Content uploaded by Jonathan Dortheimer
Author content
All content in this area was uploaded by Jonathan Dortheimer on Jul 29, 2022
Content may be subject to copyright.
CHATBOTS IN THE DESIGN PROCESS - Automating Design Conversation in Urban Design
Projects
JONATHAN DORTHEIMER, NIKOLAS MARTELARO, GERHARD SCHUBERT
Technion Israel Institute of Technology, Carnegie Mellon University, Technical University of Munich
And
AARON SPRECHER
Technion Israel Institute of Technology
1. Introduction
Urban design is an institutional design process that aims to develop cities and improve communities.
However, since cities are complex systems with many sub-systems, urban design is a challenging design
task (Alexander, 1966; Portugali, 2012). Consequently, urban design is mostly a top-down design process
often characterized by a distancing between designers and end-users. Post-modern critics, such as Jane
Jacobs and Christopher Alexander, argued that the top-down design approach led to inferior design in
large-scale urban projects (Alexander, 1966; Jacobs, 1961). To address this issue, in the past 60 years,
several participatory design (PD) methods that facilitate interaction between designers and end-users have
been introduced and gained popularity, such as workshops (Luck, 2018) or tools using gamification
methods (Muehlhaus et al. 2022).
Overall, it has been argued that participatory practices are essential for creating "just cities" and
fostering inclusive communities (Fainstein, 2010; Selle, 2005). However, discussing design with all
end-users can be problematic since urban design is politically charged (Luck, 2018), end-users have
significant knowledge gaps (Dortheimer & Margalit, 2020), and PD requires much time and resources
(Gooch et al., 2018).
This research project proposes artificial intelligence (AI) tools to address these challenges by
automating and scaling design conversations with thousands of end-users in a design project. From the
point of view of PD, such a tool offers end-users the capability to influence and intervene in design
decisions before and during the design process.
2. Method
We produced an AI chatbot using the GPT-3 language model (Brown et al., 2020) to discuss
architectural design projects with end-users and produce a complete and structured design brief. The
language model can predict how an architect would respond to the user's interaction and deliver a
personalized and graceful conversation. Furthermore, the language models predicted a list of
requirements resulting from the discussion. Together, this unique chatbot platform can be scaled to
interact with thousands of people and organize their insights to be valuable to designers.
We conducted an experiment where participants (n = 54) discussed any architectural project of their
choice. Next, the chatbot generated a list of design requirements (432 items) from the discussion
transcript (2406 messages) that the participants and the research team evaluated. Finally, the participants
completed a user experience survey to evaluate their experience.
3. Results
Experimental results from the early prototype of the chatbot indicate both the promises and challenges
in this technology. The results revealed that the chatbot had demonstrated the ability to effectively convey
and augment a design discussion (83% correct response rate) with a positive user experience and a design
brief that captured 84% of the requirements with a 75% success rate.
This protocol contributes a novel way to facilitate design discussions using large language models.
Furthermore, by providing personal attention and patience to each end-user, being multilingual, and
ignoring spelling mistakes, a chatbot can increase the influence of underrepresented communities (Gooch
et al., 2018; Chen et al., 2020). Consequently, we argue that chatbots can be helpful for designers to tap
into the tacit knowledge of communities of large urban projects.
4. Future Research
In our future research, we will perform field research where the chatbot will implement in an
architectural PD project in northern Israel. In this project, a 3000 m2 community-oriented building will be
planned in the proximity of Jewish localities Rosh Pina and Hazor Haglilit and the underrepresented
Beduin Arabic community of Tuba-Zangriya. The results from this experiment will reveal the potential of
using this AI for social good, strengthening communities, and improving cities.
References
Alexander, C:1966, A City is not a tree, Design, 206.
Brown, TB., Mann B, Ryder N, Subbiah M, Kaplan J, Dhariwal P, Neelakantan A, Shyam P, Sastry G, Askell A, Agarwal S,
Herbert-Voss A, Krueger G, Henighan T, Child R, Ramesh A, Ziegler DM, Wu J, Winter C, Hesse C, Chen M, Sigler E,
Litwin M, Gray S, Chess B, Clark J, Berner C, McCandlish S, Radford A, Sutskever I, and Amodei D: 2020, Language
Models are Few-Shot Learners, arXiv, http://arxiv.org/abs/2005.14165
Chen, Z, Lu Y, Nieminen MP, and Lucero A, 2020: Creating a chatbot for and with migrants: Chatbot personality drives
co-design activities, DIS 2020 - Proceedings of the 2020 ACM Designing Interactive Systems Conference, pp. 219–230.
https://doi.org/10.1145/3357236.3395495
Dortheimer, J & Margalit, T: 2020, Open-source architecture and questions of intellectual property, tacit knowledge, and liability,
Journal of Architecture, 25(3), pp. 276–294. https://doi.org/10.1080/13602365.2020.1758950
Fainstein, SS, 2010: The Just City. Cornell University Press.
Gooch, D., Barker, M., Hudson, L., Kelly, R., Kortuem, G., Van Der Linden, J., Petre, M., Brown, R., Klis-Davies, A., Forbes,
H., Mackinnon, J., Macpherson, R., & Walton, C: (2018), AmplifyingQuiet voices: Challenges and opportunities for
participatory design at an urban scale. ACM Transactions on Computer-Human Interaction, 25(1), 1–34.
https://doi.org/10.1145/3139398
Jacobs, J: 1961, Death and life of great American cities. Vintage books.
Muehlhaus SL, Eghtebas C, Seifert N, Schubert G, Petzold F and Klinker G (2022) Game.UP: Gamified Urban Planning
Participation Enhancing Exploration, Motivation, and Interactions. International Journal of Human–Computer Interaction,
https://doi.org/10.1080/10447318.2021.2012379
Luck, R: 2018, Participatory design in architectural practice: Changing practices in future making in uncertain times. Design
Studies, 59, pp.139–157. https://doi.org/10.1016/j.destud.2018.10.003
Portugali, J: 2012, Complexity theories of cities: Implications to urban planning. In J. Portugali, H. Meyer, E. Stolk & E. Tan
(Eds.), Complexity theories of cities have come of age: An overview with implications to urban planning and design,
Springer Berlin Heidelberg, pp. 221–244. https://doi.org/10.1007/978-3-642-24544-213
Selle, K: 2005, Planen, Steuern, Entwickeln: Uber den Beitrag offentlicher Akteure zur En-twicklung von Stadt und Land.
Dortmunder Vertrieb f ur Bau-und Planungsliteratur.
