OTTAWA — The authorities preparing to regulate what Canadians see on the internet are focusing on the algorithms that decide what to show users of giant digital platforms.

It’s a big shift from the government’s thinking as recently as last fall, when the Liberals were proposing rules to fight online harms like hate speech and illegal porn that would have set up an elaborate new bureaucracy to judge the propriety of content that might be dangerous, piece by piece.

When they were running for re-election, the Liberals promised to bring in a bill with those plans within 100 days—creating a digital-safety commissioner to oversee rules for platforms requiring them to deal with complaints about any content they carried within 24 hours, under threat of multimillion-dollar fines, and an appeals body to consider complaints about the complaints.

Critics called the scheme unworkable, certain to cause platforms to censor vast amounts of material because it might be objectionable, and a model that repressive governments would follow. Heritage Minister Pablo Rodriguez, who took that job over from Steven Guilbeault after the election, pulled the plug in March.

Now, rather than talking about content, regulators, outside experts and MPs are focusing on the algorithms that promote one type of material over another—and not only to fight online harms.

Ian Scott, the chair of the Canadian Radio-television and Telecommunications Commission, was direct about how the commission is planning to get into the algorithm-oversight business, testifying to the Senate in June about Bill C-11. That’s the bill that would let the CRTC apply Canadian-content rules to streaming platforms like Netflix and Spotify.

It forbids the CRTC from imposing specific algorithms or code on those companies. But if the bill becomes law, the commission will be able to demand they do things—like sprinkling Canadian movies or music into lists of things users might like—that the companies could do only by changing their algorithms themselves.

“We will say, ‘This is the outcome we want. We want Canadians to find Canadian music. How best to do it? How will you do it? I don’t want to manipulate your algorithm. I want you to manipulate it to produce a particular outcome,’” Scott told the Senate’s communications committee. “Then we will have hearings to decide what the best ways are, and we will explore them.”

Google Canada (which operates YouTube), Meta and Twitter all declined interviews about the apparent change in approach. Netflix didn’t reply to an inquiry from The Logic.

“We are supportive of thoughtful government regulation,” Google’s Lauren Skelly wrote in an email. “We are committed to working constructively with the government to safeguard the experience of Canadian creators and users on our platforms.”

But Google filed a brief in YouTube’s name about Bill C-11—which the company objects to pretty comprehensively—that called for the clause forbidding the CRTC to order companies to use specific algorithms or code to be expanded into a prohibition on CRTC orders that would require algorithm changes.

“This will allow Canadians to continue to receive their interests, not those of the government, on YouTube,” the brief read.

When it comes to cracking down on content the government considers harmful or that’s outright illegal, the ideas are still brewing. But after the Liberals’ aborted first attempt, the panel of academics and a handful of practitioners they appointed to advise them on how to proceed concluded that regulating algorithms is likely necessary.

“They explored how actors can manipulate platforms to spread harmful, deceptive, or manipulative narratives through the use of bots, for example,” said the panel’s final report in early July. (All the experts’ work was written up as anonymized minutes of their meetings.) “They also cited research showing that women do not see online advertisements for jobs in science as much as men do.

“Experts stressed that systemic harms of this nature go beyond content and are imperative to consider. They explained that online services use algorithms and other artificial intelligence tools that perpetuate biases and harm users. They suggested that obligations regarding algorithmic accountability be introduced.”

In a separate process, a committee of MPs setting out to tackle ideologically motivated violent extremism voiced worry about the “big money” dimension to online extremism—that platforms make money by hosting content that aggravates people, and that algorithms designed to promote engagement naturally feed divisive stuff to users.

“Although Meta and Twitter told the committee they invest heavily in technology and human resources to enforce their policies, the harms that arise from the shortcomings of their efforts are undeniable and there is good reason to believe that these harms will persist until the current approach is changed,” the committee’s final report said.

Chaired by Liberal former cabinet minister Jim Carr, the House of Commons public-safety committee recommended the government “work with platforms to encourage algorithmic transparency and reduce online use by terrorist entities by identifying terminology and phraseology for better content moderation decisions.” None of the opposition MPs on the committee filed dissenting views.

This shift in approach follows a move in Europe, where the European Union’s Digital Services Act will require the largest platforms to share their algorithms with governments.

The big platforms say they’re on the case already, without new regulation. In declining an interview by email, Meta spokesperson Alex Kucharski pointed to the company’s moves toward algorithmic transparency, including plans to expand a feature allowing users to see why they’ve been targeted by particular advertising. The company said in 2021 that hate speech on the platform had declined by nearly half.

But some of the same outside observers who excoriated the government’s first attempt at fighting online harms say Canada is on a better track now.

Mandating outcomes and leaving decisions about how to achieve them makes sense in many policy areas, said Nathalie Maréchal, the policy director at Ranking Digital Rights, a Washington-based non-profit that scores big platforms and telecom companies on how well they uphold basic freedoms.

Restricting speech is different, Maréchal said in an interview. “In those cases, I think you have to take the opposite approach—you have to get really focused on the process, and process-transparency, and accountability for following the process, and trust that a good process will lead to good outcomes without mandating what those outcomes should be.”

If platforms are formally required to think about and report on the consequences of their algorithms, they will be more likely to mitigate potential harms themselves, Maréchal said.

That might not be as satisfying for a politician or policymaker, she conceded, but it’s “going to ultimately be much more protective of people’s rights.”

Focusing on content algorithms allows more flexibility than making black-and-white judgments about whether any particular material is banned or allowed, said Jason Pielemeier, the executive director of the Global Network Initiative, a U.S.-based industry and civil-society group formed to promote free expression in the digital world.

“That flexibility is really important to be able to build into any kind of regulatory framework, so that it can be adaptable and sustainable,” Pielemeier told The Logic.

Transparency, so outside experts like academics can scrutinize how platforms do what they do, is also helpful, he said—other eyes can detect problems that regulators or the companies themselves might not.

This will demand technical acuity on the regulators’ part, though, he noted.

“It’s not so simple as to say, ‘Here’s the algorithm that is intended to deliver hate speech,’” Pielemeier said. “There are oftentimes multiple overlapping algorithms that will end up determining what kind of content is surfaced, in what circumstances, for any given user.

“So it’s a little concerning to think that the government might see that as some sort of magic key—that if you dial this knob up and dial that knob down, all of a sudden you’re going to reach some perfectly harmonious state of content on that platform.”

But overall, he said, in a democracy with entrenched respect for expression rights, a back-and-forth between regulators and platforms will yield better results than dictates from above.