SCC-IRG Track 2: Real-Time Algorithms and Software Systems for Heterogeneous Data Driven Policing of Social Harm

SCC-IRG 第 2 轨：用于异构数据驱动的社会危害警务的实时算法和软件系统

• 批准号: 1737585
• 负责人: George Mohler
• 金额: $ 79.15万
• 依托单位: Indiana University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2022-02-28
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1737585&HistoricalAwards=false
• 关键词: SCC; IRG; Track; Real; Time

Communities are adversely affected by social harm events such as crime, traffic crashes, medical emergencies, and drug usages. This proposal aims to develop algorithms and software systems for the collection, analysis, and dynamic prediction of social harm events to facilitate appropriate government interventions to improve the quality of life in communities. The project has a significant community engagement component and software developed through the research will be used by the Indianapolis Metropolitan Police Department (IMPD), Indianapolis Emergency Medical Services (EMS), National Alliance of Mental Illness, the Indiana prosecutor's office, and individual citizens for sharing of social harm analytics and collaboration in social harm intervention. This objective will be achieved by: i) creating software systems for cross-agency social harm data integration, ii) developing mathematical models for capturing social harm event dynamics along with public trust and grievance towards police, and iii) conducting a field trial of the developed software system in Indianapolis. The methods developed in the project will also be applicable to other smart and connected communities across the country and could be used for data analytics integration and allocation of resources across government departments. Graduate students from both social science and computing disciplines will be trained in interdisciplinary research methods that span criminal justice, statistics, and computer science. Research interests in the domain of algorithms for heterogeneous data in smart cities will be encouraged through a workshop hosted by the investigators at Indiana University-Purdue University Indianapolis (IUPUI). Social harm data resides within a disconnected set of community databases and current methodologies for modeling social harm neglect space-time dynamics altogether or focus on a small related subset of event types. Furthermore, interventions are designated in spatial locations for several weeks or months at a time, failing to account for the daily changes in risk of social harm events where crime, traffic crashes, and medical emergencies cluster in different times and locations in communities. Current policing interventions that focus on spatial risk (i.e., hotspots) are often too narrow and seek only to optimize crime reductions. In order to address some of these limitations, this project will develop: i) software systems for heterogeneous social harm data integration, ii) new marked point processes for modeling heterogeneous social harm event dynamics including trust and grievances towards police, iii) optimal control methods for space-time point processes that are lacking in current point process research, and iv) near real time software-human systems for deploying hourly interventions to dynamically changing risk. During phases one and two, the project team will work collaboratively with IMPD's community policing unit and leverage this unit's relationships with local neighborhood watch, faith-based, juvenile diversion, and volunteer groups that are predominantly comprised of minority community members serving largely minority neighborhoods. This collaboration will facilitate broad community buy-in for phase three and enable communication with and recruitment of community groups disproportionately exposed to social harm risk. The last phase of the project will include a randomized controlled trial of heterogeneous data driven policing in Indianapolis in collaboration with IMPD, Indianapolis EMS, Indianapolis Mayor's Office, National Alliance of Mental Illness, Marion County Prosecutor's Office, the Indy Public Safety Foundation, and the general public who will be encouraged to download a version of the application through a press release prior to the trial launch. In the trial, the extent to which police in partnership with community stakeholders can respond to dynamic, heterogeneous social harm hotspots will be investigated and the impact across four types of social harm (crime, traffic crashes, EMS calls for service, and community trust in police within high risk communities) will be measured.

社区受到犯罪、交通事故、医疗紧急情况和吸毒等社会危害事件的不利影响。 该提案旨在开发用于收集、分析和动态预测社会危害事件的算法和软件系统，以促进政府采取适当干预措施，改善社区生活质量。 该项目有一个重要的社区参与组成部分，通过研究开发的软件将被印第安纳波利斯大都会警察局（IMPD），印第安纳波利斯紧急医疗服务（EMS），全国精神疾病联盟，印第安纳州检察官办公室和个人公民用于共享社会危害分析和社会危害干预合作。这一目标将通过以下方式实现：i）创建跨机构社会危害数据集成的软件系统，ii）开发数学模型，用于捕获社会危害事件动态沿着公众对警察的信任和不满，以及iii）在印第安纳波利斯对开发的软件系统进行现场试验。 该项目开发的方法也将适用于全国其他智能和互联社区，并可用于政府部门的数据分析整合和资源分配。 来自社会科学和计算机学科的研究生将接受跨学科研究方法的培训，这些方法涵盖刑事司法，统计学和计算机科学。 智能城市中异构数据算法领域的研究兴趣将通过由印第安纳州普渡大学印第安纳波利斯分校（IUPUI）的研究人员主办的研讨会得到鼓励。社会危害数据存在于一组不相连的社区数据库中，目前用于建模社会危害的方法完全忽略了时空动态，或者专注于事件类型的一个小的相关子集。此外，干预措施每次在空间位置上指定几周或几个月，未能考虑到社会危害事件风险的每日变化，其中犯罪，交通事故和医疗紧急情况在社区的不同时间和地点聚集。 目前的治安干预措施侧重于空间风险（即，热点地区）往往过于狭窄，只寻求最大限度地减少犯罪。 为了解决其中的一些限制，该项目将开发：i）用于异构社会危害数据集成的软件系统，ii）用于建模异构社会危害事件动态的新标记点过程，包括对警察的信任和不满，iii）当前点过程研究中缺乏的时空点过程的最优控制方法，以及iv）用于将每小时的干预部署到动态变化的风险的接近真实的时间的软件-人系统。在第一阶段和第二阶段，项目团队将与IMPD的社区警务单位合作，并利用该单位与当地邻里守望，信仰为基础的，青少年分流和志愿者团体的关系，主要由少数民族社区成员组成，主要服务于少数民族社区。这一合作将促进社区对第三阶段的广泛认同，并使社区团体能够与不成比例地面临社会伤害风险的社区团体进行沟通和招募。该项目的最后阶段将包括与IMPD，印第安纳波利斯EMS，印第安纳波利斯市长办公室，全国精神疾病联盟，马里恩县检察官办公室，印地公共安全基金会以及公众合作的印第安纳波利斯异构数据驱动警务的随机对照试验。 在试验中，警方与社区利益相关者合作的程度可以应对动态的，异质的社会危害热点将被调查，并在四种类型的社会危害（犯罪，交通事故，EMS要求服务，并在高风险社区内对警察的信任）的影响将被衡量。

项目成果

Reducing Bias in Estimates for the Law of Crime Concentration

• DOI: 10.1007/s10940-019-09404-1
• 发表时间: 2019-12-01
• 期刊: JOURNAL OF QUANTITATIVE CRIMINOLOGY
• 影响因子: 3.6
• 作者: Mohler, George;Brantingham, P. Jeffrey;Short, Martin B.
• 通讯作者: Short, Martin B.

Impact of social distancing during COVID-19 pandemic on crime in Los Angeles and Indianapolis

• DOI: 10.1016/j.jcrimjus.2020.101692
• 发表时间: 2020-05-01
• 期刊: JOURNAL OF CRIMINAL JUSTICE
• 影响因子: 5.5
• 作者: Mohler, George;Bertozzi, Andrea L.;Brantingham, P. Jeffrey
• 通讯作者: Brantingham, P. Jeffrey

Repurposing recidivism models for forecasting police officer use of force

重新利用累犯模型来预测警察使用武力

• 发表时间: 2020
• 期刊: IEEE International Conference on Big Data
• 作者: Khorshidi, S.;Carter, J.G.;Mohler, G.
• 通讯作者: Mohler, G.

Group Link Prediction

• DOI: 10.1109/bigdata47090.2019.9006261
• 发表时间: 2019-12
• 期刊: 2019 IEEE International Conference on Big Data (Big Data)
• 作者: Andrew Stanhope;Hao Sha;Danielle Barman;M. Hasan;G. Mohler

A modified two-process Knox test for investigating the relationship between law enforcement opioid seizures and overdoses

改进的两过程诺克斯测试，用于调查执法阿片类药物缉获和过量之间的关系

• DOI: 10.1098/rspa.2021.0195
• 发表时间: 2021
• 期刊: Physical and Engineering Sciences
• 作者: Mohler, G.;Mishra, S.;Ray, B.;Magee, L.;Huynh, P.;Canada, M.;O’Donnell, D.;Flaxman, S.
• 通讯作者: Flaxman, S.