Bureaucratic Silences: What the Canadian AI Register Reveals, Omits, and Obscures
This paper argues that Canada's new Federal AI Register, despite its stated goal of transparency, functions as an ontological design tool that obscures the human and sociotechnical complexities of AI systems by prioritizing technical descriptions, thereby reducing accountability to a performative compliance exercise rather than enabling genuine contestability.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
The Big Picture: The "Menu" That Hides the Kitchen
Imagine the Government of Canada just opened a massive, public menu called the "AI Register." This menu lists over 400 different computer programs (AI systems) that the government uses to do its job, from sorting mail to checking border security.
The government says, "Look! We are being transparent! Here is our menu, so you know what we are serving."
However, this paper argues that the menu is a trick. It's not a neutral list; it's a carefully designed tool that makes the government look efficient and in control, while hiding the messy, complicated, and sometimes risky reality of how these computers actually work.
The authors call the things the menu doesn't tell you "Bureaucratic Silences."
1. The Menu vs. The Kitchen (What's Revealed vs. What's Hidden)
The Analogy:
Think of a fancy restaurant. The menu (the AI Register) tells you the name of the dish ("Spicy AI Stew"), the ingredients (data), and that it was cooked by a famous chef (the vendor). It sounds perfect.
The Reality:
The menu doesn't tell you:
- The Kitchen Chaos: Is the kitchen on fire? Are the chefs overwhelmed?
- The Human Touch: Did a real human actually taste the stew, or did they just trust the robot to make it?
- The Mistakes: What happens if the robot burns the food? Who cleans it up?
What the Paper Found:
The Register is obsessed with efficiency. It says things like, "This AI helps us work faster!" or "This tool saves us time!"
But it stays silent on the uncertainty. It rarely admits, "This AI might be wrong," or "We don't fully understand how this works yet." It treats AI like a magic tool that just works, rather than a complex system that needs human supervision.
2. The "Sovereign" Lie (Who Actually Owns the Tools?)
The Analogy:
Imagine a country that claims, "We are building our own cars to be independent!" They put up a sign saying, "Made in Canada."
The Reality:
If you look under the hood, you find that the engine is made in Germany, the tires are from Japan, and the software is written by a company in California.
What the Paper Found:
Canada talks a big game about "Sovereign AI" (making their own AI to be independent). But the Register reveals that the government is heavily dependent on a few giant foreign companies (like Microsoft) for their tools.
The Register hides this dependency. It makes it look like Canada is in the driver's seat, when actually, they are renting the car from someone else.
3. The "Human" Ghost (Where is the Person in the Loop?)
The Analogy:
Imagine a self-driving bus. The sign says, "Fully Autonomous!"
But in reality, there is a nervous driver sitting in the front seat, gripping the wheel, ready to slam on the brakes if the computer gets confused.
The Reality:
The Register lists the AI as the "decision maker." It makes it sound like the computer decides who gets a visa or who gets a tax audit.
What the Paper Found:
In almost all cases, a human is still making the final call. But the Register doesn't explain how that human is trained, or how they are supposed to use the computer's advice.
- The Silence: It doesn't say, "The human needs 50 hours of training to understand this."
- The Risk: It leaves the human worker guessing. If the computer gives a bad suggestion, the human might not know how to fix it, or they might just blindly trust the computer because the Register made it look so "reliable."
4. The "Black Box" of Uncertainty
The Analogy:
Imagine a weather app that says, "It will definitely rain tomorrow!"
But in reality, the app is only 60% sure.
The Reality:
The Register treats AI like a weather app that is always 100% right. It uses words like "stable," "reliable," and "efficient."
What the Paper Found:
The Register hides the doubt.
- Some systems are actually just "experiments" or "pilots" (like a test drive), but the Register lists them as if they are finished products.
- Other systems are risky (like checking for fraud), but the Register doesn't say, "This might make mistakes."
By hiding the doubt, the government pushes the responsibility onto the regular people (the citizens) and the workers. If the AI makes a mistake, the Register makes it look like a human error, not a system failure.
The Main Takeaway: "Performative Compliance"
The authors use a fancy term: Ontological Design.
- Simple Translation: The way you list something changes what people think it is.
By designing the Register to look like a simple list of "cool tools," the government is actually designing a reality where:
- AI is seen as a perfect, reliable tool.
- Humans are seen as just "checkers" who don't need much training.
- Mistakes are invisible.
The Conclusion:
The Register gives us visibility (we can see the names of the tools) but no contestability (we can't question how they work or challenge their decisions).
It's like a magician showing you their hands are empty (transparency) but hiding the fact that the rabbit is still in the hat (the real risks and human work). The paper argues that for the public to truly trust the government, the Register needs to stop hiding the "silences" and start showing the messy, human, uncertain reality of how these systems actually work.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.