Stay Happy: Understanding Urban Wellbeing Using a Behavioural Machine Learning Approach

信息和通信技术（ICT）的最新进展为公共部门创造了机会，通过更好地了解公民的偏好来增加公民的福祉。机器学习是计算机科学的一个分支，是了解消费者/用户偏好的一个有价值的工具，包括与城市发展和城市服务提供相关的偏好。事实上，通过将数据驱动的方法应用于城市商业建模，它可以帮助创造新的城市价值主张。城市领域的机器学习算法为政策制定者提供了有价值的见解。虽然现有的方法产生的模型可以很好地解释一些数据，但它可以通过纳入决策和行为科学的数学模型来改进，以更好地捕捉城市福祉的行为成分。行为科学研究提供了能够产生准确预测的个人决策模型，但这些模型通常不适用于大型数据集；它们用于通过决策实验获得的少量数据。理解行为科学模型如何能够丰富现有的机器学习，以改善城市服务的个性化——并通过这样做提高公民满意度——是拟议研究计划的重点。我们的项目将通过以下方式解决城市福利问题：一种创新的数据分析方法，行为机器学习（BML）；由于个性化的提高而制定新的城市政策；以及在日常决策中与数据（包括公民自己生成的数据以及企业和政策制定者的公民数据）的交互。拟议的研究涉及从现场实验收集的大型数据集以及公开可用的数据，前者特别侧重于捕获个人公民特征和福祉的复杂和大型数据集。我们的课程具有创新性，因为它：(a)拓宽并整合了行为科学、数据分析、计算机科学和人类数据交互（HDI）方面的研究；(b)研究复杂数据驱动的城市决策的决策设计，这些决策涉及低信息量的数据集、难以管理的超大数据集和噪声数据；(c)关注公民、企业和政策制定者如何使用这些数据。它将在几个方面产生广泛的影响。首先，我们将直接与多个利益相关者合作，制定具有实际意义的解决方案，在实践中制定具有可衡量效益的新城市政策。其次，其研究成果将建议改进参与式模式，以更广泛地分析和呈现数字经济中不同类型的数据。它的第三个价值是方法论，因为我们为社会科学和自然科学研究提供了一个密切合作和整合的模型。由于这个项目，定性和定量研究人员将更好地了解对方的方法的局限性和可能性，导致更好和更适用的跨学科研究。最后一个影响是科学通识教育：社会科学和自然科学的融合丰富和激励了所有年龄段的学生，以及普通公众。

Recent advances in information and communication technology (ICT) have created opportunities for the public sector to increase citizens' wellbeing by better understanding their preferences. Machine learning, a subfield of computer science, is a valuable tool to understand consumer/user preferences, including those relevant to urban development and urban service provision. Indeed, it can help to create new urban value propositions by applying data-driven approaches to urban business modelling. Machine learning algorithms in the urban domain provide valuable insights to policymakers. While the existing methodology produces models that can explain some data well, it can be improved by incorporating mathematical modelling from Decision and Behavioural Science to better capture the behavioural component of urban wellbeing. Research in behavioural science offers models of individual decision-making that generate accurate predictions, and yet these models are not usually applied to large datasets; they are used on small amounts of data obtained through decision-making experiments. Understanding how behavioural science models can enrich existing machine learning to improve the personalisation of urban services - and by doing so increase citizen satisfaction - is the focus of the proposed research programme.Our programme will address urban wellbeing through: an innovative approach to data analytics, Behavioural Machine Learning (BML); the creation of new urban policies as a result of improved personalisation; and the interaction with data (including self-generated data by citizens as well as citizen data by businesses and policymakers) in day-to-day decision-making. The proposed research involves large datasets collected from field experiments as well as publicly available data, the former particularly focused on the complex and large datasets that capture individual citizen characteristics and wellbeing.Our programme is innovative because it: (a) broadens and integrates research in behavioural science, data analytics, computer science, and human-data interaction (HDI); (b) examines decision design for complex data-driven urban decisions that involve datasets with low informativeness, very large datasets which are difficult to manage, and noisy data; and (c) focuses on how the data is used by citizens, businesses, and policymakers.It will have a broad impact in several ways. Firstly, we will work directly with multiple stakeholders to generate solutions that have practical implications for creating new urban policies in practice with measurable benefits. Secondly, its research outputs will suggest improved participatory modes for analysis and presentation of different types of data in the digital economy more generally. Its third value is methodological, as we offer a model of close collaboration and integration for social and natural sciences research. As a result of this project, qualitative and quantitative researchers will better understand the limits and possibilities of the other's methodology, leading to better and more applicable interdisciplinary research. A final impact is on general education in science: blending social and natural science enriches and motivates students of all ages, as well as the members of the general public.