I. Can Google & Facebook Save the Republic?
Part one in a three part series
If you want to understand why it’s so difficult for Mark Zuckerberg to spin up an algorithm to save the world from the Devil and all his false news articles, it helps to stop and think about the silent transformation that has befallen the term “fact checker.”
When I was just out of college, I somehow lucked into that job at the New Yorker. In those days, most people had no idea what a fact checker was, and to the extent they did, it was because they’d heard specifically of the New Yorker’s checking department, which had been lionized in the press as something like a journalistic special forces team. The reputation had continued partly because of Jay McInerney’s quasi-auto-biographical novel “Bright Lights, Big City,” which featured a fact checker (like McInerney) at a prestigious magazine (like the New Yorker) who did a lot of cocaine (uh…); and partly, because members of the team itself fanned the flames. I certainly tried.
I was once checking Presidential advisor Clark Clifford’s autobiography, which was written primarily by the late Richard Holbrooke, one of America’s most prominent diplomats over the past 40 years. Holbrooke set me up in a large office at Clifford’s law firm in DC. I was 24 years old, sitting at a massive mahogany desk with a commanding view of the White House. I figured the best way to avoid sounding like a star-struck yokel from Alaska was to get to work. So I started to review some items in the text with Holbrooke and his researcher. The opening scene of the New Yorker excerpt introduced one of Clifford’s army colleagues from 1947. “Have you contacted him?” I asked.
“No,” said Holbrooke, “we don’t have any record of him, and it’s 45 years ago. I just don’t think there’s any way to find him. He may not even be alive.” After reviewing a few more pages of the piece, Holbrooke left, and I went through some files with his researcher and started working the phones (pre-world wide web, don’t you know). Holbrooke came back around lunch time to see how it was going. I reviewed a few small changes with him, including the army buddy’s job title and other vital details.
“What!?” Holdbrooke said. “You found him? Where?”
“Yeah,” I said, nonchalantly. “I tracked him to a hotel room in California. He was on vacation with his wife. Remembered Clifford well. I’ll send him a magazine.”
Holbrooke was an imposing presence, but for about ten seconds he stood there with his mouth open in a baffled smile, looking back and forth at me and his researcher to see if I was bullshitting him.
“You fact checkers are amazing,” he finally said.
I shrugged, knowing full well that a tremendous amount of luck had been involved in this particular stoke of genius, and said, “It’s a service we provide.” And so the legend continued.
In one important respect I was right to be blasé: that little act of sleuthing was almost irrelevant to what made the New Yorker fact checkers great. If it were only that — the collection of obscure facts — the internet would by now have made them almost obsolete. Yes, the magazine tasked us with banishing obvious error, but the real challenge was to tamp down ambiguity, lazy bias, misleading constructions and to discover the missing facts, the omitted pieces of the story that may meaningfully change the factual character of what had been written. In other words, the job required an overall understanding of context and how it adds up to a clear lens on reality or a foggy one.
So despite all the changes that have befallen the publishing industry in the past few decades, the fact checker remains, and they still have a bunch of them at the New Yorker (still led by the great Peter Canby), and they still do the kind of work they’ve always done.
But the term “Fact Checking,” as widely encountered by people nowadays, is a very different thing. Nowadays, people are generally talking not about an internal resource by which a publication protects itself, its writers, and its readers, but about a 3rd party that has set itself up as an adjudicating body to pass judgment on the veracity of statements from politicians, organizations, publications and others. On the surface, these two occupations sound more or less the same, but upon reflection, I think you’ll see that they’re not very similar at all.
In my old job, we collaborated with authors and editors to make sure we weren’t embarrassed and didn’t deceive our customers. We tried to validate the text as written, but it was not a fixed target. In a pinch, one could modify the speculative to conform to the certain. Claims could be reduced in scope, made less superlative; “the most” could become “one of the most.” The author could explain herself. The editor could ask questions. A new paragraph could be added or a troublesome one deleted. Good fact checkers might worry as much about soothing the author’s feelings or not disrupting a skittish source as they do about research technique and as they do about wordsmithing a perfect fix. In other words, good fact checkers master an art more than a science.
If the old fact checker was a doctor or a paramedic, the new one is a cop, with all of the burden and hazard that comes with being the man. Yes, they can help us indict our politicians and restore truth where there had been falsehood, but frankly, they do not always seem equal to the task. Or perhaps to be fair, I should repeat the point from the prior paragraph: they do not even have access to the task. The task, after all, is the careful dissection and reconstruction of context to provide coherent meaning.
Wrestle with that job for a while and you realize that building up the truth of a piece of writing requires that you have empathy for the writer’s intent and that the writer have fundamentally good will. These are the starting points, and they lead to a fruitful end through research, reflection, and dialogue/editing. But the last is not possible for the fact police, who arrive too late. And the reflection is often all too brief. In the old days, I might have had weeks to work on a long article. Now, the fact police try to adjudicate complex questions of context and intent in order to append a truth metric in near real time. I suppose there is something appropriately jejune about assigning a certain number of cartoon wooden boys or flaming pants from a playground taunt as the output of such an effort.
A recent AP fact checking debacle during a presidential debate provides a great example of the difficulties involved in fact policing, as opposed to fact checking. Donald Trump declared in the debate that, while he didn’t like Syria’s Bashar al Assad, at least the dictator was fighting ISIS. To which AP Fact Check responded, “Trump wrong that Assad fights IS.”
A fact check of the fact check followed and then devolved into partisan sniping. And since it is demonstrable from many, many sources (including AP) that the Syrian army has engaged ISIS in battle, AP ultimately backed down to the position that Trump’s statement was “partially true.”
“Partially true” brings us to an interesting epistemological juncture: when is something partially true, and when is it entirely true, but simply not a comprehensive statement on the subject? If I say “oranges are grown in California,” is that partially true because they also grow in Florida? Under what circumstances is my statement enough? Under what circumstances do I have an obligation to give a full accounting of the range of oranges? Will Californian and Floridian fact checkers agree?
If I take a hard-nosed view of the Syria/ISIS issue above, I’d have to conclude that AP is still wrong, and that they were wrong all the way through. Trump made no claim in his context to give a full accounting of Syrian politics. He merely claimed that, in spite of being a bad guy, Assad was killing members of ISIS — a fact which he, Trump, counts in Assad’s favor. This is all entirely true.
What Trump failed to do was give that broader accounting of Syrian politics. For instance, he did not mention that Assad’s Syria is a Russian client state and that Russia’s primary interest is in propping up Assad, not in destroying ISIS (although that may be a necessary part of the process). And he didn’t say just how bad Assad is. But AP Fact Check should not be correcting Trump on the sufficiency of the scope of his two-minute statement on Syria. Illustrating the insufficient breadth of his understanding of an issue was his debate prep team’s job; or it was Hillary Clinton’s job (and she generally did so quite credibly without help from the fact police); or it was an op-ed writer’s job.
In 1990, during the lead up to the Gulf War, the New Yorker published a commentary comparing the invasion of Kuwait to the assassination of archduke Ferdinand. I was brand new on the job. Editor-in-chief Bob Gottlieb came by my desk to review my changes, after which I added, with all the temerity of youth: “This archduke comparison is absurd. This isn’t about entangled alliances causing a world war. This is a war for economic empire.” Bob took the proof with my changes, patted me on the head, and departed saying, “Let me worry about that, my boy.” In retrospect, I can’t say my view has changed much, but then Bob was right, too: the problems were not factual, even in an expansive sense of the word.
This division of labor becomes even more stark in the fact police context. At least I had the possibility of entering the editorial conversation about what angle the New Yorker should take on the Gulf War. The piece had not been published yet. The fact police get stuck between the horns of a dilemma: on the one hand, they can only objectively indict on facts, but plenty of deceptions, if not most, are carried out by misleading selection of verifiable facts. On the other hand, if they wade into assessing the completeness and suitability of the author’s judgment and angle, they walk square into an accusation of bias. So the AP, in the Syria example, strains against the fact checking lane boundary in order to say, “Yeah, okay, Assad is killing ISIS, but that’s not his main goal.” They didn’t like Trump’s judgment about how to focus his remarks. And a large portion of their audience cried out, “Let me worry about that, my boy.”
This is the very peril that faces Facebook or Google should they try to step into the fact police business. They could — and the current fact check sites could — restrict themselves to the safe terrain of clear-cut facts (has Assad been killing ISIS or has he not?), but by doing so, they’re taking themselves out of the larger game. To really address the full problem of deception and error, you end up taking a stand on what constitutes the correct narrative. You become a partisan. A half truth is incomplete because you preferred the author to say the other half. When Planned Parenthood released the statistic that only 3% of their services were abortions, many media outlets had no problem passing the data along unchallenged, even though there were many reasons to think the value misleading. The Washington Post provided an exemplary, carefully considered 1500-word fact check on that stat and the equally misleading counter claim that 94% of what Planned Parenthood does is abortion. Both claims were “true” in some limited, highly defined scope. Both false from the perspective of getting a reasonable insight into the question of whether Planned Parenthood is largely an abortion enterprise.
Here’s the important point: not only do the fact police not get to have an editorial dialogue with the author prior to publication, the authorial intent itself may be entirely dishonest. At best, the intent is advocacy; at worst, propaganda or fraud. Off to the side, rank stupidity.
On that last category, a final example in which I personally embarrassed myself: As part of his ongoing critique of Hillary Clinton’s email hygiene, Donald Trump said that she had “acid washed” her hard drive. Various fact checking organizations called this claim false. I wrote to some friends, in an email thread on fact checking during the campaign:
This surprised me, in that, in spite of all the idiotic things Trump has said, it did not occur to me to believe that he was suggesting Hillary literally gave her computer the “Breaking Bad” bathtub treatment, but that he was merely using a colorful metaphor to describe her wiping her email server.
I mean, how dopey are these fact checkers, ignoring the obvious intent of the statement which . . . oh, wait a minute, crap. That’s right, it turns out Trump really did believe a chemical process had been used to delete the emails. The lesson for me: how hard it can be, given the fungibility of language, to even get a precise understanding of what an author was trying to say. And it also raises a further question: if you publish something like “Trump claims HRC ‘acid washed’ her email: FALSE” are you now implying to a casual reader than she did not in fact delete the mail at all?
Now enter Facebook and Google. A one-line statement handled by experienced journalists can turn into a quagmire. So can we really expect an algorithm to fix this problem of factual and linguistic slipperiness? Can even the presumptive repository of all the world’s knowledge adjudicate these disputes such that [those other idiotic] people can’t keep perpetrating their various sophistic rhetorical tricks on the [credulous/cretinous/malevolent] masses?
On the galactic-final-arbiter-of-truth scale, obviously not. But that is not to say that there is nothing they can do. If we start to tease apart the problem confronting these companies — the perils to the people who gather information there and to the fragmenting nation that bore them and that fragments ever further because of them — we may find there are some very useful and eminently achievable features they could add to their products without taking part in an ideological war.
Tune in tomorrow for another exciting episode of “Can Google and Facebook Save the Republic?”
Part II, The Epistemological Problem
Many thanks to Cynthia Cotts for her pro bono copy editing.
