109
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
1
Lok Hang Cheung,
2
Juan Carlos Dall’Asta
1
Xi’an Jiaotong-Liverpool University. lokhang.cheung19@student.xjtlu.edu.cn. ORCID: 0009-0001-2911-3733
2
Xi’an Jiaotong-Liverpool University. juancarlos.dallasta@xjtlu.edu.cn. ORCID: 0000-0002-8600-2757
Human-computer Interaction (HCI) Approach to
Articial Intelligence in Education (AIEd)
in Architectural Design
Interacción Humano-computadora (HCI), Enfocada para uso en
Inteligencia Articial de Educación (AIEd) y Diseño Arquitectonico
EÍDOS N
o
23
Revista Cientíca de Arquitectura y Urbanismo
ISSN: 1390-5007
revistas.ute.edu.ec/index.php/eidos
Recepción: 29, 09, 2023 - Aceptación: 03, 12, 2023 - Publicado: 01, 01, 2024
Abstract:
The eld of articial intelligence (AI) in architectural
design (AIEd) has experienced signicant growth, and
there is great potential for the application of Generative
AI (GAI) in architectural design education. However,
addressing challenges associated with AI is important.
These include overhyped speculation, hidden
inherent drawbacks such as fairness and ethics,
and a trend of lacking human interaction when AI is
involved. To tackle these issues, this paper introduces
a triangulated research framework encompassing
three key perspectives: vision, technology, and user
acceptance. This framework aligns with Human-
computer Interaction (HCI) principles, AI technology
development, and past experiences in AIEd. By
adopting this multi-perspective analysis approach,
the paper aims to comprehensively understand the
phenomena surrounding AI in architectural design.
Furthermore, the research presented in this paper goes
beyond theoretical discussions and illustrates how the
research ndings are applied in practice. It showcases
the design of an ongoing architectural design course
that incorporates the insights gained from the research.
Three key observations from the ongoing designed
modules indicate the need to shift the focus towards
integrating multi-modal AIs and existing parametric
tools. Secondly, it is essential to emphasise AIs as
design partners rather than making assumptions about
AIs’ specic uses at different stages. Lastly, providing
user-friendly tools and theoretical foundations motivates
students to explore beyond the design process,
expanding their research and design boundaries.
Keywords: Human-computer interaction (HCI); AI
in education (AIEd); architectural design process;
architecture education.
Resumen:
El campo de la inteligencia articial (AI) en el diseño ar-
quitectónico (AIEd) ha experimentado un crecimiento
signicativo y existe un gran potencial para la aplica-
ción de la AI generativa (GAI) en la educación del dise-
ño arquitectónico. Sin embargo, es importante abordar
los desafíos asociados con la AI. Estos incluyen es-
peculaciones exageradas, inconvenientes inherentes
ocultos como la equidad y la ética, y una tendencia a la
falta de interacción humana cuando se involucra la AI.
Para abordar estos problemas, este artículo presenta
un marco de investigación triangulado que abarca tres
perspectivas clave: visión, tecnología y aceptación del
usuario. Este marco se alinea con los principios de in-
teracción humano-computadora (HCI), el desarrollo de
tecnología de AI y las experiencias pasadas en AIEd.
Al adoptar este enfoque de análisis multi-perspectiva,
el artículo tiene como objetivo comprender de manera
integral los fenómenos que rodean a la AI en el diseño
arquitectónico. Además, la investigación presentada
en este artículo va más allá de las discusiones teóri-
cas e ilustra cómo se aplican los hallazgos de la in-
vestigación en la práctica. Se muestra el diseño de un
curso de diseño arquitectónico en curso que incorpora
las ideas obtenidas de la investigación. Tres observa-
ciones clave de los módulos diseñados en curso indi-
can la necesidad de centrarse en la integración de AI
multimodal y herramientas paramétricas existentes. En
segundo lugar, es esencial enfatizar que las IAs son
socios de diseño en lugar de hacer suposiciones sobre
sus usos especícos en diferentes etapas. Por último,
brindar herramientas amigables para el usuario y fun-
damentos teóricos motiva a los estudiantes a explorar
más allá del proceso de diseño, ampliando sus límites
de investigación y diseño.
Palabras clave: Interacción humano-computadora
(HCI); AI en educación (AIEd); proceso de diseño ar-
quitectónico; educación en arquitectura.
110
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
1. INTRODUCTION
Articial Intelligence (AI) provides
pedagogical opportunities (Celik, 2023)
and has been a discipline in the eld of Ar-
ticial Intelligence in Education (AIEd) for
over 30 years (Hwang et al. 2020; O’Shea
& Self, 1986). In recent years, there has
been increasing exploration of AI in the
context of AIEd, as evidenced by several
researchers (Celik, 2023; Holmes & Tuomi,
2022; Namatherdhala et al., 2022; Pham &
Sampson, 2022).
AIEd can be categorised into three
main areas. The rst category focuses on
management tasks, such as the automat-
ed marking of exams and the efcient ar-
rangement of class schedules. The second
category is teacher-focused and involves
designing and preparing teaching materi-
als using AI technologies. Finally, the third
category is student-focused, emphasising
the use of AI to enhance learning incen-
tives, processes, and outcomes (L. Chen
et al. 2020; X. Chen et al. 2022; Haderer &
Ciolacu, 2022; Huang et al. 2021). These
categories encompass various aspects of
AIEd and highlight the broad range of ap-
plications and benets that AI can bring to
the eld.
In this research, our primary focus
is on the second and third categories of
AIEd in the context of architecture educa-
tion. It is worth noting that digital tools have
played a signicant role in architectural
design studios (Hettithanthri & Hansen,
2022), particularly in the aftermath of the
COVID-19 pandemic (Bakir & Alsaadani,
2022), which led to the widespread adop-
tion of digital education worldwide. While
AIEd has been primarily studied in the
eld of architecture in terms of manage-
ment tasks, such as assessment (Smolan-
sky et al. 2023; Tack & Piech, 2022), and
teaching and learning, such as automatic
generation of facades (Sun et al. 2022)
and automatic oorplan generation (We-
ber et al. 2022), the potential of AIEd in the
student-focused aspect remains largely
unexplored within the architectural design
process.
Therefore, to address this gap,
this research focuses on how architectural
design studio modules can be designed
within the context of AIEd. This will be ac-
complished by considering past teaching
experiences and the rapid development of
AI technology. By exploring the integration
of AI technologies in the architectural de-
sign studio, this research seeks to uncover
new possibilities for enhancing the learn-
ing experience and outcomes for students
in architecture education.
2. THREE OBSERVATIONS AND
RESEARCH GAPS
2.1 Observation 1: Overhype of AI
The phenomenon of overhyp-
ing AI applications, much like any other
technological breakthrough, has been
acknowledged in the literature (Nemorin
et al. 2022; Perez, 2002). As such, it is
imperative to exercise caution and avoid
becoming overly optimistic about the
widespread application of AI tools in the
architectural design process without a
thorough understanding of their capabili-
ties and limitations.
One area of AI that shows prom-
ise for integration into education is Natural
Language Processing (NLP), which can
enhance various aspects of education by
enabling computers to understand and in-
teract with human language (Villegas-Ch
et al. 2020).
2.2 Observation 2: Inherent challenges of AI
The use of AI in various contexts
presents challenges that need to be ad-
dressed, including concerns related to
fairness and bias (Huang et al. 2021;
Qiu, 2020). Additionally, there is a need
to ensure that AI is used practically and
meaningfully (Chatterjee & Bhattacharjee,
2020). These challenges are particularly
relevant when considering Generative Ar-
ticial Intelligence (GAI) systems.
One of the major obstacles to
utilising GAI is the limited understanding
of its inner workings. GAI models, often
based on deep neural networks, are in-
herently complex and difcult to compre-
hend in technical and conceptual ways.
The “black-box” nature of GAI systems
111
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
poses challenges in understanding their
decision-making processes. However,
it is important to recognise that continu-
ous learning and practice can contribute
to an increased understanding over time,
echoing the signicance of conversational
properties in human-computer interactions
(Cheung & Dall’Asta, 2023; Cheung et al.
2023b).
The learning curve for students
from non-computational disciplines to ap-
ply advanced computational tools in the ar-
chitectural design process is a well-known
challenge (Qattan & AUSTIN, 2016). To
address this, it is crucial to consider a re-
alistic learning curve by balancing the se-
lection of user-friendly user-interface tools
and novel technologies. The combination
of online tutorials and in-class exercises is
well-received by students (Holzer, 2019).
Another important consideration
is the presence of bias in data utilised by
AI systems (Baidoo-Anu & Owusu Ansah,
2023). While it is technically possible to miti-
gate bias through continuous ne-tuning,
this process often requires an unrealistic
timeline for comprehensive bias elimination.
2.3 Observation 3: Lack of Human
Collaborations
The current application of AI often
lacks human interaction, which has raised
concerns about the loss of human purpose
in the face of the Pandemics of Today’s AI,
as coined by Pangaro (2020). This aware-
ness has prompted a need to reevaluate
the relationship between humans and AI
(Markauskaite et al. 2022). Achieving a
balance between human-centred educa-
tion and AI applications has become cru-
cial (Nabizadeh et al. 2023).
Despite recognising this phenom-
enon, the actual implementation of these
ideas within the architectural discipline
still needs to be explored. It is important to
note that attributing AI technology as the
sole cause would provide an incomplete
picture, as this observation should also be
examined from a human perspective.
This tendency of modern human
consciousness to overlook its intercon-
nected nature is a concern (Goodbun &
Sweeting, 2021; Sweeting, 2022). It aligns
with the worry from the architectural de-
sign discipline that the master-slave model
(Glanville, 1992) is more akin to an “idiot-
slave” model (Negroponte, 1976); tools are
gradually taking over human designers.
The concept of conversation,
which involves maintaining differences
and achieving agreement among differ-
ent parties, differs from mere communica-
tion, where identical information is passed
and received (Goodbun & Sweeting, 2021;
Pask, 1980). Human-Computer Interaction
(HCI) has the potential to balance between
overhyped AI applications and a human-
centric approach.
A framework proposing Human-
Computer Interaction as an approach
has been previously suggested (Cheung
& Dall’Asta, 2023; Cheung et al. 2023b).
However, applying this framework in de-
signing an AI-integrated course remains
largely unexplored within the architectural
eld. Further research is needed to inves-
tigate and develop effective strategies for
incorporating HCI principles into design-
ing AI-integrated courses.
3. RESEARCH QUESTIONS
The research seeks to address the
following three main research questions,
which correspond to the theoretical foun-
dation of Human-Computer Interaction
(HCI), AI technology, and the practical im-
plications of their connection:
1. What and how can HCI principles
be effectively applied in designing
an AI-integrated architectural design
course?
2. What AI technologies should be em-
ployed in the architectural design
studio context?
3. How can a balance between inte-
grating novel AI technologies and
user acceptance be achieved?
4. RESEARCH FRAMEWORK AND
METHODS
The study will adopt three distinct
perspectives to address these research
112
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
questions: vision, technology, and user
acceptance.
The vision perspective will explore
the application of HCI principles in the
design of an AI-integrated architectural
design course, considering the peda-
gogical aspects and learning objectives
involved. The technology perspective will
examine various AI technologies suitable
for implementation within the architectural
design studio, considering factors such as
automation, generative design, and data
analysis. The user acceptance perspec-
tive will focus on understanding students’
and educators’ attitudes, perceptions, and
preferences towards integrating AI into the
architectural design process.
The research will employ qualita-
tive method primarily. It involves literature
reviews and case studies to gain insights
into the theoretical foundation of HCI, AI
technology, and user acceptance, fol-
lowed by case studies of experience to
gather data on user acceptance and the
effectiveness of the AI-integrated architec-
tural design course. Examples of students’
works from each experience were extract-
ed and displayed for intuitive understand-
ing and comparison.
To develop forward from the re-
search above, an architectural design
module is designed and implemented for
the academic year 2023-2024, incorporat-
ing the ndings and insights obtained from
the previous research phases.
Figure 1. Vision-Technology-User Acceptance Research
Framework
Source: created by the author.
5. RESEARCH - THREE PERSPECTIVES
5.1 Vision
Corresponding to the rst research
question:
The vision perspective of the re-
search focuses on Human-Computer In-
teraction (HCI), which originates from the
eld of cybernetics (Cheung et al. 2023b).
HCI establishes the foundation for un-
derstanding how humans interact with
computers and machines. With the rapid
development of various AI technologies,
they can provide insights into answering
Cedric Price’s famous question, “Technol-
ogy is the answer, but what was the ques-
tion” (Price, 1979).
Considerations of HCI have been
raised about AI (Hwang et al. 2020; Muller
et al. 2022), and the collaborative relation-
ship between humans and computers has
been addressed from the outset (O’Shea
& Self, 1986).
In terms of learning machines,
there are two types. Teaching machines,
such as SAKI (self-adaptive keyboard in-
structor) or Eucrates, designed by Pask,
act as tutors or teaching assistants to stu-
dents (Husbands et al. 2008; Pask, 1961)
(Skinner, 1958). Conversely, URBAN5
(Negroponte, 1967) and Musicolour (Pask,
1971) illustrate computers as design part-
ners or collaborators.
In recent theoretical research,
conversational properties have been em-
phasised (Cheung et al. 2023b). Conver-
sation, as the main type of interaction,
focuses on how humans interact with com-
puters. A unique feature of conversation is
its learning process, where not only do hu-
mans learn about the tools iteratively dur-
ing the design process, but the tools also
have opportunities to understand the de-
signers’ intentions or preferences through
AI techniques like ne-tuning (Cheung &
Dall’Asta, 2023).
From the ecological perspective
(Sweeting, 2022), the interconnectedness
of technology, population, and hubris is
highlighted. Hubris refers to the notion that
113
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
humans are separate from the environ-
ment. Bateson argued that technology and
population are difcult to reverse, mak-
ing hubris the easiest choice (Bateson,
1972/2000).
In summary, the vision perspective
of the research explores the application of
HCI principles in designing an AI-integrat-
ed architectural design course, consider-
ing the historical context, collaborative re-
lationships, learning machines, precedent
studies, recent projects, and the ecologi-
cal aspects of technology and human in-
teractions.
From the perspective of architec-
tural education, it is important to empha-
sise reection within the architectural de-
sign process. In this context, the outputs
generated by AI art tools should be viewed
as part of a thinking process rather than
simply a tool for generating explicit de-
sign representations (Silva, 2019). Taking
another perspective from the technology,
programming can also be seen as an in-
tegral part of the design process rather
than an “additional” component to an ar-
chitectural design brief (Qattan & AUSTIN,
2016). Finally, applying HCI emphasises
the importance of “collaboration” (Cole-
man, 2023) within the ecology between
humans and GAI. Providing this full pic-
ture to the students potentially helps to ad-
dress the hubris problems.
5.2 Technology
Corresponding to the second re-
search question:
The technology perspective of the
research focuses on Generative Articial
Intelligence (GAI), with a particular empha-
sis on computer vision (CV) applications
and Natural Language Processing (NLP).
Regarding the CV GAI, its poten-
tial uses extend beyond image genera-
tion based on prompts. The aim is for the
usual design process, including sketches
and physical models, to be able to inter-
act with AI tools. Consequently, image-
to-image methods will be the main explo-
ration approach. Previous research has
demonstrated the importance of selecting
tools students can understand rather than
solely focusing on the fastest or highest-
quality image production (Cheung &
Dall’Asta, 2023).
Regarding NLP GAI, it acts as a
discussion partner and bridge between
other AI technologies. Pretrained Lan-
guage Models (LLMs) like llama devel-
oped by Meta and GPT developed by
OpenAI can be introduced for immediate
application. Students can converse and
discuss the studio design brief with the AI.
Additionally, LLMs provide an ideal plat-
form for students to discuss prompt engi-
neering and generate images at different
design stages using AI art generation tools
based on diffusion models.
Ethical issues, such as author-
ship, have been discussed in the school
policy, and ethical uses of AI should be
emphasised within the context of different
modules. Furthermore, other forms of GAI,
such as text-to-music or image-to-anima-
tion, can be introduced. On the one hand,
these technologies enhance architectural
representation by providing audiovisuals
and animations that complement static
images, opening up new possibilities for
design exploration and expression. On the
other hand, students are encouraged to
become accustomed to working with AI as
a team, preparing them for the future impli-
cations of human-AI collaboration.
From the teaching team’s per-
spective, tools for immediate application
and less sophisticated but advanced tools
are prepared. The latter serves as addi-
tional knowledge to provide students with
a glimpse into AI’s future implications and
mentally prepare them for its integration
into their architectural practice.
The potential use of Explainable AI
(XAI) (Khosravi et al. 2022) can enhance
user acceptance (Luckin et al. 2022). Al-
though Large Language Models (LLMs)
are not inherently designed as explain-
able AI, their interaction properties offer
insights to human users regarding the
thinking logic of models like GPT within the
educational context (Kasneci et al. 2023;
Villegas-Ch et al. 2020).
114
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
5.3 User Acceptance
The third point of focus in this re-
search pertains to user acceptance, spe-
cically addressing how students perceive
and accept the use of AI in architectural
design. The authors draw from their past
teaching experiences to shed light on this
aspect, supplemented by the perspec-
tives of the teaching team. Since 2020,
the authors have been actively involved
in teaching and exploring the integration
of articial intelligence (AI) in the architec-
tural design process. Feedback gathered
from previous teachings provides valuable
insights.
During the initial year, the teach-
ing team introduced Google Colab, a free
cloud-computing platform for students
to run AI programs online. However, stu-
dents’ unfamiliarity with the Jupyter Note-
book format posed a challenge initially.
With the assistance of a computer engi-
neering expert on the teaching team, stu-
dents were successfully taught how to use
the tool to generate images. Given the lim-
ited number of architectural precedents
in this area, extensive discussions arose
regarding the types of images that would
be “useful” for the AI to produce mean-
ingful results. As students became more
procient with the tools, discussions and
reections shifted towards their implica-
tions in the architectural design process.
For instance, students explored how de-
sign strategies inspired by AI-generated
images could be employed in their ar-
chitecture design studio projects. During
that semester, two tools, Deepdream and
StyleTransfer, were introduced. Although
Deepdream yielded less “promising” re-
sults, it provided insights into how com-
puters perceive the world differently from
human eyes.
In 2020 and 2021, the focus shifted
to using StyleTransfer and introducing us-
er-friendly applications, allowing students
to explore various tools independently.
With increased practical experience, over
Figure 2.Google Colab interface
Source: created by the author.
Figure 3. StyleTransfer experiments and inspired design strategies
Source: created by the author.
115
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
90 % of students expressed a positive at-
titude towards applying AI in their design
approaches, and nearly 90 % envisioned
its relevance in their future practices
(Dall’Asta, 2023). Some students who par-
ticipated in these courses or workshops
have even been employed due to their
unique prociency in employing AI tools in
practice and research environments.
Our teaching team successfully
conducted a four-day online workshop
titled “Hacking Machine Learning Style-
Transfer” on the DigitalFutures platform.
This workshop attracted the participation
of twenty students from around the world
who had little to no coding knowledge. Our
primary objective for this workshop was to
enhance user acceptance by providing
students with a clear and familiar workow.
To achieve this, we structured the
four-dayday workshop to allow immediate
use of a web application for StyleTransfer
inspiration. The students could create 2D
collages by manually editing selected AI-
generated images and 3D modelling their
designed collages. The nal step involved
presenting their work (as shown in Fig. 4
and 5). By providing a well-dened objec-
tive and a clear path to achieve their de-
sired AI-generated results, the students
exhibited a notable increase in condence
throughout the design process.
However, it is worth noting that
although the workshop successfully in-
creased user acceptance, we observed a
limited utilisation of AI capabilities. In this
particular case, AI served merely as an in-
spiration generation tool due to the domi-
nant involvement of manual image editing
and 3D modelling. Nevertheless, user ac-
ceptance has dramatically increased.
Two main ndings emerged from
the previous study. Firstly, the importance
of conversations between students and AI
in the architectural design process was
emphasised. Secondly, the probabilistic
nature of AI allows for unlimited design
suggestions, indirectly encouraging stu-
dents to reect on their choices rather than
relying on absolute design decisions.
Figure 4. User interface of StyleTransfer on a web application, experiment by Lakita
Source: Ralph Spencer Steenbilk,2021.
116
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Since 2022, new tools have
emerged due to advancements in AI,
particularly in Diffusion Models, including
large pre-trained language models like
GPT. Consequently, the teaching team re-
assessed existing tools and reected on
how the latest AI developments could be
taught and applied in architectural design.
This involved testing different AI art gen-
eration tools and exploring applications of
large language models while delving into
technical and theoretical aspects. As a
result, a collaborative and intuitive frame-
work for the combined application of AI art
generation tools in the architectural design
process was being researched in 2022,
and proposed in early 2023 (Cheung &
Dall’Asta, 2023). Through reection, three
potential areas for improvement and fur-
ther exploration were identied.
Firstly, instead of using plain imag-
es, students could attempt to apply lters
such as depth detection to their physical
model photos. Secondly, a streamlined ap-
plication for students, including a 3D intui-
tive workow like 3D modelling in AI appli-
cations, could enhance usability. Thirdly,
when feasible, real-world scenarios, par-
ticularly within the academic environment
of architectural design studios, should be
incorporated.
During the summer of 2023, the
rst author participated in a Summer Un-
dergraduate Research Fellowship (SURF)
project, designing a streamlined workow
using Google Colab as a platform to pro-
vide students with a free tool. The project
spanned eight weeks and aimed to pack-
age three AI models (BLIP for Visual-Ques-
tion-Answering, GPT as a large language
model, and SD as an AI art generation tool)
into a single Colab notebook. However, it
became apparent in the second week that
students preferred to nd separate web
applications for BLIP and GPT and a local
application for Stable Diffusion. Some even
explored alternative tools like Midjourney
and additional applications of StableDiffu-
sion, including customised AI models. This
reection indicated that streamlined appli-
cations with script-based user interfaces,
such as Google Colab, had lower prefer-
ence among students.
Considering the importance of AI
ethics, especially fairness, it is crucial to
address potential issues. Students from
economically challenged backgrounds
may need more access to AI applica-
tion opportunities, highlighting that user
acceptance is not solely a cognitive
matter but also inuenced by nancial
considerations. Therefore, prioritising
Figure 5. Edited collages of AI images inspired 3D modelling, by Xinyi Zhang
Source: (Spencer, 2021).
117
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Figure 6. Three simulated scenarios in the architectural design process, by Lok Hang Cheung
Source: created by the author.
Figure 7. A streamlined AI toolkit designed for the SURF project using Google Colab
Source: created by the author.
Figure 8. Parametric optimisation tool and AI tools in the same workow, by Chuwen Zhong and Yian Shi, supervised by Likai
Wang, Leyuan Jiang and Lok Hang Cheung
Source: reproduced from author, with permission
118
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
open-source or free applications over
pursuing the most advanced or newest
tools, which often require costly compu-
tational power, is essential.
Several strategies can be em-
ployed to increase user acceptance.
Firstly, providing user-friendly interfaces
for immediate application during in-class
exercises is important. Secondly, incor-
porating precedent studies, in-class dis-
cussions, and post-exercise reections
can enhance students’ knowledge and
theoretical understanding of AI develop-
ment and tools. Lastly, setting clear stu-
dent goals, establishing explicit expected
outcomes, and providing short-term, im-
mediate feedback are crucial for ensuring
students can effectively utilise the tools.
Additionally, encouraging reections from
students after each in-class exercise fos-
ters long-term application in the design
studio and future design projects.
6. RESEARCH – SYNTHESIS IN
PRACTICE
After reecting on the framework
from the perspectives of vision, technol-
ogy, and user acceptance, we designed
modules for the architectural design
course in the academic year 2023-24.
6.1 Introduction
The architectural design module
titled ARC 411, known as Practice Based
Enquiry and Architectural Representa-
tion, is designed for second-year Master
of Architectural Design students in their -
nal year of the RIBA Part 2 course at Xi’an
Jiaotong-Liverpool University. The module
consists of one weekly class and adopts
a highly practice-based approach, incor-
porating in-class exercises, tutorials, and
discussions. For 14 weeks, students must
submit and present their work every four
to ve weeks, each being an independent
coursework assignment.
The teaching team responsible for
the module comprises two individuals. The
rst author serves as the teaching assis-
tant. In contrast, the second author fulls
the role of the module leader for ARC 411
and the programme director for the Master
of Architectural Design course.
With a focus on advanced prac-
tice-based methodologies, this module
introduces students to the latest articial
intelligence (AI) technologies and their
application in critical, creative problem-
solving and communication within archi-
tecture. The course encourages students
to explore Western and Eastern art prac-
tices, enabling them to engage with and
perceive such engagement as a form of
critical inquiry into prevailing architectural
presentation and representation methods.
Through the re-presentation of ar-
chitectural projects and the utilisation of
various media, such as drawings, models,
video, sculpture, interactive digital media,
and installation art, students are exposed
to novel approaches for identifying ques-
tions, addressing them, and effectively
communicating their ideas to audiences
that may not possess specialised knowl-
edge in reading architectural plans and
models. This approach acknowledges and
accommodates differing interpretations of
architecture, broadening students’ hori-
zons and fostering an expanded under-
standing of the discipline.
6.2 Coursework 1 – AI as a Design Partner
The rst coursework is titled “AI as
a Design Partner, Generative Articial In-
telligences as a Design Team.” Within the
eld of architecture, Generative Articial
Intelligence (AI) has garnered signicant
attention, particularly with the emergence
of techniques like Deepdream and Style
Transfer, which have been explored since
2015. This exploration has paved the way
for advancing more sophisticated AI tech-
niques in computer vision (CV), such as
diffusion models. These advancements
have led to the rapid development of
various AI art generation tools capable of
producing high-quality images, including
notable examples like Midjourney and Sta-
bleDiffusion. However, it is crucial to ac-
knowledge that many current applications
tend to view these AI art generation tools
merely as random “inspiration generators”
or, in the words of Nicholas Negroponte,
“a fast draftsman who doesn’t eat.” (Ne-
groponte, 1976).
To fully harness the potential of
AI in the architectural design process, it
119
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
is crucial to view AI as a design partner
rather than a mere tool. This coursework
aims to explore practical applications of
different AI technologies at various de-
sign stages, from conceptual exploration
to architectural representation. Genera-
tive AI techniques, particularly in Natural
Language Processing (NLP), have made
signicant advancements. Large Lan-
guage Models (LLMs) like ChatGPT will be
utilised in this assignment to facilitate dis-
cussions, explorations, and interpretations
of the design process.
The exercise has three primary ob-
jectives:
Firstly, it provides an overview of
AI applications, emphasising the concept
of “AI as a design partner.” This ensures
that students clearly understand different
AI categories and encourages them to en-
vision the untapped potential of AI in archi-
tecture. Secondly, the exercise introduces
the latest generative AI techniques, such
as StableDiffusion, PlaygroundAI, and
GPT, among others, and explores their
practical applications in real-world design
scenarios. Emerging techniques like BLIP,
a Visual-Question-Answering (VQA) AI,
will also be introduced and tested. Lastly,
the exercise prompts students to reect
on how designers can collaborate with AI
as a design team in different stages and
scenarios of the design process, prevent-
Figure 9. A Conversation with a Machine by Erik Ulberg
Source: Erik Ulberg, https://www.erikulberg.com/#/conversationwithamachine
120
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
ing them from solely relying on supervisor
comments (Sweeting, 2014). Particular
emphasis will be placed on reecting on
the experiences of working with different
AI technologies, enabling students to de-
velop their approaches to collaborating
with various AIs in their current and future
architectural projects.
Although the module has only been
underway for one week, it has already been
observed that students feel comfortable us-
ing StableDiffusion through PlaygroundAI,
a user-friendly and free web application.
They have engaged in image-generation
experiments by modifying prompts and
employing text-to-image and image-to-
image techniques. The intuitive interface
of StableDiffusion has encouraged stu-
dents to ask design-related and theoreti-
cal questions regarding AI creativity and
its implications in design studios and the-
sis work. Furthermore, students with prior
experience using other AI art applications,
such as Midjourney, have taken the initia-
tive to compare the outcomes produced by
StableDiffusion. A few students have also
independently explored StableDiffusion on
their computers before starting the module.
In the initial week of the course, the
focus was on introducing the background
of AI, emphasising understanding its ap-
plication in the architectural design con-
text rather than approaching it solely from
a computational or technical perspective.
Subsequently, case studies will be pre-
sented weekly to broaden students’ knowl-
edge and enhance their acceptance of AI
by providing multiple perspectives. For in-
stance, the case study of URBAN5 by the
Architecture Machine Group has been in-
troduced. Relevant projects incorporating
recent advancements in AI technology,
such as “A Conversation with a Machine”
by Erik Ulberg (2019), have also been in-
cluded to enable students to observe the
continuous development in this eld.
Throughout the course, each les-
son will include two to three in-class ex-
ercises to engage students in practical
applications of AI techniques. Exercise
1 introduces the Large Language Model
(GPT), available on poe.com. Students
are tasked with exploring how they can
brainstorm ideas or engage in discussions
with GPT, including generating prompts
for AI art generation tools. Following this
activity, students are encouraged to re-
ect on their challenges and potential op-
portunities for applying these techniques
in the design process. In the exercises,
students must prepare four sets of im-
ages, with at least one example for each
set. These images serve as a means for
students to explore their creative applica-
tions. Each set focuses on a specic topic,
Figure 10. An example template slide for the students
Source: created by author
121
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
such as site exploration or a design idea.
In the reection section, students present
their original intentions behind generat-
ing the images, discuss the challenges
they faced, and highlight any opportuni-
ties they discovered. Students are guided
to practice various techniques, including
text-to-image and image-to-image skills,
and gain an understanding of parameters
such as prompt strength, image selection,
and style manipulation.
Exercise 2 introduces the AI art
generation tool, StableDiffusion, to stu-
dents with little to no prior experience with
AI. PlaygroundAI, a user-friendly applica-
tion platform, is utilised to facilitate their en-
gagement. After registering, students can
immediately generate free images using
StableDiffusion. While some AI techniques
and tools may be introduced as extensions
of knowledge without immediate practical
use, they contribute to the student’s under-
standing of the broader AI landscape. For
instance, the Visual-Question-Answering
AI model, BLIP2, is introduced in the rst
week, showcasing the potential merging
of Computer Vision (CV) and Natural Lan-
guage Processing (NLP) in seamless ap-
plications. Although BLIP2 demonstrates
the feasibility of combining CV and NLP
through image-captioning and question-
answering, its limitations in handling com-
plex inquiries highlight the need for further
development in real-world scenarios.
During the second and third weeks,
the coursework will explore StableDiffu-
sion, which is considered the assignment’s
core tool. StableDiffusion is chosen due to
its open-source nature, availability, and
extensive range of applications. While one
student uses a Mac system, the remaining
students utilise Windows, and all have suc-
cessfully installed and run StableDiffusion.
The widespread exploration of StableDiffu-
sion in the global AI community has led to
the development of installation packages
for both Windows and Mac users. During
this period, the in-class exercises will re-
volve around architectural design-based
techniques. Students will learn to create
design inspiration within a site context
using existing site plans or photos. They
will also explore ways to achieve similar
effects to StyleTransfer using StableDiffu-
sion and generate design ideas based on
their sketches or model images.
Parts of two presented projects closely
related to the design studio project
are displayed below. In the rst case,
the student expressed a keen interest
in painting art styles but needed more
certainty regarding their application
within the design studio project. He
rst generated inspirational images by
Figure 11. An example of applying StableDiffusion in generating controlled variations as a design exploration method.
Source: created by author.
122
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Figure 12. Student iterative AI exploration in the design project, by Huanyue Gao
Source: reproduced from author, with permission
Figure 13. Student AI exploration in different design stages, by Yibo Zhao
Source: reproduced from author, with permission
123
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
inputting different art style names and
Barcelona, the project site. He then
selected two desired results and integrated
them with his site impression drawing.
This iterative approach continued as he
incorporated the generated images into
his physical conceptual model, as shown
in Fig.12. The second student showed
his passion for applying computational
skills, including AI, in the very early stage.
Upon discovering StableDiffusion and
grasping the fundamentals of ControlNet,
the student embarked on extensive self-
experimentation throughout various
stages of the design studio project. Fig.13
showcased some extracted attempts at
different stages.
In the rst case, it appears that the
student utilised AI to generate inspiration,
similar to previous cases. However, what
sets this case apart is the student’s abil-
ity to exert control over the selected tools,
such as GPT and StableDiffusion. This con-
trol enabled the student to transition from
inspiration-only to a collaborative design
approach. The second student, who was
more proactive in using AI tools and trying
to apply them in every design step, illus-
trates the possibility of smooth workow in-
tegration once tools become familiar.
6.2 Coursework 2 – Translations,
Synthesizing Reality
Visual language is only one of the
mediums expressing architectural design.
Therefore, in the second coursework, stu-
dents will explore different audiovisual me-
diums or interactive techniques to design
a spatial performance.
The rst week of teaching will focus
on expanding students’ understanding of
various AI tools, similar to the beginning of
Coursework 1. Using web applications that
can be implemented immediately is crucial
to kick off the coursework and increase
user acceptance among the students.
Three tools will be introduced: image-to-
video generation through the web appli-
cation developed by RunwayML, text-to-
music using the web application Mubert,
and text-to-speech narration using the free
platform Topmediai. These tools offer a
range of AI methods for video generation,
music creation, and speech synthesis,
providing students with a comprehensive
overview of available techniques.
Introducing these tools aims to as-
sist students in exploring audio-visual and
3D mapping applications. The possibility
of deriving ideas or prompts from discus-
sions with the GPT model will be explored,
integrating it with the concepts covered in
the rst coursework.
Figure 14. Framework of coursework two
Source: created by author.
In the second part of the course,
the focus is on establishing connections
with existing architectural design tools,
such as Grasshopper. This is particularly
important as students would have made
progress in their design studios after ap-
proximately one month of the course. In-
stead of relying solely on site context or
design intentions expressed in text, stu-
dents now have additional information and
data from their design projects, such as
building parameters. Building on previ-
ous teaching and learning experiences,
plugins and electronic tools that directly
translate architectural model parameters
into audio-visual representations will be in-
troduced. This allows for a more intercon-
nected conversation between AI and the
students, as they can input design ideas
verbally through prompts and utilise famil-
iar parametric tools.
Similar to the rst part of the
course, projects, case studies, and theo-
retical discussions related to multimodal
architectural representations and inter-
active architecture will be explored. Ad-
ditionally, a brief introduction to ongoing
research on exploring AI-integrated archi-
tecture (Cheung et al. 2023a) will be pro-
vided. It was not focused on AI as part of
the computational tool but integrated into
the built environment, such as a design
studio. The course will introduce audio-
visual techniques within standard archi-
tectural software such as Rhinoceros and
124
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Grasshopper and explore AI tools such as
text-to-music and image-to-animation. By
integrating AI and parametric tools within
the design framework, the course extends
the concept of “AI as a design partner”
from the rst coursework, offering a holistic
approach to architectural design.
Two ongoing projects were pro-
gressing in noteworthy and surprising
ways. In the rst case, the student ob-
served from Coursework One that AI-gen-
erated images in pixel format couldn’t be
directly used in the design process, lack-
ing the necessary vectorised information
such as points and lines. Consequently,
she integrated a large language model
into the 3D modelling software Rhino and
Grasshopper. Through experimentation,
she developed codes with the AI capable
of generating geometries that could be
effectively utilised in the design process,
as depicted in Fig 16. The current video
production adopts a video-blogging style,
capturing the conversation between her
and the AI partner within the Grasshop-
per environment.
The second student expressed res-
ervations about employing AI as a design
partner due to its unpredictable nature, ap-
proaching the matter from a philosophical
standpoint. In Coursework Two, she took
the opportunity to compare how different
Figure 16. GPT integrated into Grasshopper to generate vectorised geometries instead of pixels, by Zhiyan Peng
Source: reproduced from author, with permission
Figure 15. Interactive performance for coursework two from the same module in the previous year
Source: created by author.
125
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
entities (human, machine, and AI) perceive
the world in distinct ways. Starting with the
observation of a moss-covered wooden
branch, she proceeded to perform a 3D
scan. Then, she provided the resulting 3D
model image to the AI, allowing it to reimag-
ine the object, as illustrated in Fig. 17.
Both students demonstrated origi-
nal and unexpected approaches to their
coursework. The rst student adopted a
pragmatic parametric design workow
and successfully integrated AI. On the
other hand, the second student explored
the philosophical aspects by investigat-
ing how humans, traditional machines
(such as cameras and 3D scanners), and
AI perceive the world differently. They
delved into the iterative conversation
among these entities, presenting a novel
approach to design.
These “out-of-box” experiences
deviate from previous research and teach-
ing experience, highlighting the value of
introducing user-friendly AI tools. Students
can immediately engage in experimenta-
tion by eliminating the need for coding
knowledge, providing greater freedom for
imagination and exploration.
6.4 Coursework 3 – Final Exhibition, Video
performance /video reportage and joint
exhibition with Design Studio
The nal exhibition in ARC411 will
showcase all the coursework developed
throughout the semester. In addition to
the coursework, students will apply AI and
parametric tools to assist in designing and
developing an exhibition centred around
the project from ARC413. This exhibition
will include a 3D video mapping perfor-
mance that tells the story of the ARC413
project’s process. By utilising computa-
tional tools, students can work alongside
them as a design team, from the concep-
tual stage to the nal presentation.
The 3D video mapping perfor-
mance is a pedagogical project that builds
upon previous experiences with 3D map-
ping. This approach to architectural design
connects content to representation directly
and effectively, incorporating elements
such as image, video, and sound editing.
Final presentations of this nature encour-
age student participation and engagement
with the subject matter while also fostering
critical thinking and creativity.
In previous attempts, students
shared the same site model, allowing
each student to prepare their model to
be placed on the site. Some students ex-
plored creative approaches, such as pre-
senting the design process or simulating
Figure 18. A 3D mapping performative presentation for
coursework three from the same module in the previous year
Source: created by author.
Figure 17. Experimentation of how we, camera, computer
and AI see differently, by Yuheng Liu.
Source: reproduced from author, with permission
126
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
a conversation between the users of the
buildings. However, there needed to be
more connections between the models
and the students, with the students func-
tioning more as MCs.
Other attempts involved using
larger models, which allowed students
to interact and perform, creating a more
immersive experience. However, the so-
phistication level of the models could
have been improved. Conversely, a more
controlled environment was created us-
ing a smaller 60x60cm model, enabling
greater detail and more controlled map-
ping. However, the immersive perception
of the performance was diminished. Find-
ing a balance between delicate models
and immersive interactivity in the form of
an interactive performative presentation
requires further planning that aligns with
the students’ design direction.
7. CONCLUSION
7.1 Contributions
Table 1 summarises the aforemen-
tioned teaching and research experience
based on the triangulated framework of
technology, vision and user acceptance.
In addition, each experience’s ndings are
concluded so each iteration’s inuence
can be observed.
This paper presents a human-
computer interaction (HCI) framework to
bridge the gap between AI technology
and the conventional human-centric ap-
proach in architectural design studios.
The framework contributes in several
ways. First, it highlights the importance of
controlled variations throughout the de-
sign process. Second, it seeks to extend
the potential collaboration with AI beyond
inspiration, encompassing the design
process and presentation stages. The
framework goes beyond mere attempts
and actively strives to integrate AI into the
architectural design workow.
Several key ndings can be high-
lighted based on the observations pre-
sented in Table 1.
1. Regarding technology, it is advisa-
ble to shift the focus away from AI
art generation tools and instead pro-
mote the integration of multi-modal
AIs and existing parametric tools.
This approach would result in a more
comprehensive toolkit for architectu-
ral design.
2. Regarding vision, it is important to
prioritise the role of AIs as design
partners rather than making assump-
tions about their specic usage at
different stages of the architectural
design process.
3. About user acceptance, it is crucial
to provide user-friendly tools and
theoretical foundations. By offering
tools that can be immediately utili-
sed and supporting them with rele-
vant theoretical knowledge, students
are encouraged to explore beyond
the design process and expand their
research and design boundaries.
7.2 Challenges
There are inherent challenges in
the AI and architecture domains that must
be addressed. AI is a rapidly developing
discipline, and even with a conversational
framework in place, there is an inevitable
learning curve. Furthermore, architectural
design processes are highly subjective
and iterative, often characterised as “wick-
ed” problems (Rittel & Webber, 1973).
These problems do not have one-size-ts-
all solutions, necessitating frequent reec-
tions and adaptations.
Figure 19. Timeline of AIEd in architectural design discipline, by the authors
Source: created by author.
127
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
7.3 Future Opportunities
First, there is an opportunity to ex-
tend the framework to other aspects of AI
in Education (AIEd), including system-fo-
cused and teacher-focused approaches.
The framework can encompass a broader
range of educational contexts by exploring
these areas.
Second, it is important to incor-
porate feedback from the ARC411 and
ARC413 modules to reect on and im-
prove the framework continuously. This it-
erative process will contribute to its rene-
ment and effectiveness.
Third, the potential of AI extends
beyond being a design partner in the form
of digital applications or IoT devices. In-
tegrating AI directly into the design studio
is possible, creating an AI-integrated built
environment. This opens up new avenues
for exploration, where human-computer
Table 1. Summary of AIEd in architectural design by the authors
Date
Design
activities
Number
of
students
Technology
Vision
User
Acceptance
Findings
AI tools Platforms
2020
XJTLU Master
Year 2 Module
ARC411
12
Deepdream,
Style-
Transfer
Google Colab
Explore
what AI art
generation
tools can
do in the
architecture
discipline
Involve a
tutor with a
computing
background
to teach AI
techniques
Depends
heavily on the
designer’s
imagination
2021
“Hacking
Machine
Learning Style
Transfer”,
Digital
FUTURES
online
workshop
20
Style-
Transfer
Web application
(Web application
(DeepArt.io))
Involve AI art
generation
tools in
the design
process
Integrate AI
into a common
design
workow
Still depends
heavily on
imagination
and manual
inputs
Underused AI
2022 Self-research N/A
Stable-
Difsion,
DALLE-2
Web application
(PlaygroundAI)
Google Colab
Web application
Framework for
applying AI
art generation
tools in
the design
process
Simulated
scenarios for
designers
with different
levels of
computational
knowledge
Proposed
an intuitive,
collaborative,
combined
application
of AI art
generation
tools
2023
XJTLU SURF
(Summer
Undergraduate
Research
Fellowship)
6
BLIP2, GPT,
Stable-
Diffusion
Google Colab
Explore a
workow
applying AI
art generation
tools with a
parametric
optimisation
tool
Compiled
different AI
tools into
one single
streamlined
web
application
Students
prefer user-
friendly
UI over a
streamlined
application
that requires
coding
knowledge
2023
(on-
going)
XJTLU Master
Year 2 Module
ARC411
12
GPT, Stable-
Diffusion
Web application
Local computers
(WebUI)
HCI-based
AIEd
framework for
multi-modal AI
tools as design
partners and
representation
Introduced
tools that
required
minimum
computational
knowledge
More
seamless
integration into
the process.
Multimodal
AI and
integration into
parametric
tools allowed
a variety of
applications
(Still ongoing)
128
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
interaction focuses on the human-ma-
chine conversation (HMC) (Cheung et al.
2023b), emphasising the physicality of
architecture and the design environment.
The learning process for AI during its em-
ployment within the design studio is also
a valuable area to be explored, extending
the notion of reection not only for students
or designers but also for AI systems.
ACKNOWLEDGEMENTS
This research is supported by XJT-
LU Postgraduate Research Scholarship
(PGRSB211206) offered by Xi’an Jiaotong-
Liverpool University.
8. REFERENCES
Baidoo-Anu, D., & Owusu Ansah, L. (2023).
Education in the Era of Generative Articial
Intelligence (AI): Understanding the Po-
tential Benets of ChatGPT in Promoting
Teaching and Learning. SSRN Scholarly
Paper 4337484. https://doi.org/10.2139/
ssrn.4337484
Bakir, R., & Alsaadani, S. (2022). A mixed
methods study of architectural education
during the initial Covid-19 lockdown: Stu-
dent experiences in design studio and
technology courses. Open House Inter-
national, 47(2), 338–360. https://doi.org/
10.1108/OHI-09-2021-0206
Bateson, G. (2000). Step to an ecology of
mind. (Original work published 1972). Uni-
versity of Chicago Press.
Celik, I. (2023). Towards Intelligent-TPACK:
An empirical study on teachers’ profes-
sional knowledge to ethically integrate
articial intelligence (AI)-based tools into
education. Computers in Human Behavior,
138, 107468. https://doi.org/10.1016/j.
chb.2022.107468
Chatterjee, S., & Bhattacharjee, K. K.
(2020). Adoption of articial intelligence
in higher education: A quantitative analy-
sis using structural equation modelling.
Education and Information Technologies,
25(5), 3443–3463. https://doi.org/10.1007/
s10639-020-10159-7
Chen, L., Chen, P., & Lin, Z. (2020). Arti-
cial Intelligence in Education: A Review.
IEEE Access, 8, 75264–75278. https://doi.
org/10.1109/ACCESS.2020.2988510
Chen, X., Zou, D., Xie, H., Cheng, G., & Liu,
C. (2022). Two Decades of Articial Intelli-
gence in Education: Contributors, Collabo-
rations, Research Topics, Challenges, and
Future Directions. Educational Technology
& Society, 25(1), 28–47.
Cheung, L. H., & Dall’Asta, J. C. (2023).
Exploring a Collaborative and Intuitive Fra-
mework for Combined Application of AI Art
Generation Tools in Architectural Design
Process. In B. Muller (Ed.), Proceedings of
the AISB Convention 2023 (pp. 122-130).
ISBN 978-1-908187-85-7.
Cheung, L. H., Dall’Asta, J. C. & Di Mar-
co, G. (2023a). Exploring Large Language
Model as a Design Partner through Verbal
and Non-verbal Conversation in Architec-
tural Design Process. SiGraDI 2023. Uru-
guay 29 November – 1 December.
Cheung, L. H., Dall’Asta, J. C., Di Marco,
G. & Agkathidis, A. (2023b). Architectural
Design under Pandemic of Today’s AI:
From Human-computer Interaction (HCI)
to Human-machine Conversation (HMC).
xArch Symposium 2023 Proceedings.
Suzhou (China) 11-12 November.
Coleman, K. (2023). Generative AI and
education ecologies. Pacic Journal of Te-
chnology Enhanced Learning, 5, 19–20.
https://doi.org/10.24135/pjtel.v5i1.175
Dall’Asta, J. C. (2023). AI Machine Lear-
ning in Creative Architectural Design Pro-
cesses. xArch Symposium 2023 Procee-
dings. Suzhou (China) 11-12 November.
Glanville, R. (1992). CAD Abusing Compu-
ting. CAAD Instruction: The New Teaching
of an Architect? [eCAADe Conference Pro-
ceedings] Barcelona (Spain) 12-14 Novem-
ber, pp. 213-224. http://papers.cumincad.
org/cgi-bin/works/paper/cba7
Goodbun, J., & Sweeting, B. (2021). The
dialogical, the ecological and beyond.
Footprint, 15(1), Article 1. https://doi.org/
10.7480/footprint.15.1.5668
Haderer, B., & Ciolacu, M. (2022). Edu-
cation 4.0: Articial Intelligence Assisted
129
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Task- and Time Planning System. Proce-
dia Computer Science, 200, 1328–1337.
https://doi.org/10.1016/j.procs.2022.
01.334
Hettithanthri, U., & Hansen, P. (2022). De-
sign studio practice in the context of ar-
chitectural education: A narrative literature
review. International Journal of Technolo-
gy and Design Education, 32(4), 2343–
2364. https://doi.org/10.1007/s10798-021-
09694-2
Holmes, W., & Tuomi, I. (2022). State of
the art and practice in AI in education. Eu-
ropean Journal of Education, 57(4), 542–
570. https://doi.org/10.1111/ejed.12533
Holzer, D. (2019). Teaching computational
design and BIM in the age of (semi)ipped
classrooms. Intelligent and Informed - Pro-
ceedings of the 24th International Confe-
rence on Computer-Aided Architectural
Design Research in Asia, CAADRIA 2019, 2.
https://findanexpert.unimelb.edu.au/
scholarlywork/1403544-teaching-
computational-design-and-bim-in-the-
age-of-(semi)ipped-classrooms
Huang, J., Saleh, S., & Liu, Y. (2021). A
Review on Articial Intelligence in Educa-
tion. Academic Journal of Interdisciplinary
Studies, 10, 206. https://doi.org/10.36941/
ajis-2021-0077
Husbands, P., Holland, O., & Wheeler, M.
(2008). Gordon Pask and His Maverick
Machines. In The Mechanical Mind in His-
tory (pp. 185–211). MIT Press. https://iee-
explore.ieee.org/document/6287755
Hwang, G.-J., Xie, H., Wah, B. W., & Gaše-
vić, D. (2020). Vision, challenges, roles and
research issues of Articial Intelligence in
Education. Computers and Education: Ar-
ticial Intelligence, 1, 100001. https://doi.
org/10.1016/j.caeai.2020.100001
Kasneci, E., Sessler, K., Küchemann, S.,
Bannert, M., Dementieva, D., Fischer, F.,
Gasser, U., Groh, G., Günnemann, S., Hü-
llermeier, E., Krusche, S., Kutyniok, G.,
Michaeli, T., Nerdel, C., Pfeffer, J., Po-
quet, O., Sailer, M., Schmidt, A., Seidel,
T., Kasneci, G. (2023). ChatGPT for good?
On opportunities and challenges of large
language models for education. Learning
and Individual Differences, 103, 102274.
https://doi.org/10.1016/j.lindif.2023.
102274
Khosravi, H., Shum, S. B., Chen, G., Cona-
ti, C., Tsai, Y.-S., Kay, J., Knight, S., Marti-
nez-Maldonado, R., Sadiq, S., & Gašević,
D. (2022). Explainable Articial Intelligen-
ce in Education. Computers and Edu-
cation: Articial Intelligence, 3, 100074.
https://doi.org/10.1016/j.caeai.2022.
100074
Luckin, R., Cukurova, M., Kent, C., & du
Boulay, B. (2022). Empowering educators
to be AI-ready. Computers and Education:
Articial Intelligence, 3, 100076. https://
doi.org/10.1016/j.caeai.2022.100076
Markauskaite, L., Marrone, R., Poquet,
O., Knight, S., Martinez-Maldonado, R.,
Howard, S., Tondeur, J., De Laat, M.,
Buckingham Shum, S., Gašević, D., &
Siemens, G. (2022). Rethinking the en-
twinement between articial intelligence
and human learning: What capabilities do
learners need for a world with AI? Com-
puters and Education: Articial Intelligen-
ce, 3, 100056. https://doi.org/10.1016/j.
caeai.2022.100056
Muller, M., Chilton, L. B., Kantosalo, A.,
Martin, C. P., & Walsh, G. (2022). GenAI-
CHI: Generative AI and HCI. Extended
Abstracts. 2022 CHI Conference on Hu-
man Factors in Computing Systems, 1–7.
https://doi.org/10.1145/3491101.3503719
Nabizadeh Rafsanjani, H., & Nabizadeh,
A. H. (2023). Towards human-centered
articial intelligence (AI) in the archi-
tecture, engineering, and construction
(AEC) industry. Computers in Human Be-
havior Reports, 11, 100319. https://doi.
org/10.1016/j.chbr.2023.100319
Namatherdhala, B., Mazher, N., & Sriram,
G. (2022). A Comprehensive Overview of
Articial Intelligence Tends in Education.
04, 2582–5208.
Negroponte, N. (1976). Soft Architec-
ture Machines. https://doi.org/10.7551/
mitpress/6317.001.0001
Nemorin, S., Vlachidis, A., Ayerakwa, H.
M., & Andriotis, P. (2022). AI hyped? A
130
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
horizon scan of discourse on articial inte-
lligence in education (AIED) and develo-
pment. Learning, Media and Technology.
https://doi.org/10.1080/17439884.2022.
2095568
O’Shea, T., & Self, J. (1986). Learning and
Teaching with Computers: The Articial In-
telligence Revolution. Prentice Hall Profes-
sional Technical Reference.
Pangaro, P. (2020) Cybernetics, AI, and
Ethical Conversations. AiTech Agora Se-
ries, TU Delft, viewed 10 August 2023,
https://www.youtube.com/watch?v=VvJ
pkqKlv9Q&ab_channel=AiTech-TUDelft
Pask, G. (1961). An Approach to Cyber-
netics. https://www.semanticscholar.org/
paper/An-Approach-To-Cybernetics-
Pask/327d42a1ca1ccab0c50f1cf47d05
a8fedf9560e4
Pask, G. (1971) “A Comment, A Case His-
tory, and a Plan”, Cybernetics, Art and
Ideas. Reichardt, J., (Ed.) Studio Vista,
London, pp.76-99
Pask, G. (1980). The Limits of Togetherness.
IFIP Congress. https://www.semantics
cholar.org/paper/The-Limits-of-
Togetherness-Pask/967ab2134ffe4
acda1ed84203ec5300deb5d5149
Perez, C. (2002). Technological Revolutions
and Financial Capital: The Dynamics of Bu-
bbles and Golden Ages. Edward Elgar.
Pham, S. T. H., & Sampson, P. M. (2022).
The development of articial intelligence in
education: A review in context. Journal of
Computer Assisted Learning, 38(5), 1408–
1421. https://doi.org/10.1111/jcal.12687
Price, C. (1979). Technology is the answer
but what was the question? [Sound Re-
cording]. Retrieved from PidgeonDigital:
https://www.pidgeondigital.com/talks/
technology-is-the-answer-but-what-
was-the-question-/
Qattan, W., & Austin, M. (2016). I’m a Vi-
sual Thinker. Rethinking algorithmic edu-
cation for architectural design. 829–838.
Qiu, Y. (2020). Education Informationi-
zation: Education Rain Run the World,
Science and Technology Spring Breeze
Promote Equality. Proceedings of the
2020 6th International Conference on Edu-
cation and Training Technologies, 40–43.
https://doi.org/10.1145/3399971.3399978
Rittel, H. W. J., & Webber, M. M. (1973).
Dilemmas in a general theory of plan-
ning. Policy Sciences, 4(2), 155–169.
https://doi.org/10.1007/BF01405730
Silva, D. (2019). Digital Fabrication—From
Tool to a Way of Thinking—Toward rede-
ning architectural design methodology.
463–470. https://doi.org/10.52842/conf.
caadria.2019.2.463
Skinner, B. F. (1958). Teaching machines.
Science, 128, 969–977. https://doi.org/
10.1126/science.128.3330.969
Smolansky, A., Cram, A., Raduescu, C.,
Zeivots, S., Huber, E., & Kizilcec, R. F.
(2023). Educator and Student Perspec-
tives on the Impact of Generative AI on
Assessments in Higher Education. Proce-
edings of the Tenth ACM Conference on
Learning @ Scale, 378–382. https://doi.org/
10.1145/3573051.3596191
Sun, C., Zhou, Y., & Han, Y. (2022). Au-
tomatic generation of architecture facade
for historical urban renovation using ge-
nerative adversarial network. Building and
Environment, 212, 108781. https://doi.org/
10.1016/j.buildenv.2022.108781
Sweeting, B. (2014). Not All Conversations
Are Conversational: A Reection on the
Constructivist Aspects of Design Studio
Education. Constructivist Foundations, 9,
405–406.
Sweeting, B. (2022). Architectural roots of
ecological crisis. Proceedings of Relating
Systems Thinking and Design (RSD11)
Symposium.
Tack, A., & Piech, C. (2022). The AI Tea-
cher Test: Measuring the Pedagogical
Ability of Blender and GPT-3 in Educatio-
nal Dialogues (arXiv:2205.07540). arXiv.
https://doi.org/10.48550/arXiv.2205.07540
131
EÍDOS 23
2023
CHEUNG, L., DALL’ASTA, J. - Human-computer Interaction (HCI) Approach to Articial Intelligence in Education (AIEd) in Architectural Design. pp. 109-131 ISSN:1390-5007
Ulberg, E. (2019) A Conversation with a
Machine. Erik Ulberg, viewed 17 Septem-
ber 2023, https://www.erikulberg.com/#/
conversationwithamachine
Villegas-Ch, W., Arias-Navarrete, A., & Pa-
lacios-Pacheco, X. (2020). Proposal of an
Architecture for the Integration of a Chatbot
with Articial Intelligence in a Smart Cam-
pus for the Improvement of Learning. Sus-
tainability, 12(4), Article 4. https://doi.org/
10.3390/su12041500
Weber, R. E., Mueller, C., & Reinhart, C.
(2022). Automated oorplan generation
in architectural design: A review of me-
thods and applications. Automation in
Construction, 140, 104385. https://doi.org/
10.1016/j.autcon.2022.104385