Computers, Environment and Urban Systems 60 (2016) 1–11

Contents lists available at ScienceDirect

Computers, Environment and Urban Systems

journal homepage: www.elsevier.com/locate/ceus

Human behavior simulation in architectural design projects:

An observational study in an academic course
Seung Wan Hong a, Davide Schaumann b,⁎, Yehuda E. Kalay c
a
Department of Architecture, Inha University, 2S302, 100 Inha-ro, Nam-gu, Incheon 402-751, Republic of Korea
b
Faculty of Architecture and Town Planning, Technion – Israel Institute of Technology, Amado Building 117, Technion City, Haifa 32000, Israel
c
Faculty of Architecture and Town Planning, Technion – Israel Institute of Technology, Segoe Building 317, Technion City, Haifa 32000, Israel

a r t i c l e i n f o a b s t r a c t

Article history: While previous human behavior simulation research indicates that autonomous, goal-oriented, anthropomor-
Received 19 June 2014 phic Virtual Users support iterations in architectural design, it is still unknown how the simulation can be used
Received in revised form 17 July 2016 to iterate goals, potential solutions, and the match between them, and what can be learned from the simulation
Accepted 17 July 2016
process and results. To investigate these topics, this observational study tracks and records the progressive refine-
Available online 4 August 2016
ments of four authentic architectural design projects of students, during an academic course. Findings of this
Keywords:
study show that (1) the analytic experimentation of the simulation enables the students to iterate the goals rel-
Human behavior simulation evant to the physical properties of design solutions responding to the behavioral performances of heterogeneous
Virtual-users users, (2) the observable representation of dynamic users' behaviors inspires them to iterate the goals that reflect
Design iteration: iterating goals and solutions the psychological and social implications of solutions, (3) the simulation enables students to iterate potential so-
Learning from simulation lutions at different scales, ranging from masterplans, to prototypes, to design details, (4) in the final stage of iter-
Architectural design projects ating design solutions, the students iterated the parameters of Virtual Users to examine the full-performances of a
final solution under what-if scenarios related to human behavior aspects (5) the students learn in the process of
modeling detailed activities, and observing unexpected behavior outcomes during the simulation, experimenting
the relationship between the properties of users and design solutions, beyond what they presumed. The evi-
dence-based approach of this study reveals the applicability of human behavior simulation for the students' iter-
ation and in-depth learning in the search for an optimal match between built environments and human activities
in architectural design.
© 2016 Published by Elsevier Ltd.

1. Introduction interactions with the built environment, such as perception, cognition,

ergonomics, and social and cultural factors (Kalay, 2004).
Architectural design can be considered as a cycle of continuous At present, a common practice to assess human behavior issues dur-
learning by refining and reconstructing design goals via experimenting ing the design process involves extrapolation, a method for evaluating
potential solutions (Kalay, 2004; Schön, 1983). The process of refining the performance of the proposed solution by comparing it to previous
goals and solutions occurs by means of iterations. The aim of iterations similar precedents. Extrapolation requires indirect inference from pre-
is to modify either the design goals or solution to maximize the match cedents, norms, and regulations. Since the nature of design problems
between the two. This match is defined in terms of design performance. and solutions is essentially unique (Rittel & Webber, 1973), such infer-
As one of the most fundamental requirements to be a professional archi- ence inherently faces gaps between new solutions and previous refer-
tect, students are expected to be familiar with ways to systematically ences. A different method involves direct-experience behavior tests
frame architectural problems and iteratively test solutions until the that enable to observe the responses of actual human inhabitants to
achievement of a satisficing condition (Rittel, 1971). Yet, some of the some proposed solutions both in real or virtual settings However, con-
performance criteria that designers are expected to achieve in their de- structing realistic experimental settings both in real and virtual envi-
sign solutions pose serious challenges even to the most experienced ronments is expensive in terms of costs, times, and techniques.
practitioners. One of such challenges involves assessing the impact of Furthermore, the method relies on a limited number of users, whose re-
a physical setting on the people that will inhabit it. This is a complex sponses to the built environment may not be representative of the actu-
task due to the large number of variables that characterize human al users.
To overcome the shortcomings of these approaches, simulation
⁎ Corresponding author.
methods were proposed by several researchers (Chu, Parigi, Law, &
E-mail addresses: [email protected] (S.W. Hong), [email protected] Latombe, 2014; Ekholm, 2001; Kalay, 2004; Simeone, Kalay,
(D. Schaumann), [email protected] (Y.E. Kalay). Schaumann, & Hong, 2013; Tabak, Vries, & Dijkstra, 2010) to test design

http://dx.doi.org/10.1016/j.compenvurbsys.2016.07.005
0198-9715/© 2016 Published by Elsevier Ltd.
2 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11

hypothesis or to evaluate undesired consequences of design solutions as the relationships among decision variables are too complex, conflicting,
far as human behavior aspects are concerned. The power of human be- or dangerous to be tested in reality, such as space utilization in airports,
havior simulation lies in the ability to iterate the experimentation as hospitals, and industrial facilities (Wurzer & Lorenz, 2014). In particu-
many times as needed to test the relationship among variables lar, simulations allow testing “what-if” scenarios in virtual environment
interacting in complex and often unpredictable ways (Kalay, 2004). to explore possible solutions without commitment. Such strategy
The rules of behavior embedded in the proposed simulation system can be considered valuable to deal with the wicked nature of design
can be deduced from observations of real life phenomena in settings problems.
similar to the simulated ones, or can be extracted by previous studies While most simulation tools at architects' disposal focus on the pre-
conducted, for instance, by environmental psychologists. While the pre- diction and evaluation of physical aspects of a building design (e.g.
vious simulation studies focus on the technical and methodological as- structure, energy, lighting, noise), fewer approaches investigated the
pects about simulation tools and simulation model developments, possibility of using virtual humans to simulate the behavior of people
empirical case studies that investigate how designers use human behav- in existing or yet to be built environments. Such studies assume that
ior simulation are have not been reported. simulating human behavior allows designers to test the relationship be-
To address this issue, the present study intends to observe the use tween people, environment, and context through a trial-and-error
and application of human behavior simulation in authentic design pro- strategy, and therefore to make more informed decisions at each design
jects developed by students in an academic course. The aim is to inves- iteration.
tigate how students used human behavior simulation to iterate
architectural design goals and solutions, and what students learned
from using human behavior simulation in their projects. The explorato- 2.2. Human behavior simulation in architectural design
ry, evidence-based approach of this present study addresses the further
applicability and affordance of human behavior simulation to architec- Understanding the relationship between people and the environ-
tural design education and practice. ment that they inhabit is of primary concern in architecture and other
This paper is organized as follows: a literature review that explores social sciences. Whyte (1980) and Gehl (1987) investigated the capabil-
the applicability of human behavior simulation to design iteration is ities of public spaces in supporting or hindering the performing of deter-
provided in Section 2. The research questions are presented in Section mined activities, such as sitting, resting, or having social encounters. The
3. The research method, including the information on the participants, influence of the built environment on the behavior of individuals was
projects, behavioral modeling, and data collection, is stated in Section also examined in terms of proxemics (Hall, 1966), personal space
4, while the in-depth case analyses are detailed in Section 5. The conclu- (Sommer, 1969), and crowding (Altman, 1975; Stokols, 1976). Barker
sions and further discussions are outlined in Section 6. (1978), instead, proposed a more general approach to examine peo-
ple-environment interactions, based on the theory of “behavior setting”.
2. Literature review According to such theory, human behaviors happen in an environment
(milieu), which is circumjacent to the standing patterns of behavior. At
2.1. The role of simulations for design iterations the same time, the behavioral patterns are also circumjacent to the en-
vironment (p.27). The aforementioned user-centered and evidence-
The architectural design process involves a set of iterations that de- based design approaches have been promoted to support architects in
signers perform to achieve a specific purpose by means of “progressive designing settings that support user needs (Gifford, 2002).
refinements” (Akin, 2011). Within iterations, architects define and re- Capitalizing on such findings, pioneering studies proposed simula-
define design goals in the attempt to discover an acceptable solution tion methods to dynamically test the impact of a physical setting on
via navigating potential solutions (Kalay, 2004). Rittel (1971) clarified users' behavior. This can be done by defining computable correlations
that in design iterations, architects should examine the variables and between dynamic behaviors of agents, and the physical elements of
their relationship as follows: ‘under context C, design configuration D built environments in which behaviors are performed. Hillier and
will lead to performance P (1971, p.20)’. A design configuration includes Hanson (1984), Schultz and Bhatt (2012), and Hölscher et al. (2006)
the physical properties of a design layout generated and controlled by proposed a computational model to analyze the impact of the built en-
architects (e.g. form, shape). The context variables indicate the factors vironment on users' perceptual and cognitive abilities. Tabak et al.
and constraints that influence the design solutions, but that are not con- (2010) developed a schedule-based system to simulate human move-
trolled by architects (e.g. physiological, psychological aspects of people). ment in office spaces. Chu et al. (2014) expanded traditional agent-
The performance of a solution expresses the expected services emerg- based models for egress simulation to account for occupants' intimacy
ing by the match between design configurations and design goals (e.g. to social group, social structure, and social norms. Simeone et al.
safety, functionality). Iterations therefore allow (re)structuring goals (2013) and Schaumann, Kalay, Simeone, and Hong (2015) developed
and configurations to produce a desired performance under the given a coupled process-based and agent-based system to represent complex
context. use processes occurring in hospital settings.
The search for an appropriate solution, however, requires much time To simulate human behaviors in built environments, Kalay and his
and effort because of the “wicked” nature of design problems (Rittel & colleagues (Kalay, 2004; Kalay & Irazàbal, 1995; Steinfeld, 1992; Yan &
Webber, 1973). According to Rittel and Webber's definition, design Forsyth, 2005; Yan & Kalay, 2004) proposed the creation of autono-
problems can be considered “wicked” in the sense that are unique, mous, anthropomorphic, goal-oriented agents, called Virtual Users
they depend on the formulation of the problem, and they have no opti- (VUsers), which aim at representing the prospective users of a specific
mal solution. A systematic way to experiment the interactions between environment. VUsers are equipped with sensors able to detect environ-
the variables constituting goals and interactions is therefore needed mental stimuli (such as the location of building elements, and the pres-
(Kalay, 2004; Rittel, 1971), both in design practice and education. ence of other VUsers), and are able to respond to them according to
Simulation is a means to iteratively experiment the dynamic in- social and cultural-based behavioral rules. VUsers behave according to
teractions among variables to comprehend and predict the behavior a set of goals, which are achieved by performing a set of rules. Both
of systems, such as the performance of an artifact under determined goals and rules can be dynamically modified according to VUsers' per-
conditions (Shannon, 1975; Simon, 1999). Simulation methods sonal characteristics, such as personality traits, preferences or physical
enable successive testing of the behavior of a system to prove (or and psychological status, allowing designers to simulate the behaviors
disprove) some hypothesized performance or phenomena by way of particular groups of users (e.g. physical disabilities or mental dis-
of experimentation. These methods are particularly useful when eases) in not-yet built environments.
 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11 3

The concept of VUsers is implemented in the presented research by comprehension of the social aspects and affordances of their designs in
leveraging 3D game engine software. The game engine affords dynami- terms of users' activities.
cally representation of human users' behaviors in a three-dimensional Twelve undergraduate students (in their fourth and fifth year) at the
space, along with a computable set of goals and rules. The VUsers' geo- Faculty of Architecture and Town Planning participated in the course.
metrical properties, animated motions, as wells as goals, behaviors, and They formed four teams of three members each. The teams investigated
rules can be fully customized by designers. Therefore, it is possible for four different design projects – a building rooftop (Team D), a bus sta-
architects and planners to directly observe the dynamic, and descriptive tion (Team A), a commercial street (Team B), and a public park (Team
interactions of VUsers in the three-dimensional virtual environments in C), ranging from a building scale to the city scale. The details of the
which they are placed to analyze the appropriateness of the proposed four design projects are introduced in Section 5.
buildings to suit the intended behaviors of prospectus users (Fig. 1).
While these studies assumed that the analytical experimentations 4.2. Course contents
and observable representations in human behavior simulation support
for analysis and evaluation of user behavior in not-yet built environ- The curriculum of this course was sixteen weeks long, with class
ments, empirical case studies that investigate how students may use meetings once per week for 3 h. The first half of the course was aimed
human behavior simulation for design developments, are still missing. at providing students with both the theoretical framework underlying
In particular, it has not yet been reported how human behavior simula- the simulation approach, and with technical capability to manage the
tion enables the iterative process of refining design goals and solutions computational tools. Weekly assignments were required to practice
based on the observation and analysis of the simulated phenomena. the computational tools, and to create digital resources (such as VUsers
physical models and motions) that will be later be used in the simula-
3. Research questions tion. The second half of the course was dedicated to the development
of the students' projects: first, the students were asked to observe actual
This paper investigates ways in which students use human behavior peoples' behavior in similar settings to define VUsers' rules of behavior.
simulation and iterate their design solutions accordingly. The aim of re- After translating the rules into a computational language, the students
search is specified by the following three questions. First, how students generated a 3D digital model of their proposed design solution, and
iterate design goals using VUsers?; how they design VUsers in order to populated it with VUsers. Further modifications and developments of
iterate the design goals? Second, how students iterate potential solu- the design solution were made according to the simulation results.
tions using VUsers? Third, what students learn from using human be- The proposed simulation method was generated by a combination of
havior simulation?; in which ways VUsers influence iterating goals different software tools, used in different phases of the simulation pro-
and potential solutions? The questions aim to address high-level find- cess. Autodesk 3DS Max was used to model the physical settings in terms
ings and larger understanding of the ways in which students use and of geometrical features and material textures, and to create VUsers,
design VUsers. (both their geometrical features and animations). Dassault Systèmes'
3DVia Virtools was used as the 3D game engine, to set up and compute
4. Research methods the behavioral rules of VUsers. In this platform, students combined sev-
eral sets of ‘Behavioral Blocks (BB)’, programming language, to set up
4.1. Introduction VUsers' parameters and to populate the virtual settings (Fig. 2).

To investigate the aforementioned research questions, this study ap- 4.3. VUsers' behavioral models
plies qualitative research methods to analyze four design projects pro-
duced in a course titled “Designing Social Places in Multi-User Virtual VUsers' behavior models were based on four main parameters: (1)
Environment” held in spring 2013 semester at the Technion–Israel Insti- grid-based obstacle detection and avoidance, (2) social proximities,
tute of Technology in Haifa. This course aimed at providing a theoretical (3) social norms, and (4) personal traits. The path-finding system of
and practical framework for predicting and simulating the behaviors of VUsers is based on a spatial discretization mechanism that allows de-
prospective users in the proposed design solutions to promote students' tecting the obstacles placed on grids and calculating the fastest route

Fig. 1. VUsers' population in a hospital environment simulated in Dassault Systèmes' 3DVia Virtools.
(Technion-Israel Institute of Technology, 2012)
4 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11

Fig. 2. VUsers' population script using Virtools.

(Weizmann, M., Beer Sheva public square design, Technion-Israel Institute of Technology, 2013)

to a target avoiding such obstacles. The path-finding performances were 5. Case analysis
also influenced by social proximities, according to which agents tend to
maintain a certain distance from other agents, proportionate to the 5.1. Iterating goals
agents' degree of acquaintance (Hall, 1966). In addition, social norms
were also reflected in modeling VUsers' behavior. For instance, the 5.1.1. Posing initial goals: functionality of design configurations
VUser resembling an adult person tends to let children and old people While the presented projects investigated different design themes
enter or exit first when approaching a door or a narrow passage. Finally, (a building rooftop, a bus station, a commercial street, and a public
personal traits were also modeled, such as individual characteristics and park), they all aimed to pursue design goals of similar nature, which re-
preferences. Based on the combinatory use of the four behavioral lated to functional variables of design configurations, such as walking
models, the students created heterogeneous VUsers that generated bot- distances, dimensional appropriateness, and accessibility. The analytic
tom-up, interdependent interactions. experimentations of human behavior simulation have been therefore
relevant to set up and examine the functional aspects of goals. The stu-
4.4. Data collection and analysis scheme dents designed VUsers' physical properties and profiles (e.g. heights,
shapes, walking speeds) in order to measure and iterate the functional
Two types of data were collected and analyzed: the first type variables of design solutions accordingly.
consisted in records of weekly assignments, videos of the simulation For instance, the team A aimed to redesign a bus station in Haifa, Is-
outputs, snapshots, and diagrams, which tracked the students' pro- rael, to improve the connectivity between the security booth located on
gresses in the project development. The second type consisted in a sur- the bus station entrance, the retailers' area, and the waiting spaces in
vey questionnaire, which aimed at investigating the students' internal front of the bus platforms. The team goals thus aimed to support (1)
decision-making process that might not emerge explicitly from the sim- walking distances, times, and patterns between the entrance of the sta-
ulation results, as well as what type of significant information they re- tion, the designed waiting spaces, retailers, and bus platforms; (2) the
ceived from the simulation processes. The survey questionnaire dimensional appropriateness of waiting spaces; and (3) the accessibility
contained the following questions: and dimensional appropriateness of the new security entrance booth.
To measure such variables, the team A defined VUsers' shapes, roles de-
1) What are the goals that your team aims to achieve by means of pending on ages (elders and adults) and occupations (students and sol-
human behavior simulation? diers), and their relative walking speeds (Fig. 3a).
2) What is the hypothesis underlying your method to achieve the stat- The team B retrofitted a famous Tel Aviv street called Nahalat
ed goals? Binyamin to support both the market activities that take place there
3) Which specific aspects of your project aim at supporting users' twice a week, as well as new types of social activities, along with the
activities? existing ones. The team aimed to support (1) the dimensional appropri-
4) In what ways did human behavior simulation help your team to test ateness of the design of a series of loop-shaped structures to support
the proposed design solutions and the design goals that you previ- both pedestrian activities, such as sitting or eating, and vendors' activi-
ously stated? ties, such as exposing products for retailing, (2) pedestrians' movement
5) What did you learn from the simulation process that helped you de- in the street to verify that no major conflict emerged by the juxtaposi-
velop your design project? Please identify your design discoveries by tion of different types of activities. To examine the goals for the dimen-
comparing your initial design solution before the simulation process, sional accommodation and users' movement, the team designed the
with the new solutions you proposed, after the simulation process. shapes, walking speeds, standing postures, and gestures of VUsers, de-
pending on their profile types (pedestrians, vendors, tourists, and art-
In order to achieve the high-level of findings and larger understand- ists) (Fig. 3b).
ing on the students' use of human behavior simulation, the collected The team C designed a public park in Be'er Sheva, Israel. Focusing on
data were analyzed via the theoretical frames of Rittel (1971) and landscape design, this project aimed at providing different types of play-
Kalay (2004), introduced in Section 2.1. The qualitative analysis aims ground areas using natural elements like water, rocks, trees, and
to find (1) ways of iterating goals using VUsers, (2) ways of iterating meadows, to support family activities such as playing, walking, or run-
possible solutions using VUsers, (3) what students learn from the simu- ning on the grass area. Therefore, the team aimed to measure (1) occu-
lation process and results. pation capacity and density responding to an increasing number of
 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11 5

Fig. 3. (a) Walking distances, times, and patterns of VUsers (Burmad, M. et al., Hof Ha Carmel bus station design, Technion-Israel Institute of Technology, 2013). (b) Pedestrians' movement
in the street (Muzychuk, I. et al., Tel Aviv street design, Technion-Israel Institute of Technology, 2013).

users, (2) users' traffic routes. The shapes, walking speeds, and postures support varying types of social interactions according to the mid-term
of VUsers were developed accordingly – whether they are children or and long-term waiting of users. Furthermore, they added the intimacy
parents (Fig. 4a). levels to the VUsers accordingly, and evaluated whether the design con-
The team D aimed to revive the rooftop area of the Faculty of Archi- figurations of waiting spaces matched to these non-quantifiable aspects
tecture and Town Planning at the Technion in Haifa to support activities of design goals (Fig. 5a).
such as exhibitions, outdoor sports, fabrication labs, workshops, resting The team B set up the initial goals to support dimensional accommo-
areas, parties, and school-related activities. To achieve the goal, the dation of retailers and traffic movements of visitors. When the team ob-
team designed new outdoor furniture (e.g. a stage, benches, and tables), served the populated VUsers near the performance place, they framed
and re-organized the location and size of the existing coffee shop and another set of goals to satisfy (1) the privacy related to individual's pri-
hallway entrances, and they wanted to support (1) dimensional appro- vate and social zones, and (2) the sense of being with others, arising
priateness of zones, (2) users' movement inside the physical layout. The from holistic atmosphere of the market place including crowds of peo-
shapes, walking speeds, and postures of VUsers were designed depend- ple, unique design configurations, and situations (Fig. 5b).
ing on genders of students, and whether they are students or visitors To support the social events, the team D designed new outdoor fur-
(Fig. 4b). niture (e.g. a stage, benches, and tables), and re-organized the location
and size of the existing coffee shop and hallway entrances. Like the
5.1.2. Re-framing goals: psychological and social implications of design cases of the team A and B, as initial goals, the team aimed to support
solutions the functionality of design configurations such as fluent movements,
While using VUsers, students not only considered the functional as- optimal accessibility, and dimensional appropriateness. When the
pects of design goals, but also the psychological and social aspects of team observed the populated users, they reflected on how the users
human behaviors. For instance, they imagined how users would feel in feel (1) the social intimacy and (2) the sense of being with friends,
a designed spatial solution and how the solutions would support responding to a holistic phenomenon in the design configurations
intended social phenomena. Observing populated VUsers inspired the (Fig. 6a).
students to calibrate the ranges of goals from merely supporting func- The team C, instead, focused on the children user's psychological sat-
tionality of design configurations to in-depth reflections of users' mind isfaction. They designed artificial structures that function as attraction
affected by emerging behaviors of users responding to the design stimuli for children VUsers. According to particular environmental con-
configurations. ditions (e.g. wind), or at certain time of the day, either a windmill or a
For instance, the initial aims of team A were to support quick move- fountain produced entertaining effects to attract users (Fig. 6b). The
ments from waiting spaces to security booths, and provide varying seats team designed the children-shaped agents who possess preferences
in the waiting spaces. After observing the behavior of VUsers, students about the park's facilities. The design analysis aimed at verifying that
reflected on (1) the coziness of the waiting spaces from the view the proposed design solution allowed children VUsers to play in the pre-
point of users – if they were the VUsers, how would they feel, (2) the ferred zones of the park, without obstruction.
privacy, and (3) the intimacy to others responding to the simulated However, unlike the other three teams, the goal to support children's
phenomena. In result, the team framed a new set of design goal to psychological satisfactions was not inspired by the observation of the

Fig. 4. (a) Users' traffic routes (Weizmann, M. et al., Beer Sheva public square design, Technion-Israel Institute of Technology, 2013). (b) Dimensional appropriateness of zones (Lavi, H. et
al., School rooftop design, Technion-Israel Institute of Technology, 2013).
6 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11

Fig. 5. (a) Reflecting the coziness and privacy of users in the waiting spaces (Burmad, M. et al., Hof Ha Carmel bus station design, Technion-Israel Institute of Technology, 2013) (b)
Reflecting the sense of being with others in the street (Muzychuk, I. et al., Tel Aviv street design, Technion-Israel Institute of Technology, 2013).

micro-behavior of children VUsers. The goal seems to be a specification spaces (i.e. standing), while the spaces under the ramp were designed
of a more general goal aimed to satisfy functional requirements includ- for the security entrance, and private waiting spaces. At this stage, the
ing the holistic usability of park. One explanation could perhaps relate to team also examined the match between the prototype and the goals
the macro-scale of the site and the massive numbers of VUsers involving to support intimacy and feeling of being with others.
in this project. While the students analyzed holistic flows of peoples, the In the last stage, the team added more varying types and numbers of
large number of VUsers may obstruct the observation of micro-levels in- waiting spaces and retailers to the prototype, then examined the holistic
teractions among VUsers. match between the final solution and whole aimed goals. The examina-
tion was conducted both under normal situation, and also when special
5.2. Iterating solutions events occurred (a what-if scenario), such as the bus arrival to prevent
the creation of bottlenecks around the designed security booth. The
Based on the analysis of the collected data, students applied three aforementioned iteration processes of the team A is summarized in
different strategies of design iterations. The first strategy (used by Fig. 7.
team A and B) consisted in iterating several prototypes that contained In another example, the team B set up the prototype of a loop-
the key variables of design configurations, selecting one optimal proto- shaped street furniture for supporting sitting and selling activities. The
type, and then developing the solution. By prototype, we mean a local main variables of the design configurations were the dimensions and
experimental setup of partial, selected design elements. The second curvatures of the loop forms. The first prototype was evaluated in
strategy (used by team D) involved iterating several masterplans that terms of its ability to support retail activities from a geometric, dimen-
organize spatial elements, and then selecting one of them to further de- sional, and ergonomic point of view. The second prototype, instead, in-
velop it. By masterplan, we mean a holistic coordination and integration vestigated the possibility of assembling multiple loop forms to support a
of all given constraints and elements of design. The last strategy (used fluent pedestrian circulation. In the first and second prototypes, the de-
by team C) involved organizing activity zones at the masterplan level sign variables were the locations and compositions of the loop shaped
to achieve holistic functions. After a preliminary organization of the de- structure, and such configurations were tested iteratively to avoid bot-
sign solution, the team members iterated VUser variables while keeping tlenecks of pedestrians.
fixed the design solution. In the last prototype, the team developed two prototypes of loop
units: one consisted of an enclosed O-shaped unit to support static ac-
5.2.1. Iteration type 1: from prototypes to masterplans tivities such as displaying retail products, sitting and eating. The other
Team A and B set up several potential design prototypes and exam- consisted of a semi-open S-shaped structure that supported dynamic
ined which of them matched the goals focusing on functional uses. users' activities such as street art expositions and performances.
Then, they selected a preferred prototype and developed it to propose In the following simulations, the students tested how the position of
a holistic solution. When the teams added more programs to the select- the different loop units supports the functional goals like the circulation
ed prototype, they examined matches between layouts and goals to patterns and flows of pedestrians. The integrated prototypes supported
support psychological and social behaviors, introduced in the previous more programs, such as street exhibitions and artists' performances and
section. the relevant simulations aimed to confirm not only the aforementioned
The team A set up a prototype that included a bridge-shaped waiting functional matches, but also qualitative implications such as intimacy
space, represented in the initial sketch and digital model, and discov- and sense of being with others. A what-if event scenario aimed at test-
ered that the prototype geometry caused users' bottlenecks. At this ing changes in users' circulation patterns and crowding response to a
stage, the main variables of design configuration were the location street performance. Such performance test was conducted during the
and size of waiting spaces and pathways. In the second prototype, last design iterations. Based on the simulation results, the team B iterat-
they tested a square-typed waiting space and found that the prototype ed the location of S shaped units and slightly carved the ground to create
was also inefficient due to the unorganized circulation patterns and the seats for pedestrians. Fig. 8 lists up the iteration processes of the team B.
lack of waiting spaces. The team also examined a circulation-typed
waiting space and discovered the prototype blocked the passage of 5.2.2. Iteration type 2: from masterplans to details
other users' moving toward the bus. Team D iterated different masterplans to match functionality goals.
As last prototype, the team proposed a new type of entrance booth, During the iterations, one masterplan was selected, and parts of the se-
which was developed on two floors connected by a ramp. Both the ramp lected masterplan were developed in detail in order to satisfy both func-
and the second floor were intended to be used both as long-term tional and social aspects of goals. This design strategy, therefore
waiting spaces (i.e. studying and eating) and as short-term waiting involved first a series of iterations at the macro-level zones, and then
 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11 7

Fig. 6. (a) Reflecting the sense of being with others in the coffee shop, a zone of the rooftop design (Lavi, H. et al., School rooftop design, Technion-Israel Institute of Technology, 2013). (b)
VUsers' behaviors responding to environmental stimuli (Weizmann, M. et al., Beer Sheva public square design, Technion-Israel Institute of Technology, 2013).

a progressive refinement of the design solution by scaling down toward 5.2.3. Iteration type 3: iterating VUser variables in masterplans
the analysis of ergonomic aspects of human behavior. The last type of design iteration involved the systematic modification
More specifically, the team D interlinked different programs of the of VUsers' parameters to evaluate the match between a fixed solution and
proposed rooftop design. In particular, they iterated different design solu- goals under varying contexts. In previous simulation studies, this type of
tions to organize the different zones from the location of the coffee shop iteration is commonly used to predict negative side-effects such as bottle-
to enable students to bring warm food from the coffee shop to the close necks and traffic jams, and is usually conducted at the end stage of design.
by sitting and eating areas. In the first masterplan layout, they discovered Unlike other teams, when the team C started the project, they had an al-
that the proposed spatial organization caused bottlenecks in front of the most fixed landscape architecture design, developed in the previous de-
coffee shop exit when many agents attempted to enter into the rooftop. sign studio, and thus examined the full performances of the solution,
In the second layout, the team repeated the analyses on the movement rather than iterating goals and potential solutions.
patterns, dimensional appropriateness, location, and people's density in In particular, the team C interlinked different playground zones to
the retrofitted coffee shop, hallway, and newly designed rooftop area. support holistic use of the park. When they organized the activity zones,
In the next step, the team selected the second layout and examined the students created a script that automatically assigns a target to each
the possible conflicts in the case of a school-related event, such as the type of users according to a random function. This target can be dynami-
exposition of students' work in an exposition area. After being satisfied cally modified according to environmental stimuli such as an amphithe-
with the holistic functionality performance of the rooftop zones, the stu- ater, a water pond, a meadow area, a playground area with swings, and
dents then iterated the key parts of the rooftop design and coffee shop an open plaza with water sprinklers. The script aimed to iterate the intrin-
in detail to simulate micro-levels of activities. At this stage, the team fo- sic preference of VUsers, a variable of users, in order to examine the occu-
cused ergonomic aspects, and in particular on the correct dimensioning pation capacity and people's density in the designed zones. In addition,
of the space among the tables to support the drinking activities as well the team iterated population amount of users from 100 to 600 (Fig. 10).
as the intimacy and sense of being with others in reflection of observing After observing the simulation results, the team found that the solution
social interactions between VUsers. The iteration processes of the team matches the proposed goals of supporting fluent movements between
D is summarized in Fig. 9. the key places of the park, even under the high-population.

Fig. 7. An example of iterating solutions from prototypes to masterplans (Burmad, M. et al., Hof Ha Carmel bus station design, Technion-Israel Institute of Technology, 2013).
8 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11

Fig. 8. An example of iterating solutions from prototypes to masterplans (Muzychuk, I. et al., Tel Aviv street design, Technion-Israel Institute of Technology, 2013).

5.3. Lessons from human behavior simulation representation of a simple activity, such as eating, required a lot of
details to be specified, like the position in which users can eat (stand-
5.3.1. Learning from representing explicit details of human behavior ing or sitting) as well as other dynamic components such as eating
A lesson mentioned by students involves the ability of the simu- while walking or eating in a hurry. This investigation of ordinary ac-
lation to describe human activities at different level of details. Such tivities at this level of detail led to several considerations related to
description inspired the continuous iteration and specification of de- the ability of potential solutions to support different aspects related
sign goals and solutions. For instance, the team A reported that the to the analyzed activity.

Fig. 9. An example of iterating solutions from masterplans to details (Lavi, H. et al., School rooftop design, Technion-Israel Institute of Technology, 2013).
 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11 9

Fig. 10. An example of iterating user variables in masterplans (Weizmann, M. et al, Beer Sheva public square design, Technion-Israel Institute of Technology, 2013).

In the interviews, to explain the lesson gained from the simulation, during the analysis phase, since behaviors in a rush hour could legiti-
one participant stated that in his previous studio practices, human be- mate people to behave differently to shorten the distance to their desti-
haviors were over-simplified like a straight line representing people's nation. This discovery provided the opportunity for in-depth reflection
movement from A to B. When he used VUsers, he found that the simu- on human behaviors beyond lay-down assumptions.
lated behaviors were not simple and predictable, unlike his original
imagination, and thus his team refined design solutions based on 6. Conclusion and discussion
what they discovered and observed in the simulation.
The paper presents an observational on how students use human be-
5.3.2. Learning from unexpected human behavior outcomes haviors simulation to iterate design goals and solutions. While previous
The students have learned ways of conducting ‘stochastic trial-and- studies assumed that analytic experimentations and representations of
error’ experimentations and capturing ‘dynamics and unexpectedness human behavior simulation merely impact on the quantitative func-
results of human behavior analysis’ in use of the simulation. Students tionality of design solutions such as fluent movements, dimensional ap-
have experimented the matches between the variables of design config- propriateness, and walking distance (Ekholm, 2001; Tabak et al., 2010),
urations and goals, and thus learned which parts of current design con- findings of the present study indicate that students iterated goals rele-
figurations should be specified and modified to suit to the goals under vant to both functional, psychological and social aspects of design solu-
the given contexts. tions. The significance of observing the dynamic interactions between
While the students conducted such experimentations, VUsers' pa- VUsers is highlighted in the empirical use of the simulation. Observing
rameters' setting enabled “serendipitous” discoveries. For instance, in three-dimensionally represented humans, in fact, inspired to reflect
the case of the team B, the unintended behaviors of pedestrian VUsers what kind of feeling and attractiveness were aroused by experiencing
trespassing across the retailers' private spaces inspired the developed the designed environments beyond geometrical and physical aspects
the semi-closed S type unit. In another example, when the team A ex- of design configurations, and thus facilitated setting up further non-
amined the bridge typed waiting space, a solder VUser jumped over quantifiable design goals.
the bridge located on the second floor and arrive at the bus parking Resembling with the finding of this study, developing spatial con-
lot. The team has not counted that the shortest-path finding, which ceptions in reflection of human bodies representation is known as “met-
compute the relationship of physical properties between users and so- onymic procedure” (Anderson, 2002; Frascari, 1987). The term means
lutions, simulated this type of violation. They initially assumed that that when architects represent the actions of human bodies in their
the VUser use a staircase like ordinary situation in reality. The team drawings (e.g. touch, lurk behind, and observe), they deduce and elabo-
learned that such abnormal behavior of the VUser should be accounted rate the details of ongoing design solutions from the experiences and
10 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11

statues of the human bodies in the architectural drawings. Such mental VUsers, spontaneous social interaction among agents and collaborative
synchronization of architects to the represented human bodies impacts behaviors to reach common goals required scripting skills far beyond
on the qualitative aspects of evaluation. For instance, the bus station the scope of this course. As Simeone et al. (2013), Chu et al. (2014),
team speculated and evaluated a feeling of privacy in the waiting spaces, and Schaumann et al. (2015) argued, a higher-level coordination system
and the rooftop team also discussed a feeling of coziness in the coffee to experiment and simulate the complex social interactions of VUsers is
shop layout. This finding and interpretation explain why designing required.
VUsers incorporating geometrical and material attributes (rather than Despite these biases and limitations, this study addresses the find-
using a common dot-like representation) is important for architectural ings in the empirical use of human behavior simulation for iterating stu-
design analysis and evaluation. dents' architectural design projects. The findings of the present study
Another finding of this study indicates that human behavior simula- suggest the possibility that human behavior simulation contributes to
tion was used for iterating solutions to develop prototypes to iterate goals and solutions, and their matches via analytical experimen-
masterplans, and specify masterplans to details. At the end stage of de- tations and explicit, observable representations. The simulation is able
sign, the students iterated the parameters of VUsers to evaluate full-per- to inspire students to investigate functional, psychological, and social
formance of a fixed design solution. In addition, while the early stage of aspects of design solutions, and it might overcome the absence of ana-
iterating solutions focused on users' behaviors in daily routines, the end lytic approaches and in-depth reflections to comprehend users' activi-
stage of iteration examined a match between a fixed solution and goals ties in students' projects.
under special events and “what-if” conditions. The design configura-
tions were expected to serve appropriate performances under both Acknowledgments
daily and special contexts, even the worst condition. In observations,
the common what-if context was a sudden change of population and This research was sponsored by a grant from the European Research
density. The types of iterations required micro-, local-interactions Council (FP-7 ADG 340753) and a grant from the Israel Science Founda-
among VUsers, as well as at a macro-level, producing an emergent, ho- tion (890015). We also appreciate the help and motivation of all the stu-
listic view of use patterns. From a methodological view, iterating solu- dents who took part in the course “Designing Social Places in Multi-User
tions from prototypes to masterplans resembles with a bottom-up Virtual Environments”, held at the Faculty of Architecture and Town
approach that adds more design configurations to the optimal proto- Planning, the Technion, and the tireless support of the technical staff –
type, while iterating solutions from masterplans to detail is more likely Dan Klinger and Doron Zur.
a top-down approach that specifies parts of a holistic plans. Therefore,
the applicability of human behavior simulation scopes those two
approaches. References
The last finding of this study informs that learning from the simula- Akin, O. (2011). Iteration: What is it good for? Spatial Cognition for Architectural Design,
tion involves not only well-established, systematic experimentations, 2011, 1–24.
but also incomplete definition of VUsers' behavior properties. When Altman, I. (1975). The environment and social behavior: Privacy, personal space, territory,
crowding. Monterey, Calif., Brooks/Cole Pub.
students miscounted the relationships between physical properties of Anderson, A. T. (2002). On the human figure in architectural representation. Journal of
VUsers and design solutions, unexpected and varying matches between Architectural Education, 55, 238–246.
them were generated. These unexpected human behavior outcomes in- Barker, R. G. (1978). Habitats, environments, and human behavior: Studies in ecological psy-
chology and eco-behavioral science. San Francisco, CA: Jossy-Bass Publishers, San
spired the in-depth reflection to iterate goals and solutions beyond pre- Francisco, California.
assumptions in students' mind. Bhatt, M., Schultz, C., & Freksa, C. (2013). The ‘Space’ in spatial assistance systems: Con-
The present study also notes the inherent biases and limitations aris- ception, formalisation, and computation. In T. Tenbrink, J. Wiener, & C. Claramunt
(Eds.), Representing space in cognition: Interrelations of behavior, language, and formal
ing from the qualitative case analysis. First, the course contents and pro- models. Series: Explorations in Language and Space. Oxford: Oxford University Press.
cedures were designed by teachers, and the design solutions were also Chu, M. L., Parigi, P., Law, K., & Latombe, J. C. (2014). SAFEgress: A flexible platform to
evaluated by them. Therefore, the students were also motivated to study the effect of human and social behaviors on Egress performance. 2014 Proceed-
ings of the Symposium on Simulation for Architecture and Urban Design (pp. 35–42).
apply the human behavior simulation to fulfill the course requirements
Ekholm, A. (2001). Modelling of user activities in building design. 2001 19th Proceedings of
and to satisfy the teachers' expectations. In addition, the statistical sig- Education and Research in Computed Aided Architectural Design in Europe (pp. 67–72).
nificance and rigorous comparative approach are limited in this study. Frascari, M. (1987). The body and architecture in the drawings of Carlo Scarpa. RES:
Anthropology and Aesthetics, 14, 123–142.
The total project samples are only four, and they have the diverse scopes
Gehl, J. (1987). Life between buildings: Using public space. English translation. New York:
of aims, topics, and parameters of users. The students also presented dif- Van Nostrand Reinhold Company.
ferent technical skills and attitudes. Gifford, R. (2002). Environmental psychology: Principles and practice (3rd. ed.). Canada:
Despite the fact that the simulation aimed to test possible scenarios Optimal Books.
Hall, E. T. (1966). Hidden dimension. New York: Doubleday.
that will occur in a proposed physical setting, discussing the validity of Hillier, B., & Hanson, J. (1984). The social logic of space. Cambridge: Cambridge University
the simulation, in terms of its accuracy in comparison to real-world phe- Press.
nomena, was beyond the scope of this study. This study, instead, found Hölscher, C., Meilinger, T., Vrachliotis, G., Brosamle, M., & Knauff, M. (2006). Up the down
staircase: Wayfinding strategies in multi-level buildings. Journal of Environmental
that simple behavioral rules related to users' spatial behavior, when Psychology, 26, 284–299.
combined and applied to many VUsers at the same time, generated Kalay, Y. E. (2004). Architecture's new media: Principles, theories, and methods of computer-
complex and unpredictable human behavior patterns in a proposed fu- aided design. Cambridge, MA: MIT Press.
Kalay, Y. E., & Irazàbal, C. E. (1995). Virtual Users (VUsers): Auto-animated human-forms
ture environment. for representation and evaluation of behavior in designed environment. Technical re-
In addition, this paper acknowledges that the negative aspects of port. Berkeley: University of California.
human behavior simulation arose from the inherent limitation of the Rittel, H. (1971). Some principles for the design of an educational system for design.
Journal of Architectural Education, 25, 16–27.
used 3D game engine and students' skills in modeling and coordinating
Rittel, H. J., & Webber, M. M. (1973). Dilemmas in a general theory of planning. Policy
VUsers' behaviors. For instance, despite the fact that research in way- Sciences, 4, 155–169.
finding proved that users rely on spatial and cognitive capabilities to Schaumann, D., Kalay, Y. E., Simeone, D., & Hong, S. (2015). Simulating human behavior in
not-yet built environments by means of event-based narratives. Proceedings of the
navigate through buildings (Bhatt, Schultz, & Freksa, 2013; Hölscher et
Symposium of Simulation for Architecture and Urban Design (SimAUD) (pp. 7–14).
al., 2006), because of the simulation engine complexity of use, students Schön, D. A. (1983). The reflective practitioner. USA: Basic Books, Inc.
relied on a pre-coded path-finding algorithm that led virtual users to Schultz, C., & Bhatt, M. (2012). Multimodal spatial data access for architecture design as-
their target using the shortest route. Further difficulties arose in ac- sistance. Artificial Intelligence for Design, Engineering, Analysis and Manufacturing, 26,
177–203.
counting for serendipitous VUsers' encounters during the simulation Shannon, R. E. (1975). System simulation: The art and the science. NJ, Prentice Hall: Engle-
process. Despite the fact that users can detect the presence of other wood Cliffs.
 S.W. Hong et al. / Computers, Environment and Urban Systems 60 (2016) 1–11 11

Simeone, D., Kalay, Y. E., Schaumann, D., & Hong, S. (2013). Modeling and simulating use Whyte, W. (1980). The social life of small urban spaces. New York: Project for Public Spaces,
processes in buildings. 2013 Proceedings of Education and Research in Computed Aided Inc.
Architectural Design in Europe (pp. 59–66). Wurzer, G., & Lorenz, W. E. (2014). Causality in hospital simulation based on utilization
Simon, H. S. (1999). The science of the artificial (3th ed.). Cambridge, MA: MIT Press. chains. 2014 Proceedings of the Symposium on Simulation for Architecture and Urban
Sommer, R. (1969). Personal space; The behavioral basis of design. Prentice-Hall: Engle- Design (pp. 161–164).
wood Cliffs, N.J. Yan, W., & Forsyth, D. (2005). Learning the behavior of users in a public space through
Steinfeld, E. (1992). Toward artificial users. In Y. E. Kalay (Ed.), Evaluating and predicting video tracking. Proceedings of IEEE Workshop on Applications of Computer Vision
design performance (pp. 329–346). New York: Wiley Interscience. (WACV) (pp. 370–377).
Stokols, D. (1976). The experience of crowding in primary and secondary environments. Yan, W., & Kalay, Y. E. (2004). Simulating the behavior of users in built environments.
Environment and Behavior, 8, 49–86. Journal of Architectural and Planning Research, 21, 371–384.
Tabak, V., Vries, B., & Dijkstra, J. (2010). Simulation and validation of human movement in
building spaces. Environment and Planning B: Planning and Design, 37, 592–609.