This past June, Canada released its national AI strategy, AI for All. It’s an industrial strategy with the aims of strengthening our economy through investment, innovation, and global trade.
Worth noting, for me and perhaps for you: it is not a service plan or service strategy, the kind that describes in detail how and where government will deploy AI to improve its services. There is a brief mention that AI will transform public service delivery, but the statements are general and left me wondering. How exactly might administrative burden be reduced? Which services will be improved? How will government identify the service areas best suited to AI, and the ones that are not? I was also keen to learn how government plans to modernize the identification, procurement, and deployment of AI tools, another commitment the strategy mentions but does not detail.
All of that will have to wait for another policy and another day.
So while it was not enlightening from a service delivery perspective, the strategy was instructive in other ways. It illustrates the variety of roles government plays and the many policy levers it holds, demonstrates how government thinks about AI and technology generally, and reveals how it imagines citizens as part of the mix.
What the strategy asks of Canadians
The gist of the strategy, as articulated in the vision section, is that AI is an autonomous, inevitable technology full of potential and future benefits, and we Canadians are behind in the race. We must adopt AI, but before we do, we need to trust it. To trust it, we need to learn it, understand it, and use it. And finally, once we trust and adopt it, and sort out the sovereignty issues tangled up in the AI supply chain… we’ll all benefit.
The trouble with inevitability
The idea that technologies are inevitable has been a long-standing source of concern for me. In preparation for celebrating OXD’s thirtieth anniversary, I recently came across a 1996 photo of myself writing a third-year university term paper for a technology studies class. Seated at the kitchen table with my hand in a bag of cookies, I had a stack of library books beside me, one of them Leo Marx and Merritt Roe Smith’s Does Technology Drive History? The Dilemma of Technological Determinism. Thirty years later, and I’m still allergic to the rhetoric of determinism and inevitability.
I’m glad to see I’m not alone. Coinciding with the national strategy’s release is the June 2026 Vertesi et al. ACM FAccT conference whitepaper, “Reckoning with the Political Economy of AI: Avoiding Decoys in Pursuit of Accountability.” The authors outline how those who fund and develop AI systems use decoys as a form of misdirection that distracts us from contemplating “what the Project of AI is, how it is working, and where the levers of power and accountability are located.”
It’s worthy of a full read, but let me draw your attention to decoy number two, The Inevitability Decoy: “when powerful actors project a story of technological determinism and inevitability, they enhance their own economic and social capital and constrain other actors and futures in the process.” In doing so, these “cultural narratives about AI therefore often conceal the material power moves that the story of inevitability enables.”
Inevitability and determinism remove our agency and choice in the matter of technology. It’s not about if we’ll use AI, it’s about when. With the conversation shifted to the timing of our adoption, we Canadians simply need to move on from our laggard position and get on with the business of believing in the promise of AI. We need to trust it. The strategy tells us so.
“Canadians need to trust in its promise. Trust that they will share in its benefits and that it will be developed, adopted, and governed in ways that reflect our shared values.”
Trust isn’t a prerequisite
The idea that trust precedes adoption also strikes me as problematic.
Helen Hayes recently dismantled this theory of change (or misguided hunch about change?) in her June 9 op-ed in The Globe and Mail, “Canada’s AI strategy won’t build necessary trust.” First, she calls out the problem of treating trust as a prerequisite to adoption, when it should be an outcome of governance. Then she challenges the assumption that personal exposure and use will breed familiarity with AI, and shows why that familiarity, even so, is insufficient for trust.
“Why? Well, the AI systems the government hopes to invest in and accelerate adoption of are not simply consumer products that Canadians can choose to use or not. They are predominantly AI systems that will be integrated into workplaces, education systems, public services, and health care. In these contexts, public acceptance depends less on familiarity than on legitimacy.”
She then points out that most Canadians, upon whom AI systems will have an effect, are in a poor position to validate the reliability of the AI systems we’re being asked to trust.
“A patient cannot independently assess the technical reliability of an AI-assisted diagnostic tool. A resident cannot audit the algorithmic systems used to allocate public services. An employee cannot evaluate the models informing workplace performance evaluation. The challenge in each of these cases is not a lack of knowledge, and literacy will not solve the skepticism that Canadians have of those technologies.”
Hayes has hit upon a subtle but important difference: the distinction between trust and trustworthiness.
To better understand the difference of the two concepts and the psychology of trusting AI, a recent paper, “Principles for understanding trust in artificial intelligence,” by UK authors at the University of Kent and the University of Oxford is also worth the time and effort to read. They note that trustworthiness is often considered a characteristic of the system, answering an evaluative question (“Is this entity worthy of trust?”), whereas trust is a subjective attitude held by the person doing the trusting. Within the concept of trustworthiness, they draw a further distinction: whether a system objectively has the characteristics that make it worthy of trust (its “actual trustworthiness”) is different from whether people perceive it as having them (its “perceived trustworthiness”).
The authors summarize this distinction and why it matters (emphasis mine):
“In sum, developers can focus on building normatively trustworthy systems, which is likely to increase people’s trust in their products. However, these points are distinct: the fact that something has characteristics that increase its actual trustworthiness does not guarantee people will perceive it as such, and the fact that people perceive a system as trustworthy does not mean it is actually trustworthy.”
The semantics and mechanics of AI trustworthiness and human trust may seem to some like dancing-on-the-head-of-a-pin territory. Yet trust and technology is a closely studied and debated field, one where scholars of technology like UCL’s Jack Stilgoe state “people trust people, not technology” and one where the answer to “Do you trust AI?” is not a simple yes or no, or where adoption rates are a measure of our collective belief in the technology.
Trust is earned, persists or is revoked over time, and rests on many factors and variables. It certainly isn’t a byproduct of government telling us what we need to do.
Designing with the public, not upon them
This detour into trust and trustworthiness, be it of AI or of government in general, takes me back to the thing I care about most here at OXD, and the thing largely absent from this strategy: the practice of service design and delivery, for and with government.
As Hayes pointed out, and as I quoted above, the opportunities to establish trust, and to judge the trustworthiness of any new technology, often arise in an institutional setting, where that technology lives further down in the value chain, embedded in a broader education, health care, or justice system.
As citizens, we might not experience AI directly (the frustrating chatbot aside), but indirectly, as an algorithmic decision somewhere in the process: triaging cases, detecting fraud, scoring eligibility. The place AI is most likely to operate, the backstage of government services, puts pressure on the values citizens need in order to trust their government: integrity, openness, and fairness.
AI may be hidden from sight, but its impacts will be felt.
The work to be done
If government wants the trust that this strategy asks Canadians to extend, the place to invest is the practice of designing services with the public rather than deploying AI upon them: participatory and value-sensitive service design methods that give citizens a voice in whether and how AI is used, that make the hidden decisions legible, and that afford a path to recourse when the system gets it wrong.
Sometimes that will mean using AI, and sometimes it will mean deciding not to.
At OXD, this is the work we do with our government partners: treating design as the process that earns trust, rather than asking citizens to give it upon demand. That is what we would urge the strategy’s authors to fund.