居民对自动驾驶汽车的偏好:基于可解释机器学习

Preferences for Autonomous Vehicles:An Interpretable Machine Learning Approach

新兴技术正在塑造城市和日常生活的同时,技术进步也可能加剧现有或带来新的不平等。本文以自动驾驶汽车为例,介绍了从交叉性视角出发的机器学习分析范式,并基于北京市居民对该技术偏好的试点调查,利用可解释机器学习—迭代随机森林模型,初步探究了个体多维特征之间相互交叉、建构所产生的行为偏好机制。结果表明,结合交叉性视角和可解释机器学习有效打破了传统方法论的局限,为推动新兴技术的公平转型提供了可操作性的分析框架。

The advent of autonomous vehicles(AVs)has given rise to a revolution in transportation systems and cities,greatly impacting people's daily life and the future of urban development.However,technological advancement affects different individuals'experiences of space and time differently.In this paper,we promote the application of the intersectional perspective in the study of emerging mobilities,and we call for intersectionality as a guiding principle in governing the development of emerging mobilities,to provide a better and more nuanced picture of how emerging mobilities might influence people's daily life.We conducted a pilot survey on residents'preferences for autonomous vehicles in Beijing in September 2022.We then used a novel interpretable machine learning approach:the Iterative Random Forests(IRF)model to quantify the preferences for taking autonomous vehicles,taking into account the intersectionality of spatial and socio-economic demographic characteristics of commuters.Finally,we present our findings and discuss the implications for future transportation planning.