Ten Across
Artificial intelligence’s emergence as a tool for city planners
Credit: Jonathan Kemper for Unsplash.
As A.I. technology develops at a rapid pace, what are city officials doing to both implement and regulate the new tool on governmental level?
Editor’s note: This article is part of a collaboration between APM Research Lab and the Ten Across initiative, housed at Arizona State University.
by RITHWIK KALALE | June 7, 2024
In 2023, artificial intelligence took the world by storm. From tools like ChatGPT to Microsoft Edge’s CoPilot, the technology found its way into many sectors – from social media and education to medicine, movies, and music. In September 2023, a survey conducted by Salesforce found that three quarters of generative AI users in the U.S., U.K., Australia and India are “looking to automate tasks at work and use generative AI for work communications” compared to just over one-third of respondents who use AI for fun or to learn.
As the technology becomes even more ubiquitous, so will its presence in various spaces– including city and government planning.
AI and local transportation planning
City adoption of AI for city and regional planning is already underway in Arizona. According to Phoenix Mayor Kate Gallego, full city adoption of the technology will be on the ballot this November in the form of the metropolitan area’s new regional transportation plan, Prop 479.
In a recent interview with Ten Across’ Executive Director, Duke Reiter, Gallego described how the plan was a collective effort by “all of the mayors in the region, with our counterpart tribal nation leaders. It has hundreds of millions of dollars in investments in smart technology, which would address things in your daily commute.”
If passed by voters, Prop 479 would set aside $12.5 million every year for emerging technologies, allowing Phoenix to maintain its status as a regional leader in technology adoption.
The city already has traffic signals that use AI to adjust to surrounding conditions, for example pedestrians or cars running red lights. Prop 479 would enact this technology throughout the whole metro area.
While AI for transportation purposes is set to be voted on by Arizonans in the fall, there is still a long way to go before the technology can be integrated into multiple city functions.
Using AI as a tool rather than the solution
“There’s a lot of excitement about AI in the city, but what I’m working on right now is getting people to pause for a moment to just think about ‘What does AI really do? What risks are in there?’ We have to talk about the quality of the data, access to data, potential bias in our data, risks associated with any decisions that might be made using output from AI,” said Stephanie Deitrick at Arizona State University’s Smart Region Summit.
Deitrick, the Chief Data and Analytics Officer for the City of Tempe, was responsible for writing the city’s ethical AI policy and getting it adopted by the city council in 2023.
Deitrick was not the only city official who has observed that the excitement for AI city implementation exceeds the ability to actually put it into practice.
At the same summit, Liz Crowe, Director of Urban Analytics and Innovation for the city of Cleveland, commented, “Cities, by nature, are very risk-averse and resource constrained. We can’t afford to have an AI model that optimizes our trash routes and maybe it forgets about 10 streets.”
Crowe mentioned that defining generative AI was the first obstacle when it came to implementing it into a governance plan. Additionally, tech and data literacy levels not being uniform among residents across the city was another factor.
Tempe, Arizona’s adoption of artificial intelligence has its drawbacks, according to Deitrick. The crux of the issue is taking enough time to weigh the costs and risks of implementing a new type of technology on a city-wide level, while also making sure that the implementation plan is not outdated by the time it has been tested and approved.
“Our policy basically is built on the fact that this needs to be very people-focused. It’s human-centered. There always has to be someone who is involved in the process and someone (who) is being held accountable for what is being done with the system, so that if something happens or if there’s a question there is someone to go back to.”
She mentioned that implementing AI as one tool among many, instead of using it as a replacement for all other tools and taking its outputs as “gospel,” lowers risk in planning, in much the same way a single human employee’s suggestion wouldn’t be implemented immediately without review and revision first.
The importance of a human team working with AI is not only to ensure that it isn’t producing any false data, but also to provide a check and balance as the data it is operating on likely has some form of bias to begin with.
According to a report issued by the United Nations Human Settlements Programme, “AI use in cities: Risks, applications and governance,” the technology is not neutral. According to the report, “formal structures, such as the design of algorithms, and informal arrangements, such as social norms, embed and propagate values.”
Additionally, AI on its own cannot understand the social norms, cultures, or context of development in every city it is used in, especially when it comes to governance.
In conclusion
“As cities like Phoenix and Tempe navigate the integration of AI into their infrastructures, the balance between technological innovation and human oversight remains crucial. By carefully weighing the benefits and risks, and ensuring AI is used as a complement rather than a replacement, urban planners can harness the power of AI to create smarter, more responsive cities while safeguarding community values and public trust.”
NOTE: Due to the nature of this article we decided to let AI write the conclusion. The above paragraph was produced by ChatGPT (free version) after feeding in the entire text of the article and providing the prompt, “Write a two-sentence conclusion to this article.” We did proofread to ensure that no errors were made, but in this case the results proved useful without any edits.