make the effects of change within a parametric model controllable, measurable,
and apparent to designers. This approach differs from conventional mathematical
or statistical sensitivity analysis methods and falls under the category of
visual (or graphical) sensitivity methods. ViSA controls and visually displays the model's
sensitivity to the designer through interactive representations that can be of
the same or different types as the model itself. The method proposes
customizable control and visualization features in the model that are decoupled
from each other at the design level while providing interfaces between them
through parametric associations (Figure
7-Left
Top). We adopted the Model-View-Controller paradigm from software design to
decouple these features while providing interfaces between them through
parametric associations (Figure
7-Left
Bottom). Figure 7 (Right) shows the effect of
changing parameters t-value and delta of perturb points TP03,
TP04, and TP06, creating a mesh structure.
The colour-coded visualization helps recognize which parts of the meshed model
experience more decrease or increase instantly and continuously. ViSA can
improve the exploration of complex parametric design models and support designers
in making informed decisions. ·
Erhan, H., R. Woodbury and N.H. Salmasi (2009). Visual sensitivity
analysis of parametric design models: Improving agility in design. Proceedings
of CAAD Futures, T. Tidafi and T. Dorta
(eds) Joining Languages, Cultures and Visions: CAAD Futures 2009, PUM, 2009,
(pp. 815- 829) ·
Erhan H., N.H. Salmasi and R.F. Woodbury (2010). ViSA: A Parametric
Design Modeling Method to Enhance Visual Sensitivity Control and Analysis. Int.
Journal of Architectural Computing, Special Issue, (pp. 461-483) Alternatives in Design Space
My group developed an active research program on design
space exploration by working with alternatives using parametric design systems.
Designers explore alternative solutions (or simply alternatives) that are said
to exist in a design space. They display distinct patterns of exploration, the
chief of which are parallel development, history revision, and solution fusion.
Current CAD systems work on largely single-state design models with little
support for alternatives. This briefly means that at any given time, only one
alternative is accessible, unlike using more conventional methods such as
sketching several alternatives on the same workspace. The “subjunctive
interfaces” introduce interaction techniques for exploratory analysis of
scenarios side-by-side and editing these scenarios in parallel. By learning
from the literature and through experimentation, we are working on developing
new interaction and interface solutions on realistic prototypes for design
exploration. Exploring in graph-based modelling: pCAD works on one parametric state at a time; alternatives are lost as the design progresses unless explicitly saved. To enable working with multiple design alternatives in parallel, we have developed a parametric modelling prototype using subjunctive graphs. A subjunctive graph is derived from another prototype dependency graph to execute "what-if" scenarios (Figure 8). It inherits all properties from a prototype graph. It can override these properties, add new properties, define parametrically or structurally different solutions, and compose properties from several prototypes. The prototype introduces features that merge variations, reuse parts from other graphs, and simultaneous manipulation. As part of the ENCAD project, this method is adopted by Prof. Wolfgang's group to develop new graphical design applications. RaBBiT also implements a prototype-based modelling technique of subjunctive graphs to support the exploration of alternative design briefs.
·
Kolarić, S., H. Erhan, R. Woodbury,
and B. E. Riecke (2010). Comprehending parametric CAD
models: an evaluation of two graphical user
interfaces. Proceedings of the 6th Nordic Conference on Human-Computer Int.:
Extending Boundaries (NordiCHI '10). ACM, New York,
NY, (pp. 707-710). · Shireen, N., H. Erhan, D. Botta, and R.F. Woodbury (2012). Parallel Development of Parametric Design Models Using Subjunctive Dependency Graphs, Proceedings of the 30th Annual Conference of the Association for Computer Aided Design in Architecture, San Francisco, California (14 pages). Parallel Editing: Designers work by exploring alternatives. They mostly edit what they have already done. They scan, remember, compare, and combine. The result is a branching structure of alternatives. Designers cover desk and wall space with alternative sketches to see many at once. Only a few tools can support this strongly preferred style of work. We need new ways to view and manipulate alternatives. Configuration Management (CM) provides tools to manage variations of well-defined designs, resulting in a product family. Design Exploration needs such structure but also visibility and direct interaction. Dr. Woodbury, Dr. Koloric, and I presented a method that builds on CM and shows a path to directly working with alternatives. Our three key principles are: (a) View many alternatives at once; (b) Edit alternatives in ad-hoc groups; and (c) Maintain the identity of parts across alternatives. We built a prototype system called CAMBRIA, adopting these principles (Figure 9). Our method can be adapted to any system and likely works best with component-based parametric models. GRAND NCE, NSERC, Bentley Systems, and MITACS have funded CAMBRIA and research on alternatives. Based on our
experimental study, we developed a user interface that aims to enable designers
to work with a large number of computationally
generated alternatives (Figure 11). It provides functions for design criteria building and
applying in a continuous search process. My PhD student Naghmi
Shireen developed prototype-based parallel editing techniques by using multiple
models in (Figure 10). · Woodbury, R., S. Kolaric, H. Erhan, J. Guenther (2013). Exploring for Designs: Five basic elements. In Armstrong R. and Ferracina S. (eds.) Unconventional Computing: Design methods for adaptive architecture, Riverside Architectural Press. · Kolaric, S., R.F. Woodbury, H. Erhan (2014). CAMBRIA: A Tool for Managing Multiple Design Alternatives, WIP, Proceedings of the Designing Interactive Systems 2014, Vancouver Canada. · Erhan H. and N. Shireen, (2017). Juxtaposed design models: A method for parallel exploration in parametric CAD. In Proceedings of CAADFutures 2017, Istanbul Turkey (12 pages). · Kolaric S., H. Erhan, and R. Woodbury (2017). Abstract: CAMBRIA: Interacting with multiple CAD alternatives. In Gulen Cagdas, Mine Ozkar, Leman Figen Gul, and Ethem Gurer (Eds.), Future Trajectories of Computation in Design: CAADFutures July 2017, page 509. · Kolaric, S. and H. Erhan, (In Review). The Case for Design Alternatives: In Support of Multi-State Interaction, ACM Transactions on Computer-Human Interaction. · Kolaric S., H. Erhan, and R.F. Woodbury (2011). Complex Floor Plans: How to Represent Them, and Interact With Them?, Poster, GRAND conference. (Poster). · Kolaric, S., R. Woodbury, H. Erhan (2012). CAMBRIA: Set-based Interaction in Design Space Exploration, GRAND Conference, Montreal, QC (Poster). · R. Woodbury, S. Kolaric, H. Erhan, J. Guenther (2013). Design Exploration and Configuration Management: Two Sides of the Same Coin?, COFES conference, Scottsdale, AZ, US. (Poster).
Figure 11. (Top) A designer (participant) studying a large number of design solutions parametrically generated. (Bottom) The interface prototypes demonstrating working with alternatives using clustering, tagging, and using SOM for similarity-based search. Alternative designs in the past: I received Boeing AeroInfo Systems and MITACS matching fund to further explore supporting design exploration using actions in CAD recorded as data. The project aimed to (a) provide feedback through design narratives; a model of the design process built through the actions and alternatives designers make and develop in terms of what, how and when; (b) facilitate a local and global analysis of the design actions and alternatives across the narrative, and (c) enable the execution of design decisions within the interface based on the preceding analysis. Design ReExplorer (Figure 12) captures design moves in a time-action-directed graph, and make the graph used to present and edit the record of design histories connected with a parametric design modeling tool. Following this project, we look at how design solutions can be developed using computational tools from a set of discrete action-level data; and how process centric interactive visualizations can support better decision making and design exploration capabilities. This work set the framework for Design Analytics (Figure 13). · Erhan, H., R. Sanchez, R.F. Woodbury, V. Muller, M. Smith (2012). Visual Narratives of Parametric Design History: Aha! Now I see how you did it!, Proceedings of the 30th International Conference on Education and Research in Computer Aided Architectural Design in Europe, Prague Czech Republic (pp. 259-268). · Sanchez, R. and H. Erhan (2014). Design ReExplorer: Interactive Design Narratives for Feedback, Analysis and Exploration, Proceedings of 32nd Education and Research in Computer Aided Architectural Design in Europe Conference, (10 pages).
Figure 12. While the designers can explore design alternatives in the past actions, it enhances the exploratory capabilities of pCAD. The editable visual history of design alternatives provides interactive controls of the parametric design by directly connecting with the pCAD systems' working models.
Design Analytics
Based on my previous research, I develop solutions for data-informed decision-making in computational design. This approach builds on research on creativity support tools and data analysis research, called "Design Analytics," which combines visual analytics with computing to explore design alternatives with form and performance data. The premise for this proposal is that we can assume descriptions or artefacts representing a design as data. These data collections contain geometric data and increasingly large amounts of numeric and textual form in structured and unstructured formats representing processes, contracts, and design performance. Traditionally such data have been limited to, for example, visual representations, specifications, and other design documents. Today, design firms generate vastly more diverse data but
need more access to tools to gain insight from such data. Yet the "visual
analytics" field provides concepts and systems for the much-needed
analysis of such data. Design Analytics research aims to visualize and analyze
design data, that is, collections of designs, their alternatives, project
documentation, and other data collected from buildings and their settings. Use of Similarity-Data for Search: We developed a design-space reduction method using Self-Organizing Maps (SOM) to enable interaction with a large number of (1000) design alternatives through a similarity-based exploration. We present a scenario on developing conceptual designs for a residential apartment to illustrate the limitation of current tools and explore potential interface solutions for parametric design (Figure 14). We formalized system features adopting existing data analysis and visualization techniques for systematic filtering, clustering, and choosing alternatives. In a follow-up project, we conducted a case study to experiment with how the reliability and validity of parametric models can be tested in a realistic collaborative design setting and how such design models can be improved using the insight gained from the interactive data visualizations. We received over 250 alternative designs and their performance data generated for a mixed-use high-rise building by eight architects from different institutions. We investigated the alternatives using Tableau, a general-purpose Visual Analytics tool (Figure 15). We observed a set of recurring issues and proposed solutions or approaches that can be used to address the concerns in this case study. For example, data from the alternatives can cover a wide range of test cases demonstrating a model’s behaviour under a diverse sample of generated input. Designers must be aware of the limitations of the performance modules suitable for the task. Using multiple models stresses the advantage of challenging the module’s generality by subjecting them to reliability tests. While it is possible for a correlation pattern or an outlier to raise questions when observed in data visualization, expert knowledge is necessary for proper interpretation.
Figure 14. A case of a massing study for a low-rise apartment building in downtown Vancouver use to generate a large number of alternatives (>5,000). We used SOM and cluster designs with various parametric combinations. The goal is to narrow the selection set to manageable clusters.
·
Erhan
H., I. Wang, N. Shireen (2014).
Interacting with Thousands: A Parametric-Space Exploration Method in Generative
Design, Proceedings of the 34th Annual Conference of the Association for
Computer Aided Design in Architecture, United States, California, Los Angeles, (11 pages). · Erhan, H., I. Wang, and N. Shireen (2015). Harnessing Design Space: A Similarity-Based Exploration Method for Generative Design, Design Agency Special Issue, International Journal of Architectural Computing Spring 2015 (19 pages). · Erhan, H., Chan, J., Fung, G., Shireen, N., and Wang I. (2017). An Epistemic Action Analysis: Understanding cognitive overload in generative design. In the Proceedings of the 22nd Int. Conf. of the Assoc. for Computer-Aided Architectural Design Research in Asia, Suzhou, China. (CAADRIA) (10 pages). · Erhan, H., A. Abuzuraiq, M. Zarei, O. Alsalman, R. Woodbury, and J. Dill (2020). What Can Data Reveal About Your Design Model? A Case Study on Reliability and Validity. 25th International Conference of the Association for Computer-Aided Architectural Design Research, Thailand, (10 pages). Combining form and performance data: Geometric forms and performance data of built environments are interrelated and equally important in design. Each presents information on different aspects of the design alternatives. Therefore, the interactive visualization for design analysis should enable the exploration of form and performance data both independently from each other and in unity. The similarity between design alternatives can be computed depending on the aspects under comparison; if the similarity between alternatives correlates with their geometric or visual similarity, then the interactions that use it, whether implicit or explicit, can be treated as form-first interactions. This applies to interactions with the results of both clustering and dimensionality reduction techniques. This form-based similarity can be computed by comparing input parameters if similar input parameters generate similar geometries. The applicability of this assumption is highly limited for design spaces with complexity where the relationship between the input parameters and outputs could be less predictable. In a design process that involves multiple generative design models, whether as iterations on a single model or in a collaborative setting, we would like to be able to explore them jointly. Relying on the shared input parameters between them might not be possible or applicable. Figure 16 illustrates similarity-based coupling with a type of form-first interaction since it is initiated from a representation derived from form-based similarity. We accomplished this through shape similarity approaches using distance functions like the Euclidean distance or calculating the distances between their corresponding mesh vertices. Not all aspects of a design can be quantified. It is then crucial for designers to have visual access to the design forms to identify these qualitative features and compare the designs' forms with each other. To support this, a grid of 3D thumbnails is filled with the selected alternatives in the Scatterplot and Dendrogram tree (Figure 16).
· Abuzuraiq, A. and H. Erhan (2020). The Many Faces of Similarity: A Visual Analytics Approach for Design Space Simplification. 25th International Conference of the Association for Computer-Aided Architectural Design Research, Thailand, (10 pages). D-Sense: Navigating design spaces: Generative design promises novel and performant solutions to architectural design problems, but it produces large design spaces that are challenging to navigate and select from due to the choice overload they create. Several design space navigation (DSN) interfaces have been proposed to support this task. A distinctive feature of DSN interfaces is the inclusion of abstract data visualizations (data views) and gallery-like views of design geometries (form views). We conducted a design study with domain experts, resulting in a domain characterization of DSN. In addition, we identified several areas for improvement in current DSN interfaces, notably the lack of tight coupling between data and form views. Furthermore, we developed DesignSense (or D-Sense), a DSN interface of coordinated and linked form and data views, as aiding mechanisms for expressive selection from, seeking through, clustering, and manually grouping design alternatives (Figure 17). We evaluate the current iteration of DesignSense through a formative focus group and present realistic case studies of its utility. Domain tasks we identified include (a) finding designs by narrowing a design space and navigating collections of designs to identify a set of satisficing designs considering changing design criteria due to the ill-defined nature of design problems; (b) relating inputs to performance and forms (sensitivity analysis) to identify dependencies and correlations; (c) assessing performance trade-offs to identify better designs or trade-offs; (d) logically couple form and data to arbitrate between two designs with similar performance or be- tween designs with trading-off performance portfolios; and (e) externalize insight to formulate hypotheses concerning the alternatives and their relation to the design problem (sensemaking loop). ·
Abukhodair, F., B.E. Riecke, H.
Erhan, C.D. Shaw (2013). Does interactivity improve exploratory data
analysis of animated trend visualizations? Proceedings of SPIE-IS and T.
Electronic Imaging, Visualization and Data Analysis 2013, (11 pages). · Erhan, H., A. Abuzuraiq, M. Zarei, O. Alsalman, R. Woodbury, and J. Dill (2020). What Can Data Reveal About Your Design Model? A Case Study on Reliability and Validity. 25th International Conference of the Association for Computer-Aided Architectural Design Research, Thailand, (10 pages). · Abuzuraiq, A., O. Alsalman, H. Erhan (2020). Game Level Design by Shopping: A Visual Analytics Approach to Exploring Generated Content, The International Conference on the Foundations of Digital Games. Bagibba, Malta 2020, (10 pages). ·
AbuZuraiq, A. and H. Erhan (2023). DesignSense: A Visual Analytics Interface for
Navigating Generative Design Datasets, <Journal TBD, Summer 2023).
D-ANZ: Design Analyzer: D-ANZ is a design analytics project demonstrating and testing how visual analytics of parametric design data can become integral to design exploration in a parametric and generative design workflow. We developed design analytics techniques first on low-fidelity prototypes (Figure 17). It categorizes the arbitrary data retrieved from design repositories into relevant datasets based on their parametric criteria. Results obtained from user testing encourage further research and exploration of the basic D-ANZ features. Timeline and Hierarchical Cluster views providing data visualization based on the time of activity, the temporal order of origin, and the evolution of design alternatives helped the user derive a holistic overview of the project. Analytical features enabling custom visualization of input and output data derived from parametric models support design cognition and rapid exploration. ·
Garg, A. and H. Erhan
(2019). Use of Data in Design Exploration: Design Analyzer, CAAD Futures 2019,
South Korea (12 pages). Combining
Directly Interactive Design Modeling with Design Analytics: We explored how directly manipulating non-parametric geometries can be
used with real-time parametric performance analytics for design-decision
making. This combination gives rise to a design process where considerations
that traditionally take place at late design phases can become part of the
early phases without the expense of developing highly labour-intensive
and limiting parametric models for form-finding. We developed D-FlowUI and
D-CAT as prototype tools for performance-based design in collaboration with our
industry partner (Figure
19).
The tools work with and respond to changes in the design modelling environment,
process the design data, present the results, and support comparative analytics
(Figure 20). Directly
interactive CAD provides agility for sculpting design geometries, making a
divergent, sketch-like exploration possible. The design criteria for D-FlowUI,
a design tool for the AEC industry, were developed through an iterative process
with professional designers, resulting in a high-level set of criteria that
include allowing form-first exploration, incorporating design program input,
enabling context definitions, managing multiple design ideas, and accommodating
diverse design-analysis scenarios. The prototype was tested with a project from
industry partner Boeing Canada, Stantec, and MITACS provided funding for the
project.
Figure 19. Directly sculpted massing models created in Rhino are analyzed, and
corresponding data is selectively presented in D-FlowUI’s dashboard in real-time or
on-demand. An arbitrary number of alternatives can be managed in the task
environment, including their CAD models, parametric definitions for performance
assessment, and data visualizations. D-FlowUI takes
advantage of the benefits of both approaches augmented with interactive
visualization of design alternatives with their form and performance data.
·
Erhan,
H., A. Abuzuraiq, M. Zarei,
A. Hass, O. Alsalman,
R. Woodbury (2020). D-FlowUI: Combining Directly Interactive Design Modeling
with Design Analytics. 25th International Conference of the Association
for Computer-Aided Architectural Design Research, Thailand, (10 pages). · M. Zarei, H Erhan, A.M. Abuzuraiq, O. Alsalman, A. Haas (2021). Design and Development of Interactive Systems for Integration of Comparative Visual Analytics in Design Workflow. 26th International Conference of the Association for Computer-Aided Architectural Design Research, Hong Kong, (10 pages). D-Predict(v1-v3): In another Design Analytics
project, we integrated generative design and surrogate modelling for data-informed
exploration in the early phases of building design. The increasing importance
of sustainable architecture has led designers to incorporate algorithmic design
generation, design analytics interfaces and machine learning-based performance
analysis into the design process. The challenge lies in integrating these
highly related tasks into a cohesive and engaging design environment that
maintains designers' creative flow while reducing system complexity. The
central question is how to support designers in making data-driven decisions by
offering a design analytics interface that combines generative design and
machine learning-based surrogate modelling for performance analysis. We propose
a workflow and a functional design analytics interface that supports the
generation of designs and their performance prediction using machine
learning-based surrogate models and interactive design-data visualizations in
the early phases of building design.
Figure 21. (Top) A workflow for iterative design analytics integrating data from generative design and performance analysis. The workflow ensures the generation of design alternatives, prediction of performance metrics using machine learning, and assessment of results against predefined criteria. (Bottom) D-Predict.v2 main interface with model-agnostic parametric controllers and data analysis panels; Parametric controllers (A) define zones and contexts; (B) select walls to apply material or openings; (C) set up Horizontal and vertical shading devices; and (D) assign material to boundary elements, such as ceilings, floors, walls, or windows. Peripheral daylight metrics (UDI, ASE, and MI) visualizations with benchmark references. Energy load visualization by predicting potential energy sources and consumption. The next version of D-Predict (v3) focuses on algorithmic
or GenAI methods to explore design space. Architectural design work is
inherently multidisciplinary, with architects constantly under time pressure.
In the face of pressing demands for creating sustainable and performant built
environments, designers increasingly rely on computer simulations to assess
their designs, e.g., how much energy they use and how well they utilize natural
light. However, such simulations are time-consuming and often require expertise
in building science, which is not typically part of an architect’s training.
Surrogate models—machine learning models trained to predict the outcomes of
computationally expensive simulations—offer a solution to these challenges. We
developed an interactive and user-friendly design tool tailored for the early
design phase, which integrates surrogate models of energy and daylighting to
provide architects with rapid and actionable insights into the design space, guiding
more informed and efficient design decisions.
Figure 22. D-Predict.v3 enables spatial zones, and parameter ranges to be selected using various sampling methods, such as the Latin square or incremental selection, to predict building alternatives and generate data for further exploration. Like D-Predict.v2, this version relies on the same daylight surrogate model and energy load prediction modules. ·
Abuzuraiq, A.M., H. Erhan, E. Muttaqhi, et al. (2024). An Integrated and Designer-Friendly Approach for
Performance-Based Design: A User Study and System Proposal. ACADIA 2024,
Calgary (10 pages, Accepted) ·
Erhan, H., E. Mottaghi,
A.M. Abuzuraiq, V. Okhoya, S. Ampanavos,
M. Bernal, C. Chen, Y. Madkour (2024). Integrating
Surrogate Modeling and Visual Analytics for Data-informed Design Exploration in
the Early Phases of Built Environment Design, eCAADe
2024, Cyprus (10 pages) ·
Mottaghi, E., H. Erhan, and A.M. Abuzuraiq (2024). D-Predict: Integrating
Generative Design and Surrogate Modelling with Design Analytics, CAADRIA 2024
(10 pages) ·
Sadeghi, H., H. Erhan, A.M. Abuzuraiq (2024).
D-GreenPlans: A prototype Design Analytics System to
Support Building Sustainability of Alternatives, eCAADe
2024, Cyprus (10 pages) Reporting
explored solutions with data: Evaluating design ideas is integral
to designing built environments. It involves multiple stakeholders with diverse
backgrounds reviewing design solutions by studying their form and performance
data. Although there are computational systems for supporting evaluation tasks,
they are either highly specialized for designers or configured for a particular
workflow with limited functions. We developed a Design Analytics method aiming
at a collaborative and data-driven evaluation of alternatives in the
design-evaluate-feedback cycle. Adopting this approach, we introduce D-ART as a
prototype system composed of customizable Web interfaces for presenting design
alternatives, enabling stakeholders to participate in data-informed discourse
on alternatives and providing feedback to the design team (Figure
22).
Its system design considers requirements gathered through literature review,
critical analysis of the existing systems and collaboration with our industry
partners. Finally, we assessed D-ART’s design through an expert review
evaluation, which generally reported positive results on the system’s goals. An
important role of D-ART is to help designers reflect on and share their search
process. D-ART consists of four main visualizations (Figure
22).
They display alternatives in a juxtaposed layout for overall evaluation, enable
stakeholders to show alternative details, together or individually, provides a
commenting features and selection of relevant criteria to their interests.
·
Alsalman, O., H. Erhan, A. Haas, A.M. Abuzuraiq, and M. Zarei (2021). Design Analytics and Data-Driven
Collaboration in Evaluating Alternatives. 26th International Conference of the
Association for Computer-Aided Architectural Design Research, Hong Kong, (10
pages). ·
Alsalman O. and H. Erhan (2022). D-ART for Collaboration in
Evaluating Design Alternatives. International Journal of Architectural
Computing. Special Issue 2021. (17 pages). ·
Piray, P. H. Erhan and A. AbuZuraiq
(2023). Annotation on Interactive Design Data in Design Analytics Tools, 28th
International Conference of the Association for Computer-Aided Architectural
Design Research, India (10 pages) Data-informed Design Democratization: As a human-centred approach, the research aims to integrate data in evaluating
design alternatives to enhance stakeholder involvement in built environment design
(Figure 23). The
projects centred on investigating the potential of design data enhanced by
visual analytics and social media-like interactions to improve public
engagement in built environment design. This study explores how diverse users, specialists
and non-specialists can utilize an interactive prototype I am developing to
make data-informed decisions. Given the varying skill levels in data handling and
comprehension among users, this tool aspires to bridge the gap between
specialists and non-specialist users, empowering them to participate in
decisions regarding the design of their built environments, significantly impacting
their social fabric. ·
Khan, Z.M., H. Erhan, and A.M. Abuzuraiq (2024).
Bridging the Gap: Design Data Democratization for Enhanced Public Participation
in Built Environment Design, CAADRIA 2024 (10 pages) ·
Khan, Z.M. and H. Erhan (2024). Data-Informed Design
Democratization: Engaging Design Stakeholders for Creating Liveable
Built Environments, eCAADe 2024, Cyprus (10 pages) ·
Behrouz, T.E, H.
Erhan, and A.M. Abuzuraiq (2024). A Location-Based Augmented Reality
Tool for Empowering Public Engagement and Design Assessment of the Built
Environments. CAADRIA 2024 (10 pages)
The adoption of democratic and participatory design
processes involving community members can lead to the creation of better and
more livable environments. In another project, we examine the potential of
location-based augmented reality (L-AR) as a system solution for engaging the
public in evaluating design proposals, intending to contribute to democratizing
built environment design. We introduced a mobile L-AR prototype, D-ARE, which
leverages mobile devices' capabilities to allow interactive and in-situ
visualization of design proposals, along with features like interactive AR form
views, performance data displays, and interfaces for facilitating discussion
threads (Figure 24). We
focus on the challenges of transforming complex design data into understandable
formats for non-specialist users. We developed insights gathered from D-ARE's
user evaluation with 20 participants, highlighting promising engagement
possibilities and identified challenges. The findings emphasize the
transformative potential of in-situ AR applications and the importance of
fostering informed dialogue between designers and community members to ensure
that built environments reflect their needs and perspectives.
Collaboration and Tools: Effective
collaboration requires aids for generating, sharing, and managing artifacts
representing products and decisions and tracing them as people work on
different parts of a task. However, not all collaboration artefacts are
recorded or can be found among others, resulting in some getting lost or
forgotten. Collaboration Support Tools (CST) are available, but why can't they
address this issue? While the availability of CST addresses one part of the
problem, the "principle of least effort" addresses the other. Agents
in collaboration minimize their effort in presenting and accepting arguments. In general, our individual and collective behaviour
form an ecological system in which we try to minimize our efforts to achieve a
particular goal. If the immediate benefits of using a tool do not outweigh the
perceived effort, the tool will not be used. Even small obstacles on the path
cause us to search for other paths with lesser effort. Therefore, a
collaboration support tool's success should be measured by how well it can
address the "principle of least effort." Collaboration is not only based
on artefacts but also on cues such as body language. Thus, it is essential to
reduce the perceived distance between collaborators. How can readily available
devices reduce this distance? We use them every day to organize our activities
and communicate with others. Can these devices better support how we work
without overloading the workflow? In this research, we seek answers to these
questions by prototyping a system called DiNa (Figure
25).
Following the DiNa framework, we develop a more generalizable
topic-centric collaboration tool which demonstrated shifting from
document-centric to topic-centric discourse between the collaborators (Figure 26). The topic-centric asynchronous
collaboration system developed uses available technologies as modular system
components with high compatibility and interoperability. These components
include peripheral devices like mobile phones, smart pens, and cameras. ·
Rajus,
V.S., R. Woodbury, H. Erhan, R. E. Riecke,
and V. Mueller (2010). Collaboration in Parametric Design: Analyzing User
Interaction during Information Sharing. Proceedings of the 30th Annual Conference of the Association for Computer Aided Design in Architecture, ACADIA 2010: LIFE in:formation, On Responsive
Information and Variations in Architecture, New York, NY, (pp. 320-330). ·
Huang, A., H. Erhan, R.F. Woodbury, K. Kazlova, and D. Botta (2011).
Collaboration Workflow Simplified: Reduction of Device Overhead for Integrated
Design Collaboration. Proceedings of 14th International Conference on Computer
Aided Architectural Design Futures CAAD Futures, Belgium, (pp. 591–601). ·
Erhan H., D. Botta, A. Huang,
R.F. Woodbury (2013). Peripheral Tools to Support Collaboration: Probing to
Design Collaboration through Role Playing, in R. Stouffs
et al. (eds.), Proceedings of the 18th International Conference of the
Association of Computer-Aided Architectural Design Research in Asia, Singapore
(pp. 241-250). ·
Erhan, H., A.
Huang, R.F. Woodbury (2014). DiNa Framework and
Prototype to Support Collaboration in the Wild, Proceedings of
the International Conference on Computer Aided Architectural Design Research in
Asia, Kyoto Japan, (10 pages). ·
Oppenlaender, J., N. Shireen, M. Mackeprang, H. Erhan, J.
Goncalves, and
S. Hosio (2019).
Crowd-powered interfaces for creative design thinking. In Proceedings of the 2019 on Creativity and
Cognition ACM, New York, NY, (pp. 722–729).
Other
research collaboration: Certainly, building a solid
research program hinges on collaboration. I feel lucky to work closely with
world-class researchers not only in my domain but also in others. In parallel
to my main research, I collaborate with researchers where I find opportunities
to learn from and contribute to solving problems in other areas. Some of this
research aims to help visually impaired people to navigate the Web, with Prof.
Kamal to develop Visual Analytics interfaces, with Prof. Shaw and Prof. Fisher;
to increase the effectiveness of social collaboration, with Prof. Woodbury; and
to learn and teach about spatial cognition, with Prof. Dill and Ms. Berry. All
these areas intersect with my background and interest in design, cognition, and
human-centred system design. My teaching practice is also informed by my and
others’ research, particularly for developing and applying interactive
computational tools in solving real-world problems in the creative domains. In
addition to the above research, I studied pedagogies for teaching spatial
thinking that I published as book chapters and peer-reviewed conference papers.
I am particularly interested in spatial thinking and cognition as it presents
fascinating insights into how people perceive, represent, and interact with
space. The further implications of spatial thinking and design go beyond
architectural design and involve VR and interactive game development. I see
strong parallels between these fields. An exciting proposal I was involved in as a project
leader was titled “MOSIM: End-to-end Digital Integration based on Modular
Simulation of Natural Human Motions.” I led the SFU’s participation in this
project. As an international effort initiated by Daimler AG of Germany, the
project is part of the ITEA 3 Eureka Program. The project’s international
partners were automotive, game, software, healthcare, and construction
companies and several universities from Germany, Canada, Finland, France, and
Sweden. Qube Building Systems Inc. and SFU SIAT headed
the Canadian Consortium. I worked with Qube Building
Systems closely on this project and invited Prof. Wolfgang and Prof. Fisher to
participate as their expertise aligned with the project goals. In addition,
Prof. Fiume (from FAS) was invited to be part of the SFU’s front in MOSIM. The
other partners in Canada were Archiact Interactive,
Finger Food Studios, National Research Council Canada, Unity Technologies, and
the University of Calgary. The project aimed to automatically simulate a rich
repertoire of realistic human motions in various task environments (e.g.,
construction, healthcare, and automotive manufacturing) by adapting a generic
framework comprising a comprehensive database of modular human motion units. In
December 2017, ITEA 3 evaluated our proposal outline, and we were invited to
submit a full proposal by 15 February 2018. SFU and our industry partners sought
funding through Canadian organizations, such as NCR and NSERC Strategic Grant
upon final approval. The Canadian consortium was aimed at an equivalent of €3
million in funding, of which SFU would be received 18%. We proposed to
investigate data-driven reasoning methods to computationally generate,
simulate, and validate sequence configurations of discrete human motion model
units (MMU). The ITEA’s assessment for our proposal was ranked highest among
the others. Unfortunately, our European partners had to remove our part from
the proposal after the industry leader, Qube Building
Systems, unexpectedly withdrew from the program. This was a missed opportunity
for Canada and SFU. However, I continue to search for international
collaboration opportunities. Recently, I submitted a support letter for another
European project to participate as an associated partner in Computational
Design for Performance Based Architecture (CODE4ARCH) submitted within the call
HORIZON-MSCA-COFUND-2022. If the proposal gets approved, my lab will contribute
to the research innovation and training activities by developing and testing
interactive Design Analytics interfaces for augmenting creative decision-making
by assessing design form and performance data. Such design data can be
specified, generated, and evaluated through diverse activities, such as
requirements specifications, concept modelling, building information modelling,
sustainability, or environmental impact analysis. The lab’s projects are
aligned with the proposed research project. I will supervise one doctoral
candidate at my lab. I will also develop online or in-person workshops on
performance analytics of design alternatives by using the system prototypes we
plan to develop. We are waiting the results to be announced in 2023. Future REsearch Focus I will continue to focus on Design Analytics as main
research program. In particular, I will collaborate
with other researchers from academia and industry to develop design analytics
interfaces and integrate them into generative design ecosystems, such as
deep-learning surrogate models developed for performance prediction of complex
building projects that retrieves design data to make sense of the potential of
possible design solutions. Unlike simulation-driven performance prediction, it
is fast, but its results must be verified empirically. Reserch Funding My current research program—funded by NSERC Discovery
Grant, SSRCH Insight Development Grant, and various industry partnerships with
MITACS—focuses on exploring design analytics interfaces and their applications.
During my academic career, I received about CAD 700,000 in research funding,
which may be considered a small amount considering other disciplines. Their details can be found in the attached documents
presenting my research record. I must note that finding funding opportunities for
projects involving AEC is highly restricted. Only small funds can be allocated
by industry partners. Large amounts need consortiums or more extensive
partnerships, which I aim to develop as I progress in my current research program. Training and Student Supervision Philosophy My supervision philosophy is founded on three basic
principles: to create an environment for intellectual growth and collegiality;
to strive to develop idealism to foster interest, engagement, challenge, and
critical thinking; and to observe agility to make timely changes in research
considering real-world problems. In my lab, I aspire to train my graduate and
undergraduate students as independent researchers and practitioners with
interdisciplinary knowledge and skills in design, cognition, and software development.
They solve research problems by focusing on real-world cases from the domain of
built environment design. I view training in my Computational Design Lab as a
dynamic social activity that engages my students in critical thinking and
allows them to practice what they have learned to create new knowledge and
develop new skills. I encourage my students to work towards their academic and
life goals while also contributing to achieving others' goals. I encourage my lab members to meet with our external
partners, build their academic and industrial networks, and help them make
connections[††]. I value exposing my graduate students to different perspectives on
research. Through this approach, I observed the students develop ideas beyond
what I could imagine. My lab members learn to collaborate with researchers
outside their comfort zone. They assume ownership of their projects, and
although I guide them along with the research in my lab, their roles and the
projects they develop are shaped by their initiatives. As such, we write papers
together, make presentations to external bodies, and publish our work on social
media. Almost all of the papers I published include
members from the graduate lab. When possible, I choose to place their name as the
first author to promote ownership and leadership qualities. The testimonials
from my past graduate students demonstrate that my approach made a positive
impact on the students I graduated. In my lab, the students learn various aspects of
conducting research. I particularly emphasize selecting and applying design
research methodology and developing high- and low-fidelity software prototypes
as research artefacts to generate knowledge. In iterative research cycles, my
lab members select and design methods for data and task abstractions, identify
use cases, develop interfaces, and evaluate and reflect on the research
artifacts and process. My students can apply these transferable skills in
solving other domain problems. I focus on training my students in my lab with
meta-research knowledge and focusing on approaches to problem-solving,
methodology and system development. Equity,
Diversity and Inclusion In Computational Design Lab, I prioritize
inclusiveness, equity, and balance. We work as a group to provide strong
collegial support, and I have a broad collaborative network that benefits my
students. Through regular weekly lab meetings, my group updates other group
members, provides feedback on each other's work, and occasionally celebrates
group and individual successes. Advancing Equity, Diversity, and Inclusion
(EDI) at SFU is a priority in my academic and personal life. I encourage my
students to take online courses on EDI and actively participate in developing
ethics applications that involve how EDI can be factored into conducting
research. I engage my lab members in discussions on EDI and respect for each
other's life choices. As much as being interdisciplinary, my group is
culturally diverse. Our discussions and meetings are helping our team grow
strong and collaborate in harmony by recognizing individual choices and
cultural backgrounds. When students face life changes, I strive to help them maintain
a balanced and healthy life. When necessary, I encourage my students to seek
help from university members, such as the SFU Center for Accessible Learning,
SFU Health and Counselling, SFU TSSU, etc. I often consult with my colleagues
about such issues. Recent
Graduate Students I was the senior supervisor of 4 master theses (Abuzareiq, Alsalman, Zarei, and Piray) the last five
years. I was also the senior supervisor of two Ph.D. students, Koloric (2016) and Shireen (2020). Kolaric completed his Ph.D. in 2016 and received a researcher position at the
Georgia Institute of Technology. His system CAMBRIA enables designers to work
on an arbitrary number of alternative designs simultaneously. Sinisa and I
published five conference papers and two book chapters while he was a Ph.D.
student. I continue to collaborate with him. We submitted a journal paper
titled "The Case for Design Alternatives: In Support of Multi-State
Interaction" for review. Shireen studied how designers interact with a large number of alternatives in a simulated design task
environment. She questioned the choice overload problem in generative design
and developed cognitive patterns after analyzing designers' moves. We published
one journal and 11 conference papers, and two book chapters together while she
was a member of my lab. We developed three different ParaXplorer
prototype systems with varying functionality for parametric design. Shireen and
I are working on a journal paper titled "Spatial Metaphors to design user
interfaces for design exploration" and will submit it to the International
Journal of Architectural Computing. Shireen is a sessional instructor at SFU
and started her parametric design company in 2021. Four of my MSc students developed the design analytics
interfaces in the Computational Design lab. Supported by two separate MITACS
funding, we developed three fully functional prototypes of D-FlowUI, D-Sense,
and D-ART and two low-fi (D-CAT and D-ANZ) design analytics systems. Abuzaraiq's thesis focused on a Visual Analytics system for navigating generative
design spaces. He worked on D-Sense as a Web application that enables designers
to filter and cluster design alternatives on coordinated views. He used
similarity-based clustering and Pareto-Frontier to help designers manage a
given a large number of alternatives. We developed
D-Sense with our industry partners from Stantec Consulting, but also, we tested
its generality on shopping for cars online and searching 2D game layout designs
based on their difficulty levels. Abuzaraiq
co-authored six conference papers with me. We are writing a journal paper
titled "D-Sense: A Visual Analytics System for Navigating Design
Spaces." Ahmed was hired as a game developer and researcher for a
well-known BC mobile game company. He returned to our school as a doctoral
student and is working on generative (AI) systems in creative practices. Zarei's thesis investigated how design models can be sculpted in the early
phases of design by taking advantage of both flexibilities of directly
interactive geometric modelling and the computational power of parametric
design for performance analysis. She also worked on using design data as a
source of information to test the design models' validity. She worked on D-FlowUI
and D-CAT with our industry partner. Zarei also
developed a novel approach for comparative analytics as a part of D-CAT, a
successor of D-ANZ as a low-fi design data analytics system we developed with
Garg—another graduate student I supervised. The system was tested and used in
Stantec Consulting's research group. Zarei was hired
as a UI/UX specialist in a multimedia company focusing on VR and AR, and now she
is as a senior UI/UX researcher at Amazon. Alsalman's MSc thesis presented a Web-based system for exploring the potentials of
interactive data visualizations in collaboratively evaluating design
alternatives. The system is called D-ART (Design Reporting Tool). We developed
D-ART based on high-level system requirements we compiled that emerged through
a design-study method. D-ART is a platform-independent and online system combining
Computer-Supported Collaborative Work and Design Analytics. Alsalman
co-authored six conference papers and one journal papers with me. Osama is
working for Amazon as a software developer. Piray,
recently expended D-ART’s functionality by proposing coordinated annotation
features on form and performance data representations by stakeholders. The
annotations serves as communication means between the
design stakeholders and design decision-makers. [*] A more comprehensive report on my teaching can be found in my Teaching Portfolio. [†] Simon, H. A. (1973). The structure of ill structured problems. Artificial intelligence, 4(3-4), 181-201. [‡] Eric Horvitz. 1999. Principles of mixed-initiative user interfaces. In Proceedings of the SIGCHI conference on Human Factors in Computing Systems (CHI '99). Association for Computing Machinery, New York, NY, USA, 159–166. [§] Simon, H. A., & Newell, A. (1971). Human problem solving: The state of the theory in 1970. American psychologist, 26(2), 145. [**] The authors indicated bold characters are graduate or undergraduate students who participated in the research and writing the paper. [††] Computational Design Lab members’ presentations and lab projects can be found at https://www.youtube.com/@computationaldesign_SIAT |
||||||||||||||||||||||||||||
