The intersection of AI and cities

Historically, cities and technological advancements have always been inseparable. From the introduction of spatial analysis to sensor-based traffic management systems, our cities have continually been reshaped alongside new innovations. Today, the emerging artificial intelligence (AI) is seen as the future of urban governance and planning. Cities have started to utilise AI at different layers of creating cities, particularly in the way it may have a crucial impact on the 'plan-making' processes. Practitioners have used it as a tool for urban planning, particularly working towards a more autonomous urban development. AI-based tools offer potential solutions for operationalising more efficient cities, improving service delivery, and even shaping urban design. There have also been ongoing conversations on further leveraging AI in response to pressing, complex challenges such as climate change.

As urban practitioners advocating for more inclusive and sustainable cities, we must not only acknowledge AI’s inevitable influence but also more critically explore the potential of utilising it to enable solutions that are relevant and contextual. However, like any new technology, further utilising AI in urban planning and design also presents critical challenges, particularly related to its ethical concerns. We might want to ask questions: What does it mean to incorporate AI into our planning and design processes? To what extent does AI truly support the creation of more just and democratic urban spaces? And how do we ensure its application reflects the principles of inclusivity and environmental justice?

AI, the panacea that it is not

One of the concerns of using AI for city-making is how we navigate the likely possibilities that there may be inherent biases embedded in the data and assumptions used. As AI is trained on existing datasets, it could reflect biases that put specific communities in a more disadvantaged position—for example, biases and stigmas related to gender, ethnicity, abilities, and sexual orientation. We need to consider this challenge, particularly in regard to making sure that our efforts won’t reinforce exclusionary practices or marginalise vulnerable communities.

Furthermore, context matters. In places where data infrastructure is insufficient, it becomes more difficult to ensure the data used by the AI-based solution does not overlook the often marginalised communities and their needs. Missing data or paper-based records are still a common problem in Indonesian cities. We might need to consider many more factors when applying the same logic for a model in different contexts where data availability differs. For instance, in places where sustainability means different things, we might want to include meanings and local understanding, which, in the end, might require different sets of data. Thus, in the context of generating urban solutions, considering elements of contextuality would be fundamental so that we come up with solutions that are actually relevant.

An increasingly common case for AI in city-making is the use of generative AI to visualise future urban scenarios. Based on our experience, while the immediate conjuring of images displaying a variety of scenarios can be compelling in facilitating discussions, the visual output remains shaped by the data it has been trained on. 

Our initial impression in using generative AI  for a co-design process is that the lack of representation of diverse urban contexts could generate images and solutions that may be disconnected from the realities of many communities. Kota Kita's urban designer, Bima Pratama, shared this challenging experience, "As I facilitated a co-design workshop with communities, I was asked to add a traditional ethnic house as part of the design. It went through many iterated prompts to finally get to the one participants were most okay with. Because when we generate pictures using AI, mostly it generates visuals of European cities". 

There have been conversations around focusing on personalised algorithms, instead of generalised algorithms when it comes to using generative AI for planning and designing spaces. This is especially important to accommodate contexts and different needs that will be reflected in the output. Specific question-based modelling would also be potentially useful to train AI for particular purposes, with questions disclosed so that users are well-informed. Another interesting idea is how we might want to incorporate a ‘positive bias’ into the model—assumptions that affirm certain values such as justice and inclusivity, making sure it is responding to the limitations of the existing data and prejudice. In the end, the important thing for planners is a “proactive approach to mitigate ethical risks associated with AI”, that we acknowledge certain limitations embedded within the technology and not taking them for granted. 

Transitioning towards future cities

Eventually, the use of technology reflects power, and the further direction will depend on how we put meaning to it. In this sense, we might want to think of AI as part of a transition towards ‘the imagined’ future cities. Technology such as AI should develop into something that we can use as a tool to empower practitioners and communities to come up with more informed decisions about the direction of our cities. It certainly does not replace human decision-making with design and planning—but instead enhances the processes. Bima also emphasised how it helps generate options based on things we know. Here, providing multiple alternatives facilitates a more meaningful conversation about what is actually possible.

As AI’s integration into urban making will be more common in the future, the conversation is not just about technological advancement but also about the kind of cities we want to live in. It goes back to asking ourselves what would be the collective vision wherein AI can serve as a tool for us to reimagine and build cities that reflect our aspirations. The influence of AI for efficiency, for better delivery of public services or more informed urban planning, is not the end goal; it is the means. Addressing this emerging trend, I propose to base our engagement with AI on the understanding that cities are not passive recipients: we might respond or even resist certain technologies—we might reshape it in many ways. 

What are your thoughts about the use of AI in city-making?