The physical structures in which urban life occurs are an underappreciated determinant of how grassroots urban politics unfolds. In many rapidly growing cities, housing scarcity forces residents into multifamily buildings that create daily exposures to neighbors. We argue that these exposures affect political behavior by shaping residents’ access to political information and capacity for collective action. We focus on the informal, vernacular architecture of West Africa’s dominant urban housing form -- the compound house. Compound house residents in urban Ghana participate more in politics than similar residents of other housing types. Leveraging an original survey, including novel measures of tenants’ spatial network centrality within their residential buildings, we suggest that key mechanisms for this relationship emerge from the effects of architectural design on visibility and social ties among co-tenants. Ultimately, built environments must be studied alongside demographic environments to best understand contextual effects on political behavior.

Researchers across disciplines increasingly use Generative Artificial Intelligence (GenAI) to label text and images or as pseudo-respondents in surveys. But of which populations are GenAI models most representative? We use an image classification task—assessing crowd-sourced street view images of urban neighborhoods in an American city—to compare assessments generated by GenAI models with those from a nationally representative survey and a locally representative survey of city residents. While GenAI responses, on average, correlate strongly with the perceptions of a nationally representative survey sample, the models poorly approximate the perceptions of those actually living in the city. Examining perceptions of neighborhood safety, wealth, and disorder reveals a clear bias in GenAI toward national averages over local perspectives. GenAI is also better at recovering relative distributions of ratings, rather than mimicking absolute human assessments. Our results provide evidence that GenAI performs particularly poorly in reflecting the opinions of hard-to-reach populations. Tailoring prompts to encourage alignment with subgroup perceptions generally does not improve accuracy and can lead to greater divergence from actual subgroup views. These results underscore the limitations of using GenAI to study or inform decisions in local communities but also highlight its potential for approximating “average” responses to certain types of questions. Finally, our study emphasizes the importance of carefully considering the identity and representativeness of human raters or labelers—a principle that applies broadly, whether GenAI tools are used or not.

In recent decades, activists and leaders of government and nongovernment organizations have increasingly and explicitly called for greater attention to human dignity in their efforts to promote pro-social relations. In this study, we investigate whether appeals to this core human value actually influence how individuals act with regard to those who might be otherwise ignored or neglected. Using the digital advertising platform on Facebook, we randomly assign ads to over 90,000 adult American users to estimate the effects of dignity appeals on their likelihood of engaging with content concerning people facing homelessness or incarceration. Consistent with preregistered hypotheses and specifications, we find that adding dignity appeals increases the likelihood of positive reactions to such ads, but only when the vulnerable are considered less “blameworthy” for their situation.

Can repeated casual contact between members of different groups reduce exclusionary attitudes and behaviors? I assess the finding that casual contact exacerbates conflict between groups by distinguishing whether casual contact is repeated. Repeated casual encounters create “familiar strangers” (Milgram 1970), a common, yet often neglected, social relationship that can soften boundaries between groups by providing information about out-group members' behavior. I test this proposition in urban South Africa. I analyze a dataset of 1.2 billion geographically precise cellphone pings to differentiate types of contact. I demonstrate that while contact with a larger number of immigrants predicts more extreme xenophobic attitudes and higher vote shares for an anti-immigrant party, this propensity is reduced where the incidence of familiar stranger ties is higher. I complement these results with a survey experiment that randomly varies the familiarity of a migrant featured in photos of neighborhoods in Johannesburg and find that the visual familiarity of the immigrant influences respondents' perception of the safety. By distinguishing when casual contact moderates political conflict, I demonstrate the promise of incorporating familiar stranger relationships into theories and measures of intergroup contact.

I introduce the concept of spatial capital to help explain where cooperation can emerge in urban communities featuring weak social institutions. Defining it as a social resource created through superficial, nonverbal, but repeated encounters, I theorize that it can support community cooperation by reducing anonymity and increasing findability. I test this hypothesis in urban Ghana, a context characterized by transience and diversity. I propose and validate a measure of spatial capital and link it to a behavioral measure of cooperative capacity - trash accumulation - that draws on image data. I demonstrate that intersections with higher levels of spatial capital also exhibit lower litter levels. Using original surveys and a survey experiment, I show that repeated encounters shift individuals' sense of anonymity and findability but do not increase social engagement, trust, or cohesion. These findings suggest the need for further attention to the political and social effects of shared spatial networks.