Putting humans back in the loop of machine learning in Canadian smart cities

Abstract

Even as researchers recognize smart cities as sociotechnological assemblages, the attractiveness of artificial intelligence and machine learning (AI/ML) continues to drive cities towards automating urban process and analysis. Rather than arguing whether researchers should automate their analysis, we are interested in identifying where non-automated, manual judgement calls are made and the analysis is more subjective than presented. We examine topic modelling, an ML method, in an analysis of applications to a pan-Canadian smart cities grant competition. We document 11 steps in topic modelling from data collection to interpretation. At each step, including the choice of the topic modelling method, some degree of human intervention is required. We draw on human-centred ML research to argue for a greater recognition of the role of humans-as-researchers to preclude further uncritical adoption of AI/ML to research smart cities. | INTRODUC TI ONSmart cities, through their sensors and software dashboards, produce enormous volumes, velocities and varieties of real-time streaming data. Such data come in large part from volunteered geographic information (VGI), whether that is from mobile devices as urban residents traverse space-time, or are made available from private services like Uber. We have created new methods (e.g., predictive analytics) and new sciences (e.g., urban science, data science) to take advantage of these data as a window into the city. While artificial intelligence and machine learning (AI/ ML) algorithms can analyse the vast amounts of data in the smart city, the methods also reinforce approaches that reify automation, not only in the management but also in the understanding of the city. This aligns with the long-standing association of "smartness" with the assumption that technology can improve urban functioning, despite the gradual transition from a technologically deterministic paradigm towards urban sustainability and well-being (Albino, Berardi, & Dangelico, 2015;Batty, 2013; Goodspeed, 2015). Municipalities continue to credit technologies with the ability to transform cities into places of high efficiency and innovation (Marshall, 2021). Kandt and Batty (2021) address the epistemological problems of urban data analytics and, by extension smart cities, both of which "can be automated with little human intervention" (Kandt & Batty, 2021, p. 1). They argue that it is doubtful whether big data can provide us with a general and in-depth understanding of urban problems. Big data solve problems on the surface; not why they happen. Urban data analytics offers a simple solution to urban problems, which have long been called wicked problems (Rittel & Webber, 1973). Despite a desire by some researchers and practitioners for straightforward answers, a wicked problem cannot be reduced to a single cause; the problem is incomplete and conflicting in understanding, and lacking a straightforward presentation. Instead of treating immediate symptoms of city problems, we should try to qualitatively understand why cer...

Investigating the Key Aspects of a Smart City through Topic Modeling and Thematic Analysis

“…Their study aimed to investigate the polysemic nature of the smart city concept and underscore the diversity of the prospects presented by smart city policies. Zheng et al [24] contemplated an instance of utilizing the LDA topic modeling technique on 137 smart city proposals submitted to the Government of Canada's Smart Cities Challenge (SCC) that was operational from 2017 to 2019. Similarly, proposals submitted for the "Inclusive Smart City" project were investigated in [5] through the application of topic modeling techniques.…”

Symbolic and subsymbolicGeoAI: Geospatial knowledge graphs and spatially explicit machine learning

“…(3) Given the increasing popularity of foundation models (FMs) in the natural language and vision communities, such as GPT-3 (Brown et al, 2020), CLIP (Radford et al, 2021), PaLM (Wei et al, 2022), and DALL•E2 (Ramesh et al, 2022), could we build a FM for GeoAI which, after pretraining, can be easily adapted to multiple symbolic GeoAI and subsymbolic GeoAI tasks, involving the use of different data modalities (Mai et al, 2022a)? (4) How can we address important ethical aspects, such as better accounting for and mitigating issues of bias, fairness, and transparency (Shin & Basiri, 2022;Zheng & Sieber, 2022), how to reduce the environmental footprint of model training, and how to better connect to communities studying ethics of technology (Goodchild et al, 2022).…”