Human-Centered Artificial Intelligence Institute (HAI), Stanford University, CA, USA.
Clark School of Engineering, University of Maryland, College Park, MD, USA.
Risk Anal. 2019 Oct;39(10):2160-2181. doi: 10.1111/risa.13313. Epub 2019 Apr 22.
The rise in economic disparity presents significant risks to global social order and the resilience of local communities. However, existing measurement science for economic disparity (e.g., the Gini coefficient) does not explicitly consider a probability distribution with information, deficiencies, and uncertainties associated with the underlying income distribution. This article introduces the quantification of Shannon entropy for income inequality across scales, including national-, subnational-, and city-level data. The probabilistic principles of Shannon entropy provide a new interpretation for uncertainty and risk related to economic disparity. Entropy and information-based conflict rise as world incomes converge. High-entropy instances can resemble both happy and prosperous societies as well as a socialist-communist social structure. Low entropy signals high-risk tipping points for anomaly and conflict detection with higher confidence. Finally, spatial-temporal entropy maps for U.S. cities offer a city risk profiling framework. The results show polarization of household incomes within and across Baltimore, Washington, DC, and San Francisco. Entropy produces reliable results at significantly reduced computational costs than Gini coefficients.
经济差距的扩大给全球社会秩序和地方社区的弹性带来了重大风险。然而，现有的经济差距衡量科学（例如基尼系数）并没有明确考虑到与基础收入分布相关的信息、缺陷和不确定性的概率分布。本文引入了跨尺度的收入不平等的香农熵量化，包括国家、次国家和城市层面的数据。香农熵的概率原理为经济差距相关的不确定性和风险提供了新的解释。随着世界收入趋同，熵和基于信息的冲突会增加。高熵实例既可以是幸福和繁荣的社会，也可以是社会主义-共产主义的社会结构。低熵信号表示异常和冲突检测的高风险临界点，具有更高的置信度。最后，美国城市的时空熵图提供了城市风险分析框架。结果表明，巴尔的摩、华盛顿特区和旧金山内部和之间的家庭收入出现两极分化。熵的计算成本显著低于基尼系数，但结果可靠。
