AI and the cityscape: Finding a compromise

By Vince Chong

SMU Office of Research Governance & Administration – Artificial intelligence, or AI, is now fairly ubiquitous in daily life. Virtual assistants, language translators, educational and entertainment platforms, are but some examples of how AI algorithms have massively impacted humanity.

On the other hand – and this might surprise some – AI systems face a much steeper slope when it comes to syncing with the physical landscape. As an SMU study has found, the “unpredictable, complex, and dynamic aspects of urban spaces” presents many challenges, particularly in cities like Singapore that strive to integrate nature into its metropolitan visage.

As Navigating AI–nature frictions: Autonomous vehicle (AV) testing and nature-based constraints set out, nature is a “paradigmatic example” of this unpredictability as it defies “consistent patterns and precise data-based modelling” that AI programmes rely on to work. 

The crucial question then, the study’s lead author and SMU Research Fellow Prerona Das said, is whether AI can be finetuned to co-exist with the natural environment.

“There is no easy answer,” she told the Office of Research Governance and Administration (ORGA). 

“As environmental conditions become more unpredictable, AI will need to be far more adaptive, and even then it won’t always be able to replicate human expertise or experiential knowledge, especially in complex, fast-changing natural settings.” 

Take a tree, for example. While humans process it as flora that grows and changes, an AI programme only recognises the entity that was first mapped, not the one whose foliage droops and alters shape after a storm. Hence, an AV may simply stall on a road flanked by lush greenery, since it is unable to sense visual changes wrought by the weather, or pruning schedules. A veritable “blind spot,” the study noted. 

AVs trials can thus reveal “a broader set of tensions” in AI’s integration to urban environments.

So rather than expecting AI to fully “capture” the dynamics of the natural environment, said Dr Das, the focus should be on “combining artificial and natural intelligences to arrive at better outcomes.” 

“Human judgement, local knowledge and lived experience should still play a crucial role alongside AI systems,” she said, adding that Singapore poses the ideal environment for such research because it is not only a well-connected city in terms of infrastructure and data, but also a tropical locale where “heavy rain, heat, humidity and increasing weather variability” adds complexity to its urban sprawl.

“Any compromise will likely involve careful integration, where AI supports decision-making but remains grounded in human oversight and local environmental contexts.” 

Navigating AI-nature frictions is co-authored by two other SMU experts, Professor of Geography Orlando Woods, and Professor of Social Sciences Lily Kong, who is also SMU President.

“Nature-based constraints” 

Published in leading social sciences journal Big Data and Society, the paper is one of Dr Das’ latest studies on the intricacies of AI urbanism, a body of research that in 2025 won her a SMU Research Staff Excellence Award. This endeavour, the University said, included “extensive fieldwork” across multiple Southeast Asian hubs, “as well as analysis and conceptualisation of key findings” on the topic.

Dr Das has coauthored at least nine papers, SMU noted, including Navigating AI-nature frictions. The SMU Urban Institute Research Fellow has further won a study grant from Singapore’s Ministry of Education (MOE) for exploring the intersections between AI deployment and climate change governance in Asian cities, as well as a prestigious visiting fellowship from the University of Birmingham.

Navigating AI-nature frictions canvassed in-depth workaday perspectives from experts in Singapore involved in AV testing, including stakeholders engaged in AI governance and implementation. These spanned public agencies, institutions, and private companies including the Agency for Science, Technology and Research (A*STAR), JTC Corporation, three Singapore universities, as well as tech corporations like Amazon Web Services and Huawei Singapore.

The earlier example of how AI construes a tree came from one such interview with a physicist, who said elements of nature represent obstructions due to their unpredictability. Outwardly this jars with Singapore’s City-in-Nature approach and its ambition to be a leading AI research hub, but as another scientist remarked as well, the solving of such issues will give Singapore “first mover advantage” as similar phenomena emerge globally in the wake of climate change. 

Other challenges include the sensing of accompanying traffic technology such as traffic lights, and navigation of the “vertical realm” such as multi-storey car parks. Hence the concept of frictional urbanisms, the study noted, helps make sense of such challenges not as errors, but “symptomatic of deeper nonalignments between AI systems and urban environments that characterize AI deployment within cities.”

Differing interpretations, same problem

AI–nature frictions aside, the differences in priorities and mindsets across various stakeholders represent yet another “significant” obstacle, Dr Das continued.

“Different actors bring distinct priorities and ways of interpreting the same problem,” she explained, citing technical experts who often focus on feasibility, efficiency, and scalability, while planners, policymakers, and legal professionals tend to emphasise socio-cultural impact, governance, and regulatory constraints.

The real difficulty “lies in bridging these perspectives” to allow the various groups to work toward shared goals, rather than “operating in parallel.” 

In the meantime, other cities like San Francisco and Phoenix in the U.S. have successfully rolled out AVs, even if operations are not entirely perfect.  In general, Dr Das noted that AVs perform better “in relatively controlled and predictable environments, with clear lane markings, consistent road widths, and simpler traffic patterns that have supported smoother deployment.” 

On the other hand, Singapore’s dense urban environment is “inherently more complex,” with elements like the tropical weather, high pedestrian volumes, and constant interactions across different modes of transport working against sensor reliability. 

“Taken together, the lesson is not to replicate conditions from other cities, but to adapt by improving the technology and infrastructure where possible,” she added. 

“The final result must be to make AV systems robust enough to account for both urban complexity and tropical conditions.”

Back to Research@SMU May 2026 Issue