This paper examines an important but underappreciated mechanism affecting urban segregation and integration: urban venues. The venue- an area where urbanites interact- is an essential aspect of city life that tends to influence residential location. We study the venue/segregation relationship by overlaying venues onto Schelling’s classic (1971) [1] agent-based segregation model. We show that a simulation world with venues makes segregation less likely among relatively tolerant agents and more likely among the intolerant. We also show that multiple venues can create spatial structures beyond their catchment areas and that the initial location of venues shapes later residential patterns. Finally, we demonstrate that the social rules governing venue participation alter their impacts on segregation. In the course of our study, we compile techniques for advancing Schelling-style studies of urban environments and catalogue a set of mechanisms that operate in this environment.

This paper recommends that critical attention towards machine learning should be focused on the ordering procedures at work in these models. More precisely, it draws attention to the central role of ‘latent spaces.’ The paper first explores ‘latent space’ through a series of analogies, and then briefly situates the concept in relation to a genealogy reaching back to developments in mathematical statistics at the turn to the 20th century.