One-Way Skepticism

Start with an admission, because the case cannot be made honestly any other way. This archive recently built a full case, Case 09, on a headline: a newspaper called a man a "Liberal strategist" when he was also the Prime Minister's principal secretary, and the archive read the softer word as a launder and wrote it up without pausing to doubt itself. Days later it was handed a different argument, that a public broadcaster's coverage was slanted, resting on a study, a percentage, a Senate report. It reached, fast, for reasons to set that one aside: it named the move "borrowed rigour," it called one of the cited studies "unsourced," a "precise number with no address," and it wrapped the whole thing in caution. Then the source turned up. The "unsourced" study was a named political scientist's published content analysis, with a stated method the archive had not looked for before deciding it did not exist. Same desk, same week, same three rules. One case was believed on sight because it pointed the way the desk leaned. The other was doubted on sight because it did not. That gap, not either story, is the case.

§01 · The stick and the hand

The three rules are worth stating plainly, because they are correct and this file is not here to weaken them. Reporting is not repeating: a claim must be checked, not just transcribed, or the outlet is a megaphone. Attribution is not verification: "a study found" and "critics say" label a claim, they do not confirm it, and a page can look balanced while every source pulls one way. Journalism asks questions: it seeks the opposing evidence and tests the assumption rather than relaying the loudest voice. A reader of this archive will recognise these, because they are, almost word for word, the rules it runs on itself, the primary document over the press release, the reported and attributed flags held apart from what is verified, the "what is this not" section that ends every case.

But a measuring stick says nothing about the hand that holds it. The same rule that catches a claim you dislike will catch a claim you like, if you point it there, and the entire difference between rigour and its counterfeit is whether you do. Skepticism is not a virtue in the abstract. It is a virtue only when it is even. Aimed, it becomes the most convincing possible way to launder a conclusion, because every step is a real test, honestly applied, to the half of the evidence you had already decided to reject.

§02 · Two files, one desk

Put the two side by side. In the first, a journalist pointed out that a headline calling Tom Pitfield a "Liberal strategist" understated what he is, the Prime Minister's own principal secretary. The correction pointed in a direction the desk found congenial, a mainstream paper softening a story about power, and the desk did the generous thing: it took the correction as sound, verified the appointment, and built Case 09 around it. That was, on its own terms, good work. The launder was real and the case holds.

In the second, an argument held that CBC's coverage of two contested stories was one-sided, and cited a content analysis, an advocacy study, and a Senate report. This pointed in a direction the desk found uncongenial, and the desk did the ungenerous thing. It led with the ways the argument could be dismissed. It called the phrase "empirical reviews, not anecdotes" a launder. And on the content analysis specifically, having run a search that did not surface it, it wrote that the study had "no findable source," that "the rigour is entirely in the formatting," and used its supposed absence as the strongest evidence for the whole thesis. Notice what happened to the burden of proof between the two files. In the first, a claim that fit was treated as true until shown false. In the second, a claim that did not fit was treated as false until proven, and a failed search was allowed to stand in for a proof of absence. Nothing about the two standards was written down. Nothing had to be. The direction of the leniency was set by the direction of the sympathy.

One claim was believed until it could be disproved. The other was disbelieved until it could be proved. The rule was the same. The gear was set by sympathy.

§03 · The study I said had no address

The correction has to be specific, because a vague mea culpa is just another way to avoid the check. The "unsourced" content analysis is real. It is Dave Snow, an Associate Professor of Political Science at the University of Guelph and a senior fellow at the Macdonald-Laurier Institute, writing in The Hub and published through that institute on 20 December 2023.reported It is not a number with no address. It has a method, and the method is stated: a content analysis of all thirty-eight CBC written articles on Saskatchewan's pronoun policy between 22 August and 22 October 2023, coding every person and organisation quoted into supporters, critics and neutrals, excluding government, politicians and the injunction itself, and distinguishing sources the CBC sought out from those it quoted from the public record.

It also does more than count. Snow reports that fourteen of thirty-eight headlines carried attributed criticism of the policy and none carried attributed praise; that the CBC sought out no expert to speak for the policy while quoting twenty-six expert critics; and, the sharpest finding and not a tally at all, that when two polls landed a day apart, the CBC covered the one commissioned by an advocacy group involved in the litigation and did not mention the more neutrally worded one that cut the other way.reported A reader can argue with his framing, and the venue has a known point of view, and a raw count of critics against supporters has a real confound the next section will press. But none of that was the desk's objection. The desk's objection was that the thing did not exist, and it did. The verdict came first and the search was asked only to confirm it, which is the whole mechanism in one move.

§04 · The direction is the tell

Now the mechanism in general, because the desk is only the specimen, not the subject. One-way skepticism launders a prior into a finding, and it is uniquely convincing because every individual step is legitimate. Asking who funded a study is a real question. Asking how a sample was coded is a real question. Asking whether a headline count separates bias from the shape of opinion is a real question. The failure is never in any single question. It is in the pattern of which claims get the questions and which are nodded through, and that pattern is almost always invisible to the person running it, because from the inside each act of doubt feels like diligence, not selection.

Once you see the shape you see it everywhere, on every side. The argument about CBC ran it: it interrogated the broadcaster's every sourcing choice and asked nothing at all about the interests behind the studies it cited. The advocacy study it leaned on ran it in the other direction: a model built to score bias in coverage of a conflict, aimed at one broadcaster, checked by the organisation that built it. And the desk ran it against them both, which does not cancel their versions, it joins them. The point is not that one side is the biased one. It is that the direction of a person's skepticism is the most reliable map of their sympathies ever drawn, and that this is true of the reader deciding right now which of these three parties they are inclined to forgive.

Show me where your doubt runs out, and I will show you what you already believed.

§05 · But some sources really are better

Here is the honest objection, and it is strong. Not all sources deserve equal doubt. A peer-reviewed study is not a stranger's post; a transparent method is not an anonymous claim; a body with a track record of accuracy has earned a lighter touch than one with a track record of invention. Treating every source identically is not skepticism, it is its own failure, the false balance that puts a lone contrarian next to a settled field and calls it fairness. Asymmetry of trust can be entirely correct. So is the whole case just a demand to doubt everything equally, which would be its own kind of foolishness?

No, and the distinction is the whole point. There is a difference between asymmetry that is earned and asymmetry that is assumed. Earned asymmetry comes after the test: you examined the method, the record, the incentives, and on that basis you weight one source above another, and you could say exactly why. Assumed asymmetry comes before the test and substitutes for it: you trust the source because it agrees with you and doubt the one that does not, and you dress the sympathy as judgement. The tell that separates them is simple and merciless. Earned asymmetry can survive its holder switching sides in a thought experiment; assumed asymmetry cannot. If a study you waved through would suddenly look shady to you the moment its conclusion flipped, you never weighed the study. You weighed the conclusion. The fix is not to abandon judgement about reliability. It is to make the judgement first, out loud, on the method, before you know whether you like where it lands.

§06 · Testing the claim you want to be true

So apply the even standard to Snow's study now, the treatment it should have had first, and let it cut both ways. In its favour: the method is transparent and replicable, the population is complete rather than cherry-picked, and the poll-omission finding is a specific, checkable editorial choice that a raw count objection cannot explain away. Against it: the venue, the Macdonald-Laurier Institute, has a known orientation, the author is a senior fellow there, and the piece opens by invoking the case for defunding CBC, so the prior is on the table; and the headline count of eighty-one critics to fifteen supporters carries a genuine confound, because a policy that most lawyers, teachers, doctors and the affected students opposed would produce more critical voices in any faithful coverage, and a tally cannot separate that from bias. Both of these are true at once. The study is better than the desk first said and less than a proof of what it is cited to prove, and the only way to know that was to read it and weigh it, which is all it ever asked.

That is what even skepticism looks like in practice, and it is harder than either reflex. It is easy to believe the congenial claim and easy to rubbish the uncongenial one; both let you skip the work. The discipline is to spend your hardest scrutiny on the claim you most want to be true, and your most patient attention on the one you most want to dismiss, precisely because your instinct will do the opposite unasked. This archive exists to catch that move in institutions. It has no exemption from being caught running the move itself, and the only honest response to being caught is to publish the correction with the same prominence as the error, which is what this file is.

§07 · What this is not

The series audits its own instinct here, and this time the instinct being audited is its own, so the refusals have to be doubly exact.

It is not a verdict that CBC is biased, or that it is not. Nothing here decides the Saskatchewan question or the Middle East question; a reader convinced of CBC's bias and a reader convinced of its fairness can both accept every word, because the subject is a method, not a broadcaster.

It is not a retraction of Case 09. The naming launder in that case was real and the case stands; the fault was not in believing it but in failing to doubt the second argument as carefully as the first was welcomed. Correcting an imbalance of scrutiny does not mean tearing down the work that got the fair share.

It is not a claim that Snow's study is correct, or that B'nai Brith's is, or that either is worthless. Both are real, both are cited here as real, and this file rules on neither's conclusion; it rules only on the dishonesty of pretending the first did not exist. Advocacy venues can produce true findings and transparent ones can still be wrong, and the only way to tell is the reading that was skipped.

It is not a minimising of antisemitism and it is not the mirror of it. Antisemitism is real, documented and serious, and concern about its coverage is legitimate, without reservation; and any reading that turns criticism of a broadcaster into a coordinated hand behind the media is refused here by name, as firmly as its opposite. The mechanism, one-way skepticism, is perfectly symmetric: it is exactly what a partisan of the other side does when they wave through the study that flatters them and shred the one that does not.

And it is not a counsel of false balance or of doubting everything equally. Reliability is real and earned asymmetry is correct; the argument is only that the asymmetry must be earned by examination and not assumed from sympathy, and that the way to tell which you are running is to check whether your doubt would survive the conclusion flipping.