Urban AI as a sociotechnical system: Ecosystem factors for responsible innovation

Municipalities are increasingly turning to artificial intelligence to optimize mobility, automate administrative processes, anticipate climate risks and improve services to citizens. Yet AI is a profoundly ambivalent and disruptive technology, and current research rarely examines it as a sociotechnical system shaped by multiscalar ecosystem factors. Empirical analyses remain limited on how actor dynamics, power relations, resources and multilevel institutional frameworks shape the development, deployment and regulation of urban AI. The risk is a reinforced technocentric governance, marked by technological solutionism and growing municipal dependence on external actors, and an uneven development of territories in which only the best-equipped cities benefit from AI.

This dissertation argues that municipalities cannot seize the opportunities of AI without orchestrating the innovation ecosystems in which they operate. It analyzes the dynamics that structure the use and regulation of urban AI by municipalities within these ecosystems, with the ultimate goal of helping cities harness the benefits of AI while limiting its risks.

The research asks how the innovation ecosystems surrounding municipalities shape their urban AI practices, through four questions concerning: (1) how municipal actors’ visions and representations of AI structure the deployment and regulation of urban AI systems; (2) the relationship between actor interdependencies and the shape of the innovation ecosystem; (3) how the structure of the ecosystem influences deployment and regulation practices; and (4) which approach can enable municipalities to govern urban AI responsibly given the ecosystem that hosts them.

Theoretically, the dissertation approaches urban AI as a sociotechnical system by articulating three concepts: urban imaginaries, innovation ecosystems and responsible innovation. Empirically, it adopts an exploratory mixed-methods design centred on the case of Quebec, drawing on 825 press articles, 20 institutional documents, 26 semi-structured interviews with ecosystem actors, a network analysis of actors and their interdependencies, and a multi-stakeholder participatory workshop.

The findings are presented in four articles. The first shows that the imaginary of the calculable city, centred on efficiency, has become the dominant representation of the city in the age of AI, and that it also structures the imaginaries of the resilient and inclusive city. The second maps Quebec’s urban AI ecosystem as fragmented yet highly mediated: municipalities remain the central holders of demand but depend heavily on technology firms and non-profit intermediary organizations, with technological and cognitive interdependencies concentrating power in the actors who control expertise, coordination and knowledge flows. The third shows that municipal AI adoption is strongly shaped by structural power asymmetries, a partial outsourcing of governance and intermediation mechanisms dominated by technology companies; while flexible coordination supports experimentation and learning, the absence of operational frameworks and control mechanisms heightens the risk of technocentric governance. The fourth, based on a participatory workshop with twelve ecosystem actors, proposes a multi-orchestration model built on four complementary layers (technological, economic, institutional and behavioural), relying on mechanisms such as cost pooling, shared infrastructure and AI clinics to promote transparency, accountability, data security, territorial inclusion and the technological autonomy of municipalities.

The dissertation makes three main contributions to the literature on urban AI governance. First, it shifts the unit of analysis, showing that urban AI governance cannot be understood through AI systems, municipalities or projects alone, but must encompass the relationships and interdependencies of innovation ecosystems. Second, it introduces the concept of latent technocentrism, showing that technocentric governance can emerge implicitly from imbalanced imaginaries, power asymmetries and interdependencies within innovation ecosystems, even in the absence of explicit technology-driven agendas. Third, it proposes a new approach to responsible governance based on the participatory orchestration of innovation ecosystems.

Keywords: urban artificial intelligence; urban imaginaries; innovation ecosystem; responsible innovation; urban governance; latent technocentrism; ecosystem orchestration.