States of computing: On government organization and artificial intelligence in Canada

Abstract

With technologies like machine learning and data analytics being deployed as privileged means to improve how contemporary bureaucracies work, many governments around the world have turned to artificial intelligence as a tool of statecraft. In that context, our paper uses Canada as a critical case to investigate the relationship between ideals of good government and good technology. We do so through not one, but two Trudeaus—celebrity Prime Minister Justin Trudeau (2015—…) and his equally famous father, former Prime Minister Pierre Elliott Trudeau (1968–1979, 1980–1984). Both shared a similar interest in new ideas and practices of both intelligent government and artificial intelligence. Influenced by Marshall McLuhan and his media theory, Pierre Elliott Trudeau deployed new communication technologies to restore centralized control in an otherwise decentralized state. Partly successful, he left his son with an informationally inclined political legacy, which decades later animated Justin Trudeau's own turn toward Big Data and artificial intelligence. Compared with one another, these two visions for both government and artificial intelligence illustrate the broader tensions between cybernetic and neoliberal approaches to government, which inform how new technologies are conceived of, and adopted, as political ones. As this article argues, Canada offers a paradigmatic case for how artificial intelligence is as much shaped by theories of government as by investments and innovations in computing research, which together delimit the contours of intelligence by defining which technical systems, people, and organizations come to be recognized as its privileged bearers.

Canadian AI: A tale of two Trudeaus

Today's announcement is an important step toward making Canada an export and AI leader. Our government launched an ambitious strategy to keep this sector competitive in the long term and support Canada's innovators. Together, we can build an innovation economy—one that benefits everyone and provides new opportunities for the middle class to succeed.

—The Rt. Hon. Justin Trudeau, Prime Minister of Canada (2018)

These comments by Prime Minister Justin Trudeau (JT) followed yet another major federal investment in artificial intelligence (AI). Awarding $230 million to fund “AI-Powered Supply Chains,” this announcement represented a key milestone in the Canadian federal government's sustained efforts to position Canada as a world leader in AI. Like its American, Chinese, French, and German counterparts, to name a few, the Canadian government promoted a vision of AI as a revolutionary technology that each country should embrace to ensure its future prosperity (see Bareis and Katzenbach, 2021). And like its counterparts, the Canadian government promoted AI as ultimately inevitable—a highly performative vision paired with substantial federal grants to AI research, massive subsidies to AI-centric firms, and ambitious programs exploring the application of AI to government processes. As the latter programs translated into major changes in the federal government's procurement policies and new guidelines on the use of automated decision-making, AI came to be conceived of not only as an economic opportunity, but also as an agent of revitalization for the government itself.

However new or revolutionary AI might be presented, this dual articulation has numerous precedents in the history of the Government of Canada and, more specifically, in the premiership of JT's father, Pierre Elliott Trudeau (PET). During his tenure as prime minister (1968−1979, 1980−1984), PET promoted a similar conceptualization of technology as a transformative force for both Canada and its government. Calling for “political instruments which are sharper, stronger, and more finely controlled than anything based on mere emotionalism” (Trudeau, 1968a: 203), PET embraced cybernetics, systems theory, and broadcasting as powerful tools to not only maintain national unity, but also restore government control in a highly decentralized state. Concerned with the development of an approach to government that would be more “realist” (Breton et al., 1964), “rational” (Trudeau, 1968a), and “coolly intelligent” (Trudeau, 1996 [1950]), as he alternately termed it, PET articulated a vision of intelligent government centered on the adoption of government structures, media, and information technologies as the main determinants of government. From one Trudeau to the next, artificial intelligence emerged, and remained, an enduring site of political investment and speculation, where the prospect of machine intelligence, centralized decision-making, and, later, Big Data came to be raised alongside that of improved government practices, motivating a wide range of targeted investments and government reforms under both administrations.

Our article works at the friction of AI being constructed as a revolutionary technology twice, by two Trudeaus. Instead of commenting on the leadership or personal vision of either figure, we approach the two Trudeaus as conceptual entry points into AI's making as a political technology. In that regard, we conceive of them more as indexes of two techno-social discourses than historical figures. These discourses, we contend, both centered on technical conceptions of intelligence and theories of state intelligence that not only permeated, but also legitimated one another. By representing key milestones in a broader re-organization of the Canadian federal apparatus both through, and on the model of, new technologies, the two Trudeaus readily capture AI's gradual articulation and deployment as an extension of government both embodying and promoting changing ideals and practices of governmentality. In our discussion of their administrations’ reforms and media policy, we then focus on their respective adoption of AI as a “political instrument,” ¹ to borrow PET's terms, with the intention of contributing to the ongoing reappraisal of the social construction of political technologies (see Jasanoff and Kim, 2015; Shapin and Schaffer, 1985; Winner, 1980). Building on the term's usual Foucauldian interpretation, which encompasses all the techniques involved in the organization and deployment of social forces toward specific ends both within and across states (Foucault, 2004: 320), we advance an understanding of political technology as a certain articulation of government as technologically enacted.

With this inquiry, we hope to contribute to the literature on the informational turn in the organization of nation-states and its relationship to the proliferation of technologies like Big Data and artificial intelligence in government settings (see Agar, 2003; Braman, 2006). In our comparative national history, we attend to how the two Trudeaus fostered a broader datafication of the state, which both anticipated and enabled recent innovations in the field of AI. By datafication, we specifically refer to the interplay between new ideas about good government, digital media, and bureaucratic processes, which together encode political affairs into data and produce the necessary infrastructure to process them (Mejias and Couldry, 2019). If contemporary states appear so receptive to the latest innovations in information processing, we argue that this might be because they themselves have been previously re-configured as information processing systems. Therefore, we call for a broader engagement with the historical construction of government itself as a form of artificial intelligence.

Making government intelligent: Global histories of political technologies

By taking the frame of the two Trudeaus' articulation and deployment of political technologies, we gain a privileged entry point into the disjuncture and politics of this informational turn in government. In recent years, several scholars have attended to the influence of postwar ideas about systems, machine intelligence, and information on government organization. In conversation with previous works on the transformative effect of cybernetic ideas on Soviet and American social science (e.g., Gerovitch, 2002; Kline, 2015; Solovey, 2012), these scholars addressed how the notion of “systems” was adopted as a powerful category to not only model entities like “population,” “the economy,” or “labor forces,” but also develop policies and mechanisms designed to manage and control them. While Hunter Crowther-Heyck highlighted how techno-cultural ideas about systems and organization came to “inform and validate policy decisions both large and small” (2015: 8) in Cold War United States, Eglė Rindzevičiūtė for her part more broadly documented the transnational adoption of systems theory as “a source of avant-garde ideas on governability” (Rindzevičiūtė, 2016:6). Covering national settings as diverse as China (Liu, 2019), the United Kingdom (Pickering, 2010), and Chile (Medina, 2011), this literature highlighted the gradual reconceptualization of government in terms of structures and organization from the 1950s onward, with the optimal management of complex systems being recognized as the ultimate end of government. If this assessment still resonates today, it is partly because these theoretical lenses and disciplinary reforms continue to animate contemporary ideas about data-driven government and AI (see Peters, 2016: 17–24). In fact, a certain optimism for optimization still endures today, with AI being regularly presented as uniquely positioned to manage complex social systems in a more sustainable, equitable way (McKelvey and Neves, 2021; Srnicek and Williams, 2016). The fact that so many of these discourses draw from the systems thinking literature of the 1970s hints at a certain nostalgia for cybernetic ideas about state management—a nostalgia which, as we find, sheds light on the many competing logics animating contemporary AI.

While some strands of AI symbolize a potential cybernetic state, other interpretations of this political technology promote opposing managerial ideas inherited from the re-organization of government practices on a neoliberal model and a digital turn in the field of economics (Beverungen, 2019; Mirowski and Nik-Khah, 2017). As Quinn Slobodian highlights, global neoliberal theory not only responded to the influence of cybernetic ideas on government, but also actively sought to counter them (Slobodian, 2018: 218–262). Making its way into government settings in the form of New Public Management and other more “business-like” models, this shift foreclosed the ambitions of cybernetic government and came in pair with “the great moving right show” (Hall, 1983) which, in the United States at least, prompted a direct rebuke of systems theory by the likes of President Ronald Reagan and an embrace of more economic-tinged theories of information. Far from being anti-statist or technologically averse, this neoliberal model encouraged a different social construction of political technology—one that centered on the meta-conditions of social, economic, and political life, with the aim of creating the necessary conditions for the development of markets and industries aligned with government goals.

Bringing these bodies of work together, our article attends to the overlaps between these two social constructions of both government and machine intelligence by comparing the discourses, policies, and overall philosophies promoted by the two Trudeaus and their administrations. As a world leader in AI research that has been mostly overlooked by the literature, ² Canada provides a rich case study to explore how Cold War ideas about government organization shaped the development of pan-national AI policies both in Canada and abroad. In the following sections, we specifically examine PET's articulation of a specific strand of systems thinking informed by a highly structural approach to, and expansive understanding of, media, alongside JT's adoption of a highly quantitative, data- and impact-driven model of government that promoted AI as the foundation for a new Canadian national project. Viewed together, these two moments highlight how ideas about intelligent government and machine intelligence operate at a similar intersection—namely, that between political imaginaries and new technologies—and as such provide the backdrop upon which new technologies get mobilized as political ones. At its core, our article then not only explores the historical interplay between ideas about intelligent government and machine intelligence but also documents how this interplay both preceded and enabled AI's ongoing promotion as a tool of government.

Government in the global village: A media history of Canadian federalism

Away from neural nets and servomechanisms, computers and black boxes, the Canadian state’s experiments in cybernetic management and new political technologies began with a somewhat more prosaic project—that of broadcasting regulation. With Québec nationalism, Indigenous claims for sovereignty, and an increasingly diverse cultural “mosaic” ³ eroding Canada’s self-image as a white, English-speaking, protestant nation, several politicians, federal administrators, and intellectuals all strived from the 1950s onward to identify how best to govern a highly decentralized state with a fractured national identity. In the wake of the Royal Commission on Radio Broadcasting, whose 1929 report identified mass media as “a great force in fostering a national spirit […] in a country of vast geographical dimensions” (Aird, 1929: 6), many of these figures singled out the creation of a unified national identity as a pressing object of state intervention. Amidst settler-colonial anxieties about nationhood, land ownership, and self-determination (see Cornellier, 2013; Coulthard, 2007), Canadian identity was recast by this Commission and others into something which, while not an object of “official responsibility […] a generation ago” (Massey, 1951: 5), required active government involvement to ensure the proper exercise of government itself. ⁴

This perceived responsibility to cultivate a shared “national spirit” contributed to the appointment of another royal commission in 1955—one specifically dedicated to the use of television to promote national cohesion. Like the railway before it (see Charland, 1986), television was seen as uniquely positioned to both connect Canada's vast territory and control the circulation of information across it. Echoing a previous memorandum by renowned scholar Harold Innis, ⁵ whose theory of communication posited that each medium “implies a bias in the cultural development of civilization” (Innis, 1950: 216), the commission recast communication technologies as privileged tools to restore federal oversight both within, and at times beyond, the Canadian government's usual legislative powers. Headed by paper industrialist Robert Fowler, the Royal Commission on Broadcasting emphasized the “unifying and cohesive functions” (Fowler, 1957: 3–6) of television and as such defended its regulation as “a legitimate role of government in Canada” (1957: 7–13). Advocating for its use to connect the country's vast territory and diverse population, the Commission mobilized media regulation as a privileged means to not only engineer national unity, but also restore the government's capacity to exercise centralized control over the circulation of information and, by extension, the nation as a whole. ⁶

Beyond questions of broadcasting regulation, the Fowler commission was then also representative of a broader concern with the organization and role of the federal government, in a way which echoed many of the reforms that were being called for at the time. By the 1960s, the federal bureaucratic machine had grown so large and disorderly that Conservative Prime Minister John Diefenbaker (1957–1963) appointed yet another commission—the Royal Commission on Government Organization—with the mandate of investigating “the “how” rather than the “why” of government” (Glassco, 1961: 2). In its final report, the Commission advocated for doing away with the personalized managerial style of the civil service. In its place, it promoted the standardization of government processes and centralization of all decision-making as a means of not only improving the circulation of information but also increasing the government's capacity to act upon it.

By the time the Commission submitted its last report, however, the reforms it promoted had already been relegated to the second rank of government priorities. Elected in 1963, the Liberal government of Lester B. Pearson (1963−1968), though sensitive to the inefficiencies the Commission had documented, remained committed to the informal, deliberative ethos of the Canadian civil service. As such, despite the many high-profile policies he introduced, ⁷ Pearson did little to appease the reformist zeal of the federal government's critics, whose ranks had come to include labor lawyer and Québec intellectual Pierre Elliott Trudeau.

From the 1950s onward, PET had become known as a leading figure of a broader movement for a “coolly intelligent” approach to government (Trudeau, 1996 [1950]: 28). Between calls for provincial autonomy and appeals for federal centralization, PET articulated a broader reconsideration of government in organizational terms, which recentered the political process on government's concrete structures and practice. While initially aimed at the conservative, Catholic Church-backed government of Québec Premier Maurice Duplessis, PET's reformism gradually took on the federal government as its main target by attributing the growing atomization of Canadian society to the latter's “great difficulty and little success […] in coordinating [its] own activities” (Breton et al., 1964: 32). Along with other Québec intellectuals, PET co-signed in 1964 a “realist manifesto” which accused the federal government to have done little to justify its existence, effectively leaving a political void that competing claims for sovereignty attempted to fill. For him, Canada's disunity was thus first and foremost the product of a certain ad hoc approach to government, which could be remedied by deploying the latest ideas and devices of the new sciences of communication. Predicating “democratic progress” on “the ready availability of true and complete information” (1964: 30), PET and his peers singled out “information media” as key to the creation of a new type of union anchored in “reason and […] objective analyses of reality” (1964: 29) instead of politicians’ whims and caprices. Faced with the risk of “the utter disintegration of the Federal State” (1964: 29), Trudeau and his co-signatories then not only promoted a re-organization of the federal government on a highly centralized, formalized model, but also conveyed a profound reconceptualization of media as an extension of the government's apparatus.

Despite its resonance with the new sciences of systems in vogue at the time, ⁸ Trudeau's organizational awareness found its closest conceptual ally in the Toronto school of communication. The Toronto school shared cybernetics’ attention to the transformative role of communication but advanced a more ontological yet actionable approach that contrasted with the abstract, mathematically inclined one of its American counterpart. ⁹ The same year Trudeau transitioned to a political career and joined the Liberal Party, Marshall McLuhan published his now canonical Understanding Media, in which he posited a similar entanglement between the organization of states and the media through which they operate. Calling for a “configurational awareness […] of the action of media” (1964: 10), McLuhan redefined media as a “man-made, social environment” (1964: 98) and recentered their study on their organizational effects. With McLuhan advising private and public entities alike on how best navigate these effects, it is not only an awareness of media's organizational underpinnings which McLuhan promoted but also the idea that such effects could be harnessed to maintain control in the face of rapid technological change. While McLuhan and Trudeau would later comment in their correspondence on the clear similarities between their respective visions, ¹⁰ it is this latter prospect—that of mobilizing media and organization as a means of control—that first united both figures. In June 1964, McLuhan and Trudeau met for the first time at the annual meeting of the Canadian Political Science Association in Charlottetown, Prince Edward Island, ¹¹ where Trudeau presented a paper entitled “Federalism, Nationalism and Reason.” ¹² Later described by McLuhan as “at the Edmund Burke level” (McLuhan, 1968b: 1), the paper echoed McLuhan's take on the entanglement between the organization of states and that of media by advancing vision of government as undergoing the same “reassembling of all [its] mechanized bits into an organic whole” (McLuhan, 1964: 93) as the globe itself.

In his paper, Trudeau most notably presented federalism as the logical outcome of a world undergoing a “cybernetic revolution” (Trudeau, 1968a: 203). As a model of government defined by its dynamic (de)centralization of competencies based on political circumstances, federalism consisted for Trudeau in “a rational compromise between […] divergent interest-groups” (1968a: 195) whose coordination was made possible by the latest communication technologies. In the image of McLuhan's global village, which posited “the utmost discontinuity and diversity in spatial organization” (McLuhan, 1964: 36) produced by electronic media as the new basis of all government, Trudeau's federalism predicated the unity of a decentralized nation like Canada on the deliberate deployment of communication technologies. Reducing all aspects of government to communication, Trudeau advocated for the development of “political instruments which are sparer, stronger, and more finely controlled” (Trudeau, 1968a: 203), predicating Canadian unity on the exercise of greater control over the government's own functions and organization. When Trudeau ultimately became the leader of the Liberal Party in 1968, he then did so as the flagbearer of a new federalist vision which mobilized government reforms and media interventions as powerful political technologies capable of artificially holding the country together in the face of both technological and political change.

In a series of letters sent to Trudeau following his landslide election in 1968, McLuhan expressed a similar understanding of the close interplay between media, government reforms, and political technologies by notably singling out television as a privileged means for the federal government to intervene in the organization of the nation itself. Describing Trudeau's “‘cool’ TV power” (McLuhan, 1968b: 1) as effectively holding the country together, McLuhan advised Trudeau to re-organize the whole of the government apparatus around a regular TV program in which the prime minister himself would take political matters directly into “the Canadian living room” (McLuhan, 1968c: 2). As he wrote, this model of a “Government of the air” (McLuhan, 1968c: 2) would allow the Canadian government to alleviate “the [decentralizing] effects of new man-made environments” (McLuhan, 1968a: 1) by using these same media to communicate its vision for the future. ¹³ A few months later, at the 1969 Liberal Policy Conference, Trudeau echoed many of McLuhan's ideas—sometimes word for word—by advancing a vision of the federal government in the image of the latest technologies, as itself an entity in flux. In an inaugural address filled with mentions of “satellite broadcasting,” “genetic engineering,” and “supersonic aircrafts,” Trudeau proclaimed that the new technologies of communication, “by transforming the control function and the manipulation of information,” could “transform our whole society” (Trudeau, 1969: 7). Promoting a vision of the government as “society’s radar” (1969: 4), Trudeau contended that, with the right information, the role of the federal government could be made into that of an active planning entity whose capacity to navigate uncertainty would be improved by the effective communication of its vision. Concluding his speech with the idea that “the image we hold of our future is itself an important element of that future” (1969: 10), Trudeau advanced a view of federalism as ultimately produced by the convergence of planning and communication in the form of collective political control.

If this vision smoothly followed McLuhan's organizational approach to media and politics, it did not sit so well with the conference's delegates. The initial purpose of the conference was to initiate a larger democratization of the Liberal Party, which ironically Trudeau substituted with reforms aimed at tightening the federal government's control over its own organization. Building on the Committee on Priorities and Planning model developed during the later years of the Pearson administration, Trudeau and many of the other signatories of his realist manifesto whom he had just appointed to key Cabinet positions most notably implemented the so-called “Cabinet Committee System.” Borrowing here and there from corporate planning, systems theory, and futurology, this system aimed to centralize all significant decision-making within the highest levels of the federal government (see French, 1984: 41–58). Organized around a handful of permanent committees chaired by Trudeau and his closest appointees, this system aggregated all relevant information about state problems within a single organ, at the expense of the departments whose portfolios were being decided upon. Developed at the same time as, and in close conversation with, the Club of Rome and its experimentations in global modeling, ¹⁴ this system echoed other contemporary attempts to rationalize policy development and reshape government as an information-driven practice. While none of the tools and models developed by the Club of Rome were ever used by the federal government, ¹⁵ they nevertheless provided direct inspiration for its re-organization on a highly centralized, information-centric model—one which framed government processes as mediating how information is not only processed but also acted upon.

Trudeau and his Cabinet's re-organization of the state as an information management system did more to alienate than rally the communities they targeted, however, as these operational reforms became quickly quite political. ¹⁶ In the name of “foster[ing] adequate communication among all people of Indian descent and […] the Canadian community” (Chrétien, 1969: 8–9), the Trudeau government controversially promoted the revocation of Indigenous people's special status, which was immediately met with nation-wide protests by virtually all Indigenous communities. When a Québec separatist cell kidnapped two government officials and demanded that their manifesto be broadcasted on public networks, Trudeau and his Cabinet invoked the War Measures Act to not only authorize the military occupation of Montréal but also tackle “the irresponsibility of the television and radio stations” in broadcasting the kidnappers’ messages (Cabinet Committee on Security and Intelligence, 1970: 1)—a decision which would remain the object of a lasting controversy in the province. ¹⁷

While framed by Trudeau's advisor Michael Pitfield as products of Canadians’ failure to view government “as a total system” (1976: 19), these backlashes testified to what can be ultimately described as the irrational underpinnings of Trudeau's “coolly intelligent” approach to government. In their account of Cold War rationality, Paul Erickson and his colleagues document how the postwar period witnessed “the expansion of the domain of rationality at the expense of that of reason” (Erickson et al., 2013: 2), with standardized procedures and information technologies displacing human judgment, experience, and reason as the main ideals of government. In its attempt to do away with emotionalism, subjectivity, and nationalism altogether, we can recognize in Trudeau's federalism a similar displacement, wherein the practice of government came to be organized around such a keen structural emphasis on organization and communication that the latter two came to be recognized as the main determinants of national unity. When turned into a model for government organization, what Trudeau described as a “coolly intelligent” approach amounted to a highly mediated form of rationality—one manifested by the exercise of control not only through media, communication, and information but also, and more importantly, over them.

Trudeau lost his electoral majority when re-elected in 1972 and failed to implement further reforms of a comparable scale. In fact, all of his government's organizational reforms had come under attack in some way or another by the end of his premiership. ¹⁸ Abroad, this coincided with vast yet similarly unsuccessful attempts by other governments to improve their ability to gather, process, and act on information, as illustrated by the Carter administration's experiments in global modeling. The Global 2000 Report from 1980, which covered the American government's predictions for the upcoming 20 years, received just a passing mention in the Canadian federal literature (Letendre, 1981), even though it employed similar modeling techniques as those PET had championed 10 years earlier at the Club of Rome’s 1971 meeting in Montebello, Québec. Meanwhile, the Canadian Communications Research Centre pursued projects in line with Trudeau and McLuhan's shared vision for the development of an integrated television and information system, notably through the creation of a videotext service called Telidon (see Devon, 1991; Gillies, 1990). And yet, though it was released and marketed as the first public information service in the world, Telidon was never popular and was soon supplanted by some of its foreign counterparts like Minitel (e.g., Mailland and Driscoll, 2017). Much like the Global 2000 Report and the modeling techniques it promoted, Telidon was ultimately abandoned by PET's successors who privatized these technologies throughout the late 1980s.

What would remain from this approach, however, which first emerged in the Canadian government's postwar media policy and culminated in the organizational reforms of the Trudeau administration, was not only a larger reconceptualization of communication as an extension of government, but also a specific way of conceiving of, and regulating, all technology in communicational terms. The 1982−1984 Royal Commission on the Economic Union and Development Prospects, convened during Trudeau's last mandate, notably described the computer as the latest manifestation of innovations in electronic media and even singled out artificial intelligence as an especially promising opportunity for Canada. Described as a “substitute for […] human mental and physical efforts” made possible by “exponential growth in micro-processors and telecommunications” (Macdonald, 1985a: 117), AI was framed as an extension of not only human intelligence, but also of government intelligence given its entanglement with its communication apparatus. Later in its report, the Commission recommended the establishment of “several technology-development centres […] in computer technology and related fields” (Macdonald, 1985b: 145) to support AI research, framing the development of a coast-to-coast pool of expertise in the field as another opportunity to foster economic development and national unity alike. Seen in such terms, AI was first articulated as the latest manifestation of the federal government's postwar experiments in organizational reforms and media regulation, while at the same time raising the prospect of another form of non-human intelligence both in continuity with, and supplementing, that of the government.

Setting the ground for the policies which would come to characterize Canada's current AI agenda and strategy, the Commission framed computing and, more specifically, AI as privileged tools to connect, control, and unify, re-locating to a technological level what PET had attempted to achieve at an organizational one. Years later, when his son—Justin Trudeau—would also run for the premiership, it would be on a platform not so much inspired by that of his father as powered by an ambitious AI agenda and data-centered rhetoric directed at similar political goals. Veering from his father's philosophically minded organizational vision, Justin Trudeau would return to a more ad hoc approach yet anchored in a similar form of political rationality—one of the more reflexive type, this time around.

The technophiles in power: From deliverology to AI patronage

While PET's transition to politics was preceded by years of essay writing and political commentary, Justin Trudeau's literary contribution conversely took place amidst a political career that was already well underway. In October 2014, one year after his successful bid for the Liberal Party leadership, Justin Trudeau did what most aspiring heads of government now do—he published an autobiography. Titled Common Ground, the book narrated the story of a true coast-to-coast Canadian, born to a French-speaking Québécois and an Anglophone Vancouverite, whose love for his country only translated into a political career after years of civic engagement and, admittedly, indecision. Recounting his journey from a prime minister's son born while his father was in office to a high school teacher and youth service advocate, JT presented the image of a true people person more in line with his maternal grand-father—James “Jimmy” Sinclair, former Minister of Fisheries and “the ultimate retail politician” (2014a: 166)—than with his father. Away from PET's philosophically minded approach, JT claimed for himself a simpler strand of federalism centered on a vision of Canada as a social experiment in diversity. “If even a country as respectful of its diversities as ours [fails] to reconcile its differences, what hope [does] the rest of the world have of getting along?” (2014a: 99), Trudeau wondered, advancing a more humanistic rationale for federalism than the one his father previously upheld.

Trudeau's appeal to a more commonsensical approach to federalism contrasted not only with that of his father but also with then-Prime Minister Stephen Harper (2006–2015). As the leader of a newly reformed Conservative Party, Harper balanced an ideological distrust of government with a more decentralized vision of federalism. Against PET's reformation of the federal government into a highly mediated system, Harper's program promoted a re-organization—or disorganization—of the Canadian civil service. During his tenure, Harper strived against decades of centralizing efforts and rejected the federal government's self-appointed role as the country's planning and coordinating body. From his election onward, Harper adopted a model of “open federalism” which combined “renewed respect for the division of powers” with “a strong central government that focuses on […] national defence and the economic union” (Harper, 2004). If his model reaffirmed the rights of provinces to govern outside of federal intervention, it also endowed private interests with the same rights and thus balanced federal withdrawal from socially inclined legislation with increased federal involvement in securing free-market principles across the nation. Moreover, Harper specifically targeted the federal government's capacity to gather, process, and exchange information through his administration's muzzling of government scientists and seclusion of the civil service, testifying to the persistence of PET's organizational approach if only in the systematic suppression of the federal government's information processing capabilities.

As a response, when JT first announced his leadership bid in 2012, he did so with a gesture to his father's legacy. In an announcement speech punctuated with passionate calls for Canadian unity and party renewal, JT emphasized that “[policy] solutions can come from the left or the right; all that matters is that they work” (Trudeau, 2014b: 302). In a pragmatist fashion, JT dismissed such ideological divisions as archaic and promoted in their place “the only ideology that must guide us […]: evidence. Hard scientific facts and data” (2014b: 302). More than a response to the defunding of government research which had been taking place throughout the previous decade, JT's promotion of evidence-based policy combined nostalgia for his father's premiership with a renewed attachment to communication technologies and rhetorical alignment with the quantitative mindset of Big Data.

In the 2015 federal election, Trudeau embraced a data-driven approach to campaigning, trying to capitalize on the optimism from Obama's technologically-intensive campaign. Adopting a mix of “GOTV (get out the vote) techniques, small-gift fundraising, and […] social media” (Trudeau, 2014a: 259), Trudeau presented the image of a highly participative, decentralized campaign, which allegedly mirrored “the new networked nature of modern political movements” (Trudeau, 2014a: 240). As documented by Munroe and Munroe (2018), however, the Trudeau campaign also centered on a predictive approach that specifically targeted likely Liberal converts in key ridings. ¹⁹ Ultimately, JT won a majority government with the second-lowest percentage of the popular vote in Canadian history. Beyond the campaign's evidence-based component, it is then a broader enthusiasm for data and, more specifically, data-enhanced decision-making that the JT campaign both promoted and mobilized—one which Trudeau carried with him into government.

JT’s 2015 electoral win inaugurated a new vision for federalism centered on several threads: an ideological commitment to the Canadian union, a return to a more centralized albeit open form of government, and a rhetorical investment in data. But as the new administration's Information Technology Strategic Plan promptly deplored, the Canadian bureaucracy was still devoid of the necessary digital skills and infrastructure to deliver on the more technical aspects of this vision (Messina, 2016). At their most visible—and advertised—level, these threads then converged in a deliberate embrace of technologies like AI, which translated into initiatives ranging from highly mediatized investments in Canada's AI industry to the promotion of a technological renewal of the government apparatus. Revamping his father's call for a modernized government machine, Trudeau appointed Canada's first Chief Information Officer and Minister of Digital Government, whose mandates consisted in overseeing reforms aimed at updating the government's procurement guidelines, improving the technological literacy of the public service, and digitizing the delivery of public services.

Of these reforms, the creation of the Canadian Digital Service (CDS) figured as one of the most high-profile initiatives of the newly elected administration. ²⁰ Officially part of a broader attempt to renew the public service, the CDS aimed to do away with the Harper government's public management philosophy and kickstart a more digital government-style approach to government operations (see Clarke, 2019: ix–xi). Mandated to improve government agencies’ service delivery capabilities through the integration of digital tools, the CDS promoted within the Canadian civil service new data management practices combined with a more user-centric approach modeled on the private sector. ²¹ A responsibility of the then newly appointed Chief Information Officer, Alex Benay, who later moved on to a heavily subsidized start-up selling AI products to the Canadian government, the CDS captured the government's emphasis on service externalization through growing investments in, and more sustained collaborations with, technology providers. While it met with resistance from many agencies, ²² the CDS readily exemplified the JT government's adoption of a more decentralized, result-driven program known as “deliverology,” which, in the name of improving public services, gave primacy to their design and delivery over more structural reforms touching on their nature, scope, and purpose.

Deliverology offered Trudeau and his Cabinet a promising framework to not only tackle high-profile issues but also recenter the practice of government on a quantifiable, data-centric mindset which, in the image of the JT campaign and, arguably, AI itself, was as much technological as rhetorical. First introduced by former Tony Blair advisor and regular Cabinet retreats attendee Michael Barber, deliverology can be understood as an impact-first approach to government. Defined as “a change in social, environmental, or economic outcomes” (Mendelsohn, 2019: 14) that can be directly attributed to a policy, the notion of impact became a cornerstone of the first JT administration, as illustrated by the creation of the new Results and Delivery Unit one month into JT's arrival into office.

The Delivery Unit supported the implementation of the “deliverological” practice of setting clear government objectives and using data, technology, and internal monitoring mechanisms to ensure they are met (see Moffit et al., 2010). As both a practice and conceptual framework, deliverology built on the New Public Management movement of the 1980s, which promoted the adoption of a more business-like model by public institutions as well as a renewed faith in the role and function of government. A highly iterative process, deliverology combined a rhetorical appeal to a mixture of top-down and bottom-up management with an emphasis on the tools, mechanisms, and technologies necessary for making this balance possible (see Barber, 2016). As one of the Unit's publications explained (see Mendelsohn, 2019), this framework promoted that all government programs be subjected to impact measurement assessments to evaluate how well they met their assigned objectives. In the name of metrological soundness, the Unit developed several metrics designed to measure the internal validity, behavioral effects, and causal impact of government programs and as such not only promoted a highly quantitative framework but also favored citizen-facing, highly measurable results over more structural programs and goals.

Deliverology, at its core, then involved a certain re-organization of the public service into an instance akin to a data broker, to the point where many of the government's highest positions came to be redefined around a similar emphasis on data collection and processing. Following the appointment of his Cabinet, JT's first initiative consisted in making public the mandate letters outlining the priorities set for each minister. Of the thirty letters made available, which together condensed and, at times, reformulated the 250 + promises brought forward during the campaign, nine promoted the modernization of the government's “data collection capacity,” with two of them singling it out as the main priority of the concerned ministers (Office of the Prime Minister, 2015a, 2015b). In many cases, this imperative translated into a series of research policies, procurement decisions, and investment strategies which, in the name of “deliver[ing] meaningful results to Canadians” (Mendelsohn, 2019: 11), singled out AI as a technology specifically optimized to produce such results.

After preliminary yet major investments in AI research at Université de Montréal and McGill University in 2016, JT first advanced a vision of AI that aligned with the data-centric rhetoric of his government. In an “AI Primer” produced by Google Canada, JT appeared alongside of leading AI researchers Geoffrey Hinton and Yoshua Bengio and promoted an understanding of AI as a decision-making technology optimized for delivering socially desirable outcomes (Google Canada, 2017). From there, as his government identified targeted investments, talent acquisition, and the creation of centers of excellence as the main pillars of its AI and, more broadly, economic strategy, JT reframed that technology in terms more closely aligned with his government's own objectives while building on a well-trod cluster approach to economic policy (Doloreux and Frigon, 2022). In an address promoting his administration's $125 million Pan-Canadian Artificial Intelligence Strategy and $230 million AI-Powered Supply Chains Supercluster program, JT emphasized how AI technologies “help create jobs, improve our quality of life, and generate new opportunities for the middle class” (2017). Phrased this way, AI emerged as an extension of the JT government's agenda, whose core priorities closely overlapped with the outcomes JT attributed to AI. During the 2018 G7 Meeting hosted by the Government of Canada, the JT administration even dedicated one of the day's working sessions to the articulation of shared commitments around the development of “human-centric AI” (Government of Canada, 2018: 2). Out of the twelve commitments Canada and its allies signed on, four of them specifically touched on economic equality, inclusivity, and accountability, conflating AI with the broader social commitments all G7 countries agreed on by the end of their 2018 meeting.

Much of the JT administration's actual application of AI, however, focused on the organization and delivery of government services. Beyond a few programs in autonomous space navigation and suicidal behavior detection in social media, the core of the federal government's AI solicitations focused on automated or algorithmically enhanced decision-making in public and/or government settings (e.g., Innovation, Science and Economic Development Canada, 2018; Public Health Agency of Canada, 2018; The Treasury Board of Canada Secretariat, 2018b), which translated into hundreds of millions of dollars in investment in a handful of companies. While these programs were subjected to the Responsible AI Initiative guidelines introduced by the Treasury Board, ²³ in practice these guidelines were articulated through a series of consultations where industry partners far outnumbered para-governmental organizations and social advocacy groups (see The Treasury Board of Canada Secretariat, 2018a), reflecting the corporate bias which informed how the success of these programs was measured.

Ultimately, of all the AI-related goals whose progress was documented by the federal government's Mandate Letter Tracker, none of them touched on the social implications of this technology. Instead, they all focused on job creation, trade, and investments in private entities (Privy Council Office, 2017), with 70% of the $1 billion of federal contributions awarded to the AI industry under the JT administration attributed to the private sector (Brandusescu, 2021: 10). Like most other policies related to digital governance and technology integration, Canada's AI policy centered on short-term benefits, contributed to an externalization of government services, and catalyzed a broader politicization of the Canadian civil service (see Clarke et al., 2017; Clarke and Francoli, 2017). By the end of JT's first mandate, his administration had even outsourced the mandate of articulating Canada's “long-term vision […] on AI both domestically and internationally” to a newly created body almost exclusively constituted of AI researchers and tech executives who had previously benefitted from the government's generous subsidies (Setlakwe, 2019). In the name of delivering results, JT's AI policy thus promoted a discretized, result-driven vision of government, where managerial and political control could be exercised by being outsourced to complex techno-corporate systems without explicitly raising the prospect of federal interference.

In continuity with its deliverological commitments, the federal government's AI policy then favored quantifiable results like job creation, corporate and foreign investments, and new companies over more structural, abstract, or socially inclined initiatives that cannot be so easily measured and quantified. As JT's first mandate came to an end, however, this political rationality came under growing pressure. For one thing, the JT administration became a regular object of inquiry of both the Ethics Commissioner and Information Commissioner of Canada. While the former documented several incidents where JT and his Cabinet allegedly committed breaches of the Conflict of Interest Act as to deliver some of the results his government so widely advertised, ²⁴ the latter undermined the JT administration's pledge for open government by exposing its repeated obstruction of the release of public information (Legault, 2017, 2018). Adding to the pressure was that despite the federal government's sustained efforts to position Canada as an AI leader, both federal and private investments in Canadian AI gradually declined. When the homegrown, widely advertised start-up Element AI was sold to foreign interests at a fraction of its estimated value, many commentators interpreted this event as a sign of the loss of the industry's Canadian momentum. Drawing any direct connection between both sets of events is indeed tenuous but their coincidence nevertheless points to a recurrent skepticism vis-à-vis a certain approach to government which re-emerged at different moments throughout Canada's postwar history. From PET to JT, from media theory to deliverology, we can recognize the persistence of a broader conception of technology as the embodiment of certain functionalist ideals regarding the practice, purpose, and optimal organization of government. In both cases, technologies like television and AI were mobilized in the articulation of a model of government that would not be bigger or smaller so much as more clearly—and definitively—defined, to the point of machine intelligence being recast from an extension of government to its reflection.

Seen together, the PET and JT administrations promoted a similar rejection of high-profile policies, multi-mandate programs, and cross-partisan endeavors in favor of managerial reforms, government structures, and technology as the defining practices and objects of government. Despite the weaker philosophical commitments of JT, both administrations endorsed a similar vision of intelligent, rational, or impact-first government as predicated on the development, valorization, and deployment of new information technologies, to the point at which their conceptions of government came to mirror these same technologies. While increasingly challenged toward the ends of both PET and JT's first mandates, this political rationality nevertheless persisted and returned throughout their premierships and beyond, if in perhaps more subdued forms. During the 2019 and 2021 elections, for instance, JT was not only re-elected with minority mandates but did so with the two smallest percentages of the popular vote any Canadian government ever had—another impressive, if more subtle, data-powered feat, enabled by further investments in new political technologies.

Conclusion: From intelligent government to machine intelligence

This article posited a connection between the Canadian federal government's ongoing promotion of artificial intelligence and its adoption of a structural approach to government from the 1950s onward. Combining a comparatist approach with a national lens, it contrasted the reforms and visions of Prime Ministers Pierre Elliott and Justin Trudeau, with a specific focus on the organizational a priori behind AI's deployment as a tool of government. Echoing Sandra Braman's work, which frames “trends in the development of information policy as manifestations of change in the very nature of governance” (2006: 9), it identified the two Trudeaus’ shared concern with national unity as a key driver behind the Canadian state's adoption of AI as a political technology. Beyond their filial relationship, the two Trudeaus crystallize two important historical moments when technologists, social thinkers, and politicians all debated how technology could not only fulfill government goals but also promote and embody them. When analyzed alongside one another, they then provide a rich perspective on the interplay between contemporary ideas about artificial intelligence and the historical convergence of political and technological goals.

Beginning our analysis with McLuhan and PET's more expansive understandings of both media and government, we recognize the latter as a manifestation of a broader “cybernetics moment” ²⁵ during which the horizon of what constituted intelligence was fundamentally expanded. Located in their original historical setting, early media theory and Trudeau's initial articulation of intelligent government both coincided with other new theories of organization that attributed intelligence to social entities of all kinds. From Friedrich Hayek’s book The Sensory Order (1952), which recognized the market as a useful example to reconceptualize intelligence as a capacity for self-organization in distributed systems, to Hebert Simon's article “A Behavioral Model of Rational Choice” (1955), which recast intelligence in managerial terms, there emerged throughout that period a consistent tendency to define intelligence first and foremost in terms of organization. Even before AI was defined as an object of inquiry—let alone a field—it was systems like the state, markets, organizations, and, crucially for the purpose of this article, government that were recognized as the bearers of a new form of non-human rationality. By recasting intelligence in organizational terms, this formulation informed how machines and humans were themselves reconceived of as bearers of intelligence, generalizing this highly managerial and structural understanding of intelligence beyond the confines of the social systems around which it was originally articulated.

In short, we posit a broader ideological and epistemological entanglement between the attribution of intelligence to entities like government and the reformulation of intelligence at large in organizational, structural terms. In continuity with Pasquinelli (2021), Bruder & Halpern (2021), Halpern (2014), and others, we find considerable proximity between technical conceptions of intelligence and theories of state intelligence, which we contend constitutes a key condition of possibility for artificial intelligence's formulation as a political technology. Not only did this proximity anticipate the ongoing adoption of AI as a tool of government, but it also continues to animate machine learning models like neural networks, whose highly distributed design mirrors the organizational ideals many governments have come to pursue. Compared to other political technologies like computer simulation, whose legitimacy relies on their capacity to performatively participate in “future making and world crafting” (Andersson, 2018: 6; see also Edwards, 2010), AI is for its part validated through its purported isomorphism with other intelligent systems like humans or states. As such, we understand AI's articulation as a political instrument as not only self-validating, but also embodying the same ideals and strategies of government alternately articulated from the 1950s onward as well as legitimizing the state actors employing it. By attending to this interplay, our paper extends the growing transnational research on the historical proximity between conceptions of machine and state intelligence, while also highlighting how this same interplay is constitutive of the normative ends that both Big Data and AI have come to embody.

As an established site of both economic and political investment in artificial intelligence, the Canadian national setting then indeed speaks to these ideals, but it is more specifically for the postwar history of its government that we singled it out as our object of analysis. From one Trudeau to the next, the Canadian government's history highlights several tensions that are now inherent in formulations of artificial intelligence. Not only do the federal government's historical and ongoing experiments in government organization and AI destabilize the latter's characterization as a stable set of technologies, but they also, and more importantly, convey an image of AI that is beset by conceptions of states, government, and markets as bearers of intelligence. By systematically pointing back to the organization of government itself, the Canadian state's experiments in that sphere then echo the historical interplay between organization and reflexivity in the history of AI (see Halpern, 2014; Hayles, 1999), while also stressing the key role of this reflexivity in engineering the perceived coherence of both AI and government (see Galloway, 2014; McKelvey, 2018). It is thus for how it sheds light on the highly reflexive nature of both AI and its deployment by governments around the world that we singled out the Canadian national context as a privileged entry point into the study of AI—one highlighting the co-constitutive relationship between politics and technology, organization and ideology in the evolution of postwar government.