CRII: CHS: Estimating the Financial, Social and Ethical Impacts of Algorithmic Crime Analysis

本研究旨在了解犯罪分析算法对金融、社会和伦理的影响。算法的误用可能造成巨大的伤害和歧视，特别是在对刑事罪犯进行判决时。检查和调查算法偏差对现实世界的影响是一个非常新颖但越来越重要的领域。目前的文献主要集中在对算法的哲学和社会学批评上。迫切需要的是建立在这些社会技术批评基础上的数学和计算调查，以便为实际影响提供背景和定量证据。我们需要知道人们的偏见、假设、规范和价值观是如何影响算法的感知和应用的。更具体地说，政府经常寻求将算法应用于犯罪分析，通常是为了提高财务效率，减少人为错误，简化官僚机构。因此，本研究旨在了解在刑事司法系统中部署算法的实际、财务、社会和伦理影响，以制定基于证据的政策。这个项目有四个主要目标。首先，使用犯罪地图作为调查的镜头，该项目将调查对算法犯罪分析的金融、社会和伦理维度的看法，以揭示外行人和专家如何看待其各种影响的定性和定量见解。其次，该研究将解构犯罪分析中使用的流行算法，以了解可能发生偏见的具体点，并在统计上更好地减轻偏见。第三，它将比较和对比每一种有偏见的变体与不那么有偏见的变体，以计算有关金融、社会和道德影响的指标，为有偏见的刑事司法政策提供定量影响。第四，作为前三个目标的结果，该提案将有助于扩大最初的努力，建立一个志同道合的学者社区，通过组织一个涉及从业者、研究人员和学生的夏季研讨会，以教育和共同创建一个围绕道德、公平、负责任和透明算法的实践社区。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This research seeks to understand the financial, social and ethical impacts of crime analysis algorithms. Misapplication of algorithms has the capacity for great harm and discrimination, notably when applied in sentencing decisions about criminal offenders. Inspecting and investigating algorithmic biases for their real-world impacts is an extremely novel but increasingly important area. The current literature has mostly focused on philosophical and sociological critiques of algorithms. What is sorely needed is a mathematical and computational investigation that builds on these socio-technical critiques in order to provide contextual and quantitative evidence for actual impact. We need to know how people's biases, assumptions, norms and values play into algorithmic perceptions and applications. More specifically, governments often seek to apply algorithms to crime analysis with the usually stated justification of financial efficiencies, reducing human error and smoothening bureaucracies. Therefore, this research seeks to understand the practical, financial, social and ethical impacts of deploying algorithms in the criminal justice system in order to develop evidence-based policies. This project has four main objectives. First, using crime mapping as a lens of inquiry, the project will investigate the perceptions of financial, social and ethical dimensions of algorithmic crime analysis, to uncover both qualitative and quantitative insights about how laypeople and experts perceive its various implications. Second, the research will deconstruct popular algorithms used in crime analysis in order to understand specific points at which bias may occur and statistically better variants that may mitigate them. Third, it will compare and contrast each biased variant with its less biased counterpart to compute metrics about financial, social and ethical impacts, providing quantitative implications for biased criminal justice policy. Fourth, as a result of the first three objectives, this proposal will help expand initial efforts to build a community of like-minded scholars, though organizing a summer workshop involving practitioners, researchers and students to educate and co-create a community of practice around ethical, fair, accountable and transparent algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.