Experimentation as governance: Publics and participation in AI testbeds

One of the most productive contributions of Phan's ‘Testing-in-the-wild’ is its insistence that contemporary AI testbeds cannot be understood independently of longer colonial histories of experimentation. By tracing how Australia has repeatedly been positioned as a site for testing, from penal management and medical science to nuclear weapons and drone delivery, the article demonstrates that experimentation is not just a technical practice but a political relation structured by asymmetries of power. The article, therefore, makes an important intervention into emerging scholarship on AI testbeds by showing that the treatment of populations and territories as experimental subjects predates contemporary digital infrastructures and is deeply entangled with colonial projects.

In this commentary, I build on and expand this argument in two ways. First, I suggest that testbeds are not only sites where technologies are trialled but can also be sites where publics emerge. Public concerns about emerging technologies, such as for instance concerns relating to drones and the data and media assemblages through which they appear, may lead to debates about desirable technology futures. Second, I argue that the proliferation of experimental forms of governance raises questions about participation, value creation and the extraction of knowledge and labour from the populations involved.

The article's emphasis on colonial continuities raises a further question: How do experimental subjects themselves experience and respond to these processes of testing? While the article carefully analyses how governments, corporations and regulators construct Australia as a desirable testing ground, the people who inhabit these testbeds primarily appear as populations to be governed, represented or enrolled. Less attention is given to how they themselves make sense of emerging technologies and the forms of experimentation in which they are implicated.

This question emerged strongly in our own research on public responses to civilian drones in Denmark (Waltorp and Bruun, 2022). Rather than examining drones as objects of innovation policy, we approached them as matters of concern that generate public reflection, anxiety and critique. We found that people rarely encountered drones as isolated technologies. Instead, they immediately connected drones to cameras, data infrastructures, image circulation and surveillance. Drones became part of what we termed a ‘drone assemblage’, activating concerns about privacy, visibility and exposure. Significantly, these concerns were highly gendered. Both men and women frequently imagined drones in relation to existing vulnerabilities associated with female bodies being watched and photographed and having images circulated beyond their control.

Bringing these perspectives into dialogue suggests that testbeds are not just territories in which technologies are trialled but also sites in which publics emerge (cf. Marres, 2005; Marres and Lezaun, 2011). Technologies generate concerns, anxieties, attachments and forms of political engagement that cannot be reduced to the intentions of innovators or regulators, and these concerns can spark public engagement and political participation. Indeed, the Wing Aviation case discussed in the article hints at precisely this dynamic when local residents object to drone noise and surveillance. Yet, these reactions appear mainly as obstacles or feedback within the company's product improvement process. What disappears from view is the possibility that such responses may constitute democratic publics and forms of public dialogue about desirable technology futures. Such publics may not reject technological innovation outright but they can articulate alternative values, priorities and visions of social life that challenge the assumptions embedded in innovation agendas.

At the time of writing this commentary, drones have acquired a very different public presence than when we conducted our study of civilian drones in Denmark in 2017. Then, privacy, surveillance and uncertainty emerged as key concerns (Bajde et al., 2017; Waltorp and Bruun, 2022). Today, after Ukraine, Gaza and the proliferation of military drone warfare, drones have acquired additional and much darker associations, and they increasingly appear as a military technology. Across the world, ongoing wars and geopolitical tensions have transformed public imaginaries of what drones are and what they do. While many of these drones are intended for military installations and infrastructure and not civilian targets, the possibility that drones may carry not just cameras but explosives and can move unpredictably across borders has introduced new and very real concerns about security, vulnerability and the proximity of war. At the same time, military planners have been forced to confront how relatively cheap and low-tech drones can challenge defence systems built around far more expensive technologies.

From an anthropological perspective, these developments underscore a point that emerged in our own research: The significance of drones lies not in the aircraft alone but in the broader socio-technical assemblages they are part of and evoke. Today, they are increasingly associated with military logistics, geopolitical conflict, cyber operations, propaganda and disinformation campaigns. Media manipulations make it clear that drones are not merely flying devices but elements within broader infrastructures of communication and political influence.

This changing drone assemblage complicates some of the concerns explored in our earlier work. Fears of surveillance, unwanted visibility and the circulation of intimate images remain relevant in many contexts, but they now coexist with anxieties about weaponisation, military escalation, infrastructural vulnerability and civilian casualties. The drone is a vivid illustration of how technologies are continuously re-signified through changing political circumstances. What appears as a matter of concern at one historical moment may be displaced, amplified or transformed as technologies become enrolled in new projects, conflicts and futures.

Another thought-provoking aspect of ‘Testing-in-the-wild’ is its demonstration that testing itself has become a mode of governing territories, populations and futures. Reading the article, I was struck by how familiar the language of experimentation has become across many domains of contemporary development and governance. During the past two decades, experimentation has emerged as a key concept and widespread practice in a multitude of fields ranging from globalised medical trials in medical science (Petryna, 2009; Kelly, 2012), urban planning (Marvin et al., 2018; Evans et al., 2016) and energy infrastructures (Levenda, 2019) to climate adaptation (Brandt and Bruun, 2023). Municipalities establish living labs, governments fund pilot projects and public authorities increasingly describe themselves as ‘learning organisations’ that test solutions before ‘scaling’ them. Much more than a scientific method, experimentation has become a way of organising action under conditions of uncertainty.

Marres and Stark (2020) push this argument one step further. For them, the significance of contemporary experimentation is not simply that tests have moved from laboratories into real-world settings. Rather, social environments themselves are increasingly organised as sites of testing. In an era of digital platforms, smart cities and real-time data collection, testing no longer takes place within a social context; instead, the social becomes the object and medium of experimentation. Streets, neighbourhoods and online platforms are deliberately modified to generate feedback, behavioural responses and new forms of knowledge. As a consequence, experimentation increasingly operates as a mode of governance that acts on social life through the continuous adjustment of environments, infrastructures and relations.

This development is also visible in urban governance and climate adaptation, where experimentation promises a way to demonstrate agency and openness to public participation while simultaneously generating knowledge about the very worlds it seeks to transform. In ethnographic research on the Danish city of Vejle, Brandt and I observed how municipal actors increasingly turned to experiments, prototypes and temporary interventions in response to uncertain climate futures (Brandt and Bruun, 2023). In addition to attempting to predict and control future developments through long-term planning, public authorities sought to create spaces in which solutions could be tested, evaluated and adjusted. In this context, experimentation, including citizens’ participation in numerous projects and tests, for instance, of apps, functioned as a governing rationality. In our research from Denmark, we reported how participation and experimentation could also lead to certain forms of ‘fatigue’ (Bruun, 2024). When citizens were repeatedly invited to contribute ideas, experiences and local knowledge, but received little feedback and saw few tangible consequences of their involvement, they experienced participation as extractive and tiring rather than empowering.

In public-sector experiments, participation is often justified through reference to public value, democratic involvement or co-creation (Bruun, 2024). Citizens are invited to contribute local knowledge, test solutions and help shape collective futures. Such ambitions are not always realised in practice but they nevertheless establish a normative expectation that experimentation should benefit the wider public. This observation resonates with debates in anthropology and science and technology studies concerning participation and co-production. Participation is often presented as an inherently democratic good (Kelty, 2019). Participation, however, often generates value through the extraction of citizens’ labour, data and situated knowledge, while the benefits of experimentation are unevenly distributed. In digital environments, users routinely provide feedback, behavioural data and forms of unpaid labour that contribute to political governance or technological development. The drone testbeds described in ‘Testing-in-the-wild’ suggest that such dynamics are also at work in contemporary experimental innovation. Citizens are invited, or in this case simply exposed, to experimentation, while the resulting knowledge is captured within proprietary innovation processes.

Phan's article may also point to a broader shift in governance that is characteristic of contemporary digital societies. Decision-making and coordination increasingly take place through privately developed digital infrastructures, data platforms and AI systems rather than through public institutions organised around political deliberation and democratic participation. This is particularly evident in software platforms that aggregate and analyse large volumes of data across domains such as healthcare, welfare administration, urban governance, policing, border control and intelligence work. While such platforms may initially have been introduced as tools to support decision-making and extract greater public value from data resources, they increasingly shape how problems are defined, which patterns become visible and what forms of intervention appear appropriate (see, e.g., Hoeyer 2023 on data systems in healthcare). Governance of public goods, critical infrastructures and population security thereby becomes dependent on proprietary technical systems whose operation, assumptions and algorithms often remain inaccessible to the publics affected by them and beyond democratic control.

Moreover, Phan's analysis points to another important feature of contemporary experimentation: The fact that those enrolled in testbeds are often not randomly selected populations. Rather, experimental interventions have a tendency to target groups and territories that are already marginalised, vulnerable or otherwise subject to intensified forms of governance. In studies of urban planning and climate adaptation, social housing areas and vulnerable neighbourhoods are frequently designated as living labs and spaces for experimentation (Marvin et al., 2018; Bruun, 2024). Similarly, contemporary AI systems are often developed and tested in domains concerned with welfare provision, fraud detection, mental health and other forms of social intervention directed at populations already positioned as at risk or in need of management. In the Dutch childcare benefits scandal, algorithmic systems contributed to the wrongful accusation of at least 35,000 parents, disproportionately affecting families with backgrounds in Suriname and the former Dutch Caribbean colonies (Arts and van den Berg, 2025). Amnesty International (2024) has noted that new forms of algorithmic governance are introduced in contexts involving socially disadvantaged populations, making them subjects of surveillance and discrimination. In this sense, contemporary AI testbeds reproduce longer histories in which marginalised populations are not only affected by AI systems through biases in the foundational data sets (Benjamin, 2019) but also serve as sites through which AI systems are developed, tested and refined.

This extends the significance of Phan's colonial analysis. The article demonstrates how histories of colonial experimentation continue to shape which populations become available for testing and which territories are rendered suitable for innovation. The existence of contemporary AI testbeds raises questions about the political economy of experimentation itself. If experimentation increasingly functions as a governing logic, we must ask not only who is subjected to experiments but also who benefits from them: Who owns the knowledge generated through experimentation and trialling? Who captures the value that emerges from citizens’ participation? And what obligations do governments and technology companies have towards the populations that contribute, willingly or otherwise, to product development? Phan encourages us to ask precisely these questions and reminds us that experimentation is never merely technical but deeply political.