Objectives

This study examines the potential of artificial intelligence (AI) to reduce the time police officers spend writing reports, a task that consumes a significant portion of their workday.

Methods

In a pre-registered randomized controlled trial, we test this claim within the patrol division of a medium-sized police department (n = 85) at the individual report level (n = 755). Analyses utilize mixed-effects regression accounting for the nested structure of report-writing.

Results

AI assistance did not significantly affect the duration of writing police reports. Alternative specifications beyond those specified in the pre-registration, including a difference-in-differences approach observing report duration over a full year (n = 6084), confirm the null findings are robust.

Conclusions

Our findings contradict marketing expectations for the effect of this technology, suggesting no time savings in report-writing can be expected when using AI-assisted report-writing. Several other potential effects remain possible and untested.

中文翻译：

---

无人之手：人工智能无法提高警方报案撰写速度

 目标

本研究考察了人工智能 （AI） 在减少警察撰写报告时间方面的潜力，这项任务占用了他们工作日的很大一部分时间。

 方法

在一项预先注册的随机对照试验中，我们在一个中型警察局 （n = 85） 的巡逻部门中在个人报告级别 （n = 755） 测试了这一说法。分析利用混合效应回归解释报告编写的嵌套结构。

 结果

AI 协助并没有显着影响撰写警方报告的持续时间。除预注册中指定的规格外的替代规格，包括观察一整年报告持续时间的双重差分方法 （n = 6084），证实了空调查结果是稳健的。

 结论

我们的研究结果与市场对这项技术效果的预期相矛盾，这表明使用 AI 辅助报告撰写时，可以预期不会节省报告编写时间。其他几种潜在影响仍然是可能的，但尚未经过测试。