Journal of Quantitative Criminology最新文献

Objectives: Police routinely collect unstructured narrative reports of their interactions with civilians. These accounts have the potential to reveal the extent of police engagement with vulnerable populations. We test whether large language models (LLMs) can effectively replicate human qualitative coding of these narratives-a task that would otherwise be highly resource intensive.

Methods: Using publicly available narrative reports from Boston Police Department, we compare human-generated and LLM-generated labels for four vulnerabilities: mental ill health, substance misuse, alcohol dependence, and homelessness. We assess multiple LLM sizes and prompting strategies, measure label variability through repeated prompts, and conduct counterfactual experiments to examine potential classification biases related to sex and race.

Results: LLMs demonstrate high agreement with human coders in identifying narratives without vulnerabilities, particularly when repeated classifications are unanimous or near-unanimous. Human-LLM agreement improves with larger models and tailored prompting strategies, though effectiveness varies by vulnerability type. These findings suggest a human-LLM collaborative approach, where LLMs screen the majority of cases whilst humans review ambiguous instances, would significantly reduce manual coding requirements. Counterfactual analyses indicate minimal influence of subject sex and race on LLM classifications beyond those expected by chance.

Conclusions: LLMs can substantially reduce resource requirements for analyzing large narrative datasets, whilst enhancing coding specificity and transparency, and enabling new approaches to replication and comparative analysis. These advances present promising opportunities for criminology and related fields.

Objectives: Examine the degree of crime concentration at micro-places across six large cities, the spatial clustering of high and low crime micro-places within cities, the presence of outliers within those clusters, and extent to which there is stability and change in micro-place classification over time.

Methods: Using crime incident data gathered from six U.S. municipal police departments (Chicago, Los Angeles, New York City, Philadelphia, San Antonio, and Seattle) and aggregated to the street segment, Local Moran's I is calculated to identify statistically significant high and low crime clusters across each city and outliers within those clusters that differ significantly from their local spatial neighbors.

Results: Within cities, the proportion of segments that are like their neighbors and fall within a statistically significant high or low crime cluster are relatively stable over time. For all cities, the largest proportion of street segments fell into the same classification over time (47.5% to 69.3%); changing segments were less common (4.7% to 20.5%). Changing clusters (i.e., segments that fell into both low and high clusters during the study) were rare. Outliers in each city reveal statistically significant street-to-street variability.

Conclusions: The findings revealed similarities across cities, including considerable stability over time in segment classification. There were also cross-city differences that warrant further investigation, such as varying levels of spatial clustering. Understanding stable and changing clusters and outliers offers an opportunity for future research to explore the mechanisms that shape a city's spatiotemporal crime patterns to inform strategic resource allocation at smaller spatial scales.