Towards the collaborative development of machine learning techniques in planning support systems – a Sydney example. (March 2021)

• Record Type: Journal Article
• Authors: Lock, Oliver; Bain, Michael; Pettit, Christopher
• Is Part Of: Environment & planning. Volume 48:Number 3(2021)
• Page Start: 484
• Page End: 502
• Publication Date: 2021-03
• Subjects: Machine learning -- urban development -- planning support systems -- co-design -- mixed methods; City planning -- Periodicals; Urban ecology (Sociology) -- Periodicals; 307.11605
• DOI: 10.1177/2399808320939974
• Languages: English
• ISSNs: 2399-8083

Abstract

The rise of the term 'big data' has contributed to recent advances in computational analysis techniques, such as machine learning and more broadly, artificial intelligence, which can extract patterns from large, multi-dimensional datasets. In the field of urban planning, it is pertinent to understand both how such techniques can advance our understanding of cities, and how they can be embedded within transparent and effective digital planning tools, known as planning support systems. This research specifically focuses on two related contributions. First, it investigates the role of planning support systems in supporting a participatory data analytics approach through an iterative process of developing and evaluating a planning support system environment. Second, it investigates how specifically machine learning planning support systems can be co-designed by built environment practitioners and stakeholders in this environment to solve a real planning issue in Sydney, Australia. This paper presents the results of applied research undertaken through the design and implementation of four workshops, involving 57 participants who were involved in a co-design process. The research follows a mixed-methods approach, studying a wide array of measures related to participatory analytics, task load, perceived added value, recordings and observations. The results highlight recommendations regarding the design and evaluation of planning support system environments for co-design and their coupling with machine learning techniques. It was found that consistency and transparency are highly valued and central to the design of a planning support system in this context. General attitudes towards machine learning and artificial intelligence as techniques for planners and developers were positive, as they were seen as both potentially transformative but also as simply another technique to assist with workflows. Some conceptual challenges were encountered driven by practitioners' simultaneous need for concrete scenarios for accurate predictions, paired with a desire for predictions to drive the development of these scenarios. Insights from this work can inform future planning support system evaluation and co-design studies, in particular those aiming to support democracy enhancement, greater inclusion and more efficient resource allocation through a participatory analytics approach.