Exploring the spatial variation in people’s everyday conceptualizations of nearness using large language models

Nearness is a central concept in geographical and planning research, yet its meaning is ambiguous and dependent on context and individual experience. At the moment, it remains unclear to what extent people’s conceptualizations of nearness vary, and whether this variation is substantial enough to justify incorporating individualized conceptualizations in research as a matter of practice. Accordingly, we pursue two complementary objectives. First, we examine how people conceptualize nearness in everyday contexts by analyzing its linguistic and spatial variation in a textual dataset using a large language model (LLM). Second, we evaluate the potential of LLMs in interpreting such intersubjective geographic information. A post-facto evaluation shows that LLMs can perform at near-human levels, although errors persist across individual observations, suggesting that LLMs are best suited for deriving general insights from aggregated datasets. Furthermore, we find that nearness expressions are often grounded in everyday mobility experiences and that types of nearness expressions tend to cluster spatially. Consequently, we argue that incorporating experiential nearness conceptualizations in research is worthwhile and can help align operationalizations more closely with people’s perceptions. While our exploratory analysis represents a first step, further research is needed to translate these insights into practical measures for research and planning.