AI reveals faster walking, less socializing in cities over 50 years

A subtle shift is occurring within our shared urban landscapes. While pavements remain bustling and parks vibrant, a deeper examination—and, more importantly, precise measurement—reveals a transformation in the very fabric of our public interactions.

Working alongside colleagues from Yale, Harvard, and other institutions, we employed artificial intelligence to compare video footage of public spaces from the 1970s with recent recordings from the identical locations in New York, Boston, and Philadelphia. The findings are striking: people now walk at a faster pace, linger less, and are less inclined to spontaneously connect with others. This phenomenon is perhaps unsurprising in an era where smartphones, streaming services, and AI companions increasingly draw us away from tangible spaces and real-world relationships. Yet, if technology contributes to the problem, it might also offer a path toward its resolution. By leveraging AI to scrutinize urban public areas, we can gather invaluable data, identify behavioral patterns, and test innovative designs that could help us reimagine the modern equivalent of the agora—the ancient Greek marketplace and primary civic gathering place.

Urban environments have long captivated curious minds, none more incisive than William “Holly” Whyte. In the 1970s, Whyte meticulously filmed plazas and parks across New York, fascinated by how people chose to sit, navigated shared spaces, and formed connections. His observations, documented in the seminal 1980 work The Social Life of Small Urban Spaces, yielded insights that were often elegantly simple, such as: “What attracts people most, it would appear, is other people.” From his extensive footage, Whyte distilled data-backed recommendations, advocating for benches “two human backsides deep” and championing movable chairs that allowed users to seek sun or shade. His analytical approach was instrumental in revitalizing New York spaces like Bryant Park and profoundly influenced contemporary people-centered urban design.

Whyte’s experiments were groundbreaking but notoriously difficult to replicate due to the sheer time commitment; analyzing footage frame by frame required months of dedicated effort from a team of assistants. Today, this challenge has been overcome with the advent of automated evaluation tools. Our team digitized Whyte’s original footage and juxtaposed it with recent videos—encompassing Bryant Park, the steps of the Metropolitan Museum of Art in New York, Boston’s Downtown Crossing, and Philadelphia’s Chestnut Street—collected by sociologist Keith Hampton. We then trained an AI model, similar to the technology enabling self-driving cars to recognize pedestrians, to analyze both sets of footage. What once took Whyte months can now be accomplished in mere minutes.

So, how have cities changed between 1970 and 2010? As detailed in a recent paper published in the Proceedings of the National Academy of Sciences, walking speeds have increased by a notable 15%. People stand still less frequently, and the occurrence of “dyads”—pairs meeting and then walking together—has declined. Boston’s Downtown Crossing, once a vibrant social hub, has largely transformed into a mere thoroughfare. Even Manhattan’s Bryant Park, despite improvements aligned with Whyte’s vision, has seen a reduction in social interactions. While our cities have not emptied, a crucial aspect of their essence seems to have diminished.

Various forces contribute to these shifts. Accelerating work rhythms mean time is increasingly perceived as a precious commodity, reducing our willingness to simply meander. Perhaps the allure of a Starbucks coffee shop now outweighs a leisurely park visit. Even in 2010, when the iPhone was barely three years old, people were already being drawn into personalized digital streams, forsaking the wandering gaze of the traditional flâneur.

This trend poses a significant threat to our social fabric. Online, we often gravitate towards curated echo chambers, easily scrolling past discomfort and filtering out dissenting viewpoints. Public space, in stark contrast, remains gloriously unfiltered. It invites friction, embraces messiness, and offers unexpected encounters: a rival football fan holding a door, or children playing across language barriers. If we spend less time in public spaces, we risk losing our tolerance for the wider public—and, by extension, the very habit of engaged citizenship.

Paradoxically, the same technologies that draw us inward might also help guide us back out. Social media’s addictive nature stems from algorithms constantly testing user preferences. If we apply AI to analyze outdoor public spaces, we can achieve a similar effect: equipping every park, plaza, and street corner with its own “personal William Whyte” to test potential improvements. What types of seating best foster interaction? Could the addition of greenery or water features create a more comfortable microclimate? Which public games might encourage strangers to break the ice? Temporary design interventions could be introduced, evaluated with AI, and iteratively refined through a process of trial and error, allowing public spaces to evolve organically, much like nature itself.

To this end, architects should embrace new AI tools, a point we emphasized at this year’s Biennale Architettura in Venice. But how should these tools be wielded?

First, with humility. Public spaces of the past were far from perfect, often excluding women, minorities, and individuals with access needs. We must neither romanticize them nor passively surrender to a technology-driven present. Optimizing public life solely through data risks repeating the errors of high modernism, a design philosophy that often imposed top-down solutions without sufficient human consideration. AI can reveal patterns, but it cannot dictate what constitutes “good.”

Second, with curiosity. Public space is not static; it is a living entity, responsive to heat, light, geometry, and programming. Small interventions—a strategically placed bench in the shade, a water fountain on a hot day, a winding path instead of a shortcut—can profoundly alter behavior. In a recent study in Milan, we observed that compliance with 30km/h speed limits had less to do with signage and more with street geometry. What truly slows us down is thoughtful design, not mere instruction.

Climate change also plays an increasingly critical role. As temperatures rise across southern Europe, many urban spaces remain shaped by outdated climatic expectations. Sicily can now cultivate mangoes, yet its public squares offer minimal protection from intense heat. We might draw lessons from cities like Singapore, where the orchestration of vegetation, water, and shading is actively employed to mitigate heat. If Europe’s climate is changing, its public spaces must adapt accordingly.

The deeper challenge lies in overcoming a long-standing disconnect: designers have often worked remotely, imagining how people should behave from studios far removed from the street. Today, we possess tools that allow us to observe how people actually behave, to test hypotheses, and to prototype joy and proximity. Yet, these tools must be employed not merely for optimization, but for careful stewardship. If used wisely, they can help counteract the hollowing out of public space. The agora is not dead; it simply requires reimagining. And if we approach this intelligently, AI might just help us achieve it, perhaps even helping us discern the fragile, elusive symphony of the commons.