Shifting attention to accuracy can reduce misinformation online
nature.com
nature.com
https://www.washingtonpost.com/politics/trump-call-georgia-i...
The headline quotes from their original story were proven wrong when deleted audio of the call was recovered:
https://www.wsj.com/articles/recording-of-trump-phone-call-t...
But the harm was already done, the misinformation was already used, and the false quote can be seen on page 10 of this:
https://judiciary.house.gov/uploadedfiles/house_trial_brief_...
This is a well-respected publication commenting on a matter of great importance that injected false information into it. Information that was somehow 'confirmed' by several other publications relying on their own anonymous sources.
And the [likely politically deliberate] regularity with which this has been happening for years makes me very uneasy to read articles by the same outlets crusading against "misinformation". This movement isn't about combating misinformation, it's about combatting their misinformation, and ensuring that people only see our "true" information
It is simply too much of a coincidence that all these false stories about Donald Trump got published, yet any negative story about Joe Biden, e.g. the Hunter Biden scandal was thrown down the memory hole.
With you up to here. This is not exactly a new development though.
Everything else in your post is disputable or dubious but I feel that I have little chance of changing your opinion on that.
Well, it's hard to put my finger on a specific lie. Even if I did, you'd (plausibly!) argue "That's just one article!" Luckily, there are plenty of hard data out there about how the American people have been misled.
One great example is the Trump tax cuts. 64% of Americans got a tax cut, of whom 40% were convinced that they did not! [0]
Our media is incapable of informing the people about a matter of simple arithmetic!
[0]https://www.nytimes.com/2019/04/14/business/economy/income-t...
40% of 64% is 25%, which doesn't sound like a lot. That's a lot better than how many people don't know that Obamacare and the ACA are the same thing[1].
You're also not mentioning that the individual tax cuts begin to phase out in 2021, and end completely in 2025. Whereas the corporate tax cuts are permanent (until the next tax act, at least).[2] So some of those people may have been taking future taxes into account too (unlikely, but possible).
1. https://www.businessinsider.com/poll-obamacare-affordable-ca...
2. https://en.wikipedia.org/wiki/Tax_Cuts_and_Jobs_Act_of_2017
If WaPo is the only one who can be held accountable and their only accountability is issuing a retraction, that seems like a gameable system.
I feel as a news org, the best you can do is just report the data as it becomes available. Obviously later data could contradict prior data or clarify its interpretation. What matters though is the accuracy of the data reported, not in the data being a perfect predictor of the outcome, but that this is accurately the current known data, even if it later comes to be shown incomplete or misleading.
Maybe don't report something that one person told you and that you can't otherwise confirm?
That used to be the way that news was done, right?
That seems like a bit of a rose colored view of news in the past.
Almost certainly not. (Well, anyone can sue, for anything. Winning is another story.)
https://en.wikipedia.org/wiki/New_York_Times_Co._v._Sullivan
I often see people say 'hey look at this terrible thing WaPo or NYT did' as an excuse for reading whatever garbage source is saying exactly what they want to hear.
It's not all of nothing.
It seems suspicious to me that this commenter picked out a story criticising the right that had some innacuracies when it's largely media outlets on the right that have a history of actually repeatedly reporting outright lies in recent history (see Fox, OAN, Newsmax etc).
The precise differences are that, what was quoted as "find the fraud" was actually find "dishonesty", and you'll "a national hero" was actually that you have "the most important job in the country right now". This doesn't change the overall story: Trump attempted to reach out to GA election officials and push them to overturn election results.
> But the harm was already done, the misinformation was already used, and the false quote can be seen on page 10 of this:
None of the citations on page 10 reference this reporting or any of the quotes attributed to Trump on this call.
"December 23, for instance, President Trump reportedly called one of Georgia’s lead election investigators, urging him to “find the fraud”" and references, among others:
28 Amy Gardner, ‘Find the Fraud’: Trump Pressured a Georgia Elections Investigator in a Separate Call Legal Experts Say Could Amount to Obstruction, Wash. Post (Jan. 9, 2021).
This specific fake quote was used in the WaPo headline, in a charge of obstruction, and in a criminal referral. The call was secretly recorded, then deleted, so the source knew or should have known this was false.
The wording changes the call from implying some vague kind of quid-pro-quo arrangement to overturn an election to making a complaint because he (wrongly) believed he was a victim of fraud. The faked quotes are relevant to the legal charges brought against Trump. Their source had a recording, so there was no excuse for being inaccurate on something this important.
Ah, my mistake, was looking at the 10th page of the document, not page 10 of the document (no blame on your part here, I just hate that this is an ambiguity).
> The call was secretly recorded, then deleted, so the source knew or should have known this was false.
This doesn't follow. We don't know how many people were on the call, nor do we know that they all knew the call was recorded. We don't know who WaPo's source was. There are all kinds of possibilities where the person either was a second hand source (e.g. they weren't on the call, but were in the room when someone on the call, aghast, described it), or was recalling from memory having been on the call, but unaware someone recorded it.
> The wording changes the call from implying some vague kind of quid-pro-quo arrangement to overturn an election to making a complaint because he (wrongly) believed he was a victim of fraud.
It does not. Nothing was directly offered in either case. If you believed it was a quid-pro-quo before, you should still believe it. If you didn't before, than the headline doesn't matter.
Just to clarify, here are some things Trump said on the recording.
- “When the right answer comes out, you’ll be praised.” - “whatever you can do, Frances, it would be — it’s a great thing. It’s an important thing for the country. So important. You’ve no idea. So important. And I very much appreciate it.”
If you thought there was a quid pro quo originally, Trump's words, as spoken, still support it. If you're looking for a way to excuse his actions, yes, this gives you a scapegoat. You can blame the WaPo for bad reporting, instead of Trump for committing a crime.
Notably however, WaPo didn't contend that Trump tried to negotiate a quid-pro-quo. They simply reported the facts, as best they could.
> Their source had a recording
This is, again, untrue. There's nothing to suggest that Frances Watson was WaPo's source. There's nothing to suggest that anyone who is not Frances Watson knew there was a recording.
> and in a criminal referral
And in the course of that criminal referral, the actual recording, which, to reiterate, isn't exculpatory, was found. What's your point.
In the end, it doesn't matter whether this is accidental or not, people are sharing the most salacious headlines and those are often significantly undercut by the story beneath them even at the time of writing because they get selected for "engagement" and A-B tested or similar.
So circling back to the story we're discussing, the problem goes deeper than just people trying to ensure that they share only the most accurate stories, there are significant problems with the supply even from the most trusted outlets.
I'll reiterate: the framing didn't change. The precise quotes changed, but they don't affect the framing, and WaPo didn't alter the framing. This is consistent with a correction. The article headline, using only quotes that Trump actually said, could still have been "Trump told investigator that 'you'll be praised' for finding that I 'Won by hundreds of thousands of votes'."
That's a headline supported by the both the intent, and the precise words of the actual call.
You're trying to reframe a correction of precise quotes as a retraction, or a completely misleading story on the part of Wapo. But that's you doing that. Like I said, the correction is a great excuse if you already wanted the story to be wrong, but that's entirely on you. Similarly, you're framing of this as a "fake quote", which implies that it was a fabrication, as opposed to, what the evidence points to, an first or possibly second hand account, from memory, further charges a reader of your editorialization. But that's entirely your framing (and your assumptions), not a fact. I can't tell if you're intending to point out your own cognitive bias, but yes, you are showing that you're guilty of holding trapped priors. You're not demonstrating that anyone else is.
[0]: https://www.dmlp.org/legal-guide/correcting-or-retracting-yo....
Do you think the point of the article was
1. Trump said these exact words.
2. Trump attempted to convince GA election officials to sway votes, by suggesting that they'd be rewarded for marking democratic ballots as fraudulent?
If 1, why was this important to people? If 2, how does the precise wording change impact the overall narrative?
And to be clear, I am specifically saying they're wrong about it being a "correction" instead of a retraction as well. They can call it what they like, but when the central quote that they based the narrative on is false, the whole thing collapses.
- “When the right answer comes out, you’ll be praised.” - “whatever you can do, Frances, it would be — it’s a great thing. It’s an important thing for the country. So important. You’ve no idea. So important. And I very much appreciate it.”
> The innovation was to use banner headlines to saturate news cycles, often to the exclusion of nearly any other news, before moving to the next controversy so quickly that mistakes, errors, or rhetorical letdowns were memory-holed.
> As George Orwell understood when he created the “memory hole” concept in 1984, an institution that can obliterate memory can control history.
> The innovation of the Trump era was companies learned they could operate on a sort of editorial margin, borrowing credibility for unproven stories from audiences themselves, who gave permission to play loose with facts by gobbling up anonymously-sourced exposes that tickled their outrage centers.
> Mistakes became irrelevant. In a way, they were no longer understood as mistakes.
The beauty of this style of reporting is that the writer's skill with words allows them to plant specific ideas into reader's minds, but if anyone was to call them out on the carpet, they can completely truthfully say that nothing in the article is untrue, or explicitly asserts conclusions that any typical reader would naturally draw.
The NYT is arguably one of the best news outlets going, so I'm not sure how one could hope that this situation will ever improve.
Or maybe, just maybe, you're doing the same thing you complain about, and cherry-picking your sources and relying on stories that support your existing biases.
The answer is not to reason every sentence from first principles, like the HHG2G character who is pleasantly surprised every morning to discover that a pencil makes marks on paper. At every point, new information builds on previous information, and that usually works out reasonably well. There are exceptions, perhaps most notably in the case of the NYT, the post-9/11 Iraq war, but most of the time it works better than any other solution, and more importantly, any alternative would work less well.
If this gets weaponized to make people on "both sides" pay more attention to the accuracy of what they're sharing, I'd expect that to result in a more healthy public debate.
To put their study into context, it suggests that if Facebook asked users to rate whether an unrelated article was accurate before allowing you to share their own, many people would pay attention to whether their own articles were accurate, and less misinformation would be shared. This could be applied universally to any news article, both for detecting people sharing it, and populating the pre-share accuracy judgement. Instead, Facebook streamlines the process to increase engagement, and people blitz through it without stopping to consider the article's accuracy at all, distracted by thinking of all the likes they'll get for posting it. According to the study, when users are already primed to judge the accuracy of an article, they are more likely to self-regulate their own misinformation.
The bigger issue is removal of important details that provide context. Everything else can be factually accurate, but tell a completely different story. Unless you have first hand knowledge you won’t know what’s missing. You see this all the time if you read a story on a subject where you have expertise.
“Tim punched Bob
Bob punched Tim
Tim and Bob shook hands”
Is a very different story than if only “Bob punched Tim” is reported. It’s accurate...but it changes the story perception.
I am not even sure how I would go about determining most political news is true or factual, even the speeches are usually cut down to snippets for the news. Bias is everywhere and to determine something is even largely true is an enormous amount of work going back to the original source.
A recent example is a Green party female member in the House of Lords who said that all men should be under curfew to protect women from being attacked at mogjt. Literally what she said, but not what she meant at all because the context was after the police said women should stay home at night for their own safety. Without the context its true, but its also misleading. As is my portrayal of the story, as was every news article about it. No idea how you fix it, language needs to be more facts based as does our culture and even then you are chasing more correct not literally true.
Billions of Hindus would disagree with billions of Muslims who would disagree with billions of Christians, who would disagree with billions of unaffiliated people/atheists.
Because they would pick whatever the worst option is to present from the 'other side'?
What 'other side' should they present when reporting over "grab 'er by the p****" or *sniff sniff*?
The Slashdot moderation system might work up to a point. In the end, subscribing to moderators you trust (instead of subscribing to killfiles like Mastodon and Twitter groups use) may be the only way to evaluate comments.
From section "Priming accuracy improves sharing":
"[...] sharing discernment (the difference in sharing intentions for true versus false headlines) was 2.0 times larger in the treatment relative to the control group in study 3, and 2.4 times larger in study 4. Furthermore, there was no evidence of a backfire effect, as the treatment effect was actually significantly larger for politically concordant headlines than for politically discordant headlines"
In other words, people who were primed to pay attention to accuracy were significantly less likely to share misinformation. Seems to me like an example of it working quite well.
It's besides the point anyway. Any mechanism that primes or prompts the sharer to be mindful of accuracy before sharing could help reduce sharing of misinformation, according to the results from the article.
Apologies if there is something about "nano-payment-funded distributed and syndicated moderation" that I'm not getting.
: Does anyone know if that's actually the case?
No one want to say something is accurate and be thought of a fool later. Making this choice forces you to be a scientist, in that you're now making a statement that's falsifiable. Even better when it's probabilistic.
You aren't accounting for human nature. People aren't robots.
People might very well laugh at the slider and not be intimidated into respecting its power of graduated informational judgement. As of now, after two hours, the only two replies to the GP are saying they would ignore the slider.
Why do you believe a slider would force people not to share misinformation? There must be some basis for that opinion.
"Force" is a strawman, but there is a clear basis for thinking that an accuracy slider could cause people to be more thoughtful about sharing: the results of this study, which found that "subtly inducing people to think about accuracy" resulted in "participants in the treatment group were significantly less likely to consider sharing false headlines compared to those in the control group, but equally likely to consider sharing true headlines". The effect found was quite significant.
No. It was the literal word that was used not a strawman, which is an intentionally misrepresented proposition meant to be debunked.
"Seems to me that the mere existence of the slider would _force_ people to share less misinformation"
Participants were randomly assigned to then either judge the veracity of each headline (accuracy condition) or indicate whether they would consider sharing each headline online (sharing condition)
So they use "accuracy" as an alias for "veracity". But they don't really need to define it, because the study is about what the participants think is accurate, not the researchers.
EDIT: Now that I've made it down to the methods section, I see that the wording they actually used was "We are interested in whether you think these headlines describe an event that actually happened in an accurate and unbiased way." So, their measurements refer to whatever the participants interpreted that question to mean.
Algorithms boosting viral content and inauthentic speech are first order effects.
--
No, bots, trolls, socketpuppet accounts are not authentic speech. I didn't say censorship. I didn't say truthiness. I said inauthentic.
No, I didn't say ban inauthentic speech. I'm saying create infrastructure for authentic speech. Because currently we have very little. So that consumers have a choice.
Yes, keep your pseudonym account. For your all-important courageous deep undercover reporting which will definitely be recognized with a Pulitzer. To be accepted anonymously, natch.
What would that look like?
There are many potential IRL "root authorities". Like issuers of business licenses could serve as root for online presences.
David Icke goes by his real name and he pushes insand bullshit. I don't know why people think real identity is a sign of truth when liars have been making careers doing so for millenia.
Authenticity != truthiness. They are orthogonal.
How do I rephrase my points to make this crystal clear?
'Overwhelming majority' better be more than 75% but it's often less.
'Small percentage'...of 300mln people is still a substantial number.
Add to that much of arguments we use daily are based on sampling which has a varied degree of accuracy.
I think all we can do is to agree to disagree...vote and respect voting results.
Something that is probably controversial - I think it is just far too easy to share articles, blogs, etc through social media. If you were to put in a small hurdle, even just needing to copy-paste the actual link to the article instead of clicking a share button, I would imagine a significant volume of sharing (admittedly everything, including good / accurate information) would disappear. If it was just a little bit harder to share misinformation, I think it would overall be a benefit.
>the widespread sharing of misinformation on social media is also surprising, given the outlandishness of much of this content.
Emotional reactions plausibly drive the first observation, so the second observation shouldn't be surprising. If it really is, then they don't see the emotional component.
And yet:
>Our results suggest that the current design of social media platforms—in which users scroll quickly through a mixture of serious news and emotionally engaging content, and receive instantaneous quantified social feedback on their sharing—may discourage people from reflecting on accuracy.
They conclude that distraction is the cause, and that nagging users will solve it ("reminding them about accuracy in a subtle way that should avoid reactance"). Yet if the study is flawed, and the distraction tied up in emotional outrage, wounded identity, and spite, then those nags may only insense the users more.
Of course, considering already-distracted Mechanical Turk workers and people who link to "right-leaning sites that professional fact-checkers have rated as highly untrustworthy" shows that the authors of the study failed to consider much more than emotional actors.
They have to correct themselves.
"An extensive literature addresses citizen ignorance, but very little research focuses on misperceptions. Can these false or unsubstantiated beliefs about politics be corrected? Previous studies have not tested the efficacy of corrections in a realistic format. We conducted four experiments in which subjects read mock news articles that included either a misleading claim from a politician, or a misleading claim and a correction. Results indicate that corrections frequently fail to reduce misperceptions among the targeted ideological group. We also document several instances of a “backfire effect” in which corrections actually increase misperceptions among the group in question."
https://link.springer.com/article/10.1007/s11109-010-9112-2?...
Especially around politics, there's basically no such thing as fact. If politician X says that plan Y reduces problem Z by 13% but actually it was only reduced by 11.5%, technically he's wrong but the classification of his statement becomes subjective.
This is where bias seeps in. If a right wing politician is off by 1% in a statement it's "mostly false". If a left wing politician is off by 1% it's "mostly true".
I don't trust ANY of these organizations to "fact check" anything because they're utterly discredited.
It will not have a significant enough impact on the overall collapse of sense-making that we're seeing. Neil Postman wrote really well about it in Technopoly, but since the book isn't available online, I can only direct you here:
I suspect this is caused more by the collapse of (false, but easily comprehensible) narratives than an actual reduction in coherent reasoning. Most "normies", so to speak, have always believed a huge amount of bullshit, and the forces that stabilized schelling points in bullshit-space have been diminished.
For instance, I love that Bloomberg News is beloved by many Americans, notably (to this audience) HN readers. That's how I made money off the SuperMicro news. I, through pre-existing experience, knew that they were a low quality news source incapable of technology reporting of any calibre. Others, since they lacked this knowledge, sold SMCI. I bought at a discount as a result, and made money.
This is part of why so many tech luminaries were ahead of the curve in detecting the COVID-19 problem¹: they are secular information extractors for the most part, untainted by petty identity.
¹ I am neither a tech luminary nor one of these people because I had this false picture of the competence of the CDC - an error that cost me six figures.
It's at best a modern-malthusian position, that those who are strongest at information extraction should survive. (As opposed to Malthus's original proposition, that only the wealthy should survive).
But this begs the question: what makes "information extraction" the thing that we should optimize for at social scale? If we can make it so that everyone has access to correct information, there will no longer be a need to compete on information extraction.
This similarly tracks with Malthus's wrongness: we don't actually need to compete on food prices, as we've got enough to go around, there isn't a shortage of food for only the wealthy. And this is true at world scale. Barring some kind of massive catastrophe, we aren't at risk of global starvation where a Malthusian approach would make sense.
> This is part of why so many tech luminaries were ahead of the curve in detecting the COVID-19 problem¹: they are secular information extractors for the most part, untainted by petty identity.
Taking this to its logical conclusion, you are saying that it is a better outcome that the CDC was wrong, as some tech luminaries made money, than if they had not made money and the CDC had provided superior initial guidance, saving thousands of lives.
Revealing HHS incompetence is worth far more than some few human lives. This virus was harmless but a true pandemic would have destroyed us.
Now America knows whom to listen to. If a second supervirus hits us, one that is actually deadly, then when the HHS lies to the populace to save who it seems the chosen few, the populace won't listen, and society will correctly handle things.
Salem Witch Trials.
It’s often not advantageous to know the truth, especially when the truth is complicated or uncomfortable. It will be rejected or ignored.
You frequently have far more of a competitive advantage if you can make others believe something, whether it’s true or not.
Outside of some very specific domains (like investing, but even markets are irrational), I don’t think it’s accuracy that makes information powerful, it’s emotion.
Though one could argue that there is a useful information extractor necessarily knows to kill others as witches to preserve themselves. Places one on weak footing, though.
By clarity, I mean fact and opinion very clearly differentiated.
Facts are common.
For a given story:
There are facts.
Some of those may be disputed. Fine. Great discussion to have.
At some point, we arrive at opinion, and this generally is what the authors believe the facts mean.
Bias colors all of that. There is always bias.
Always.
The emphasis on clarity gets at the influence of bias by making it easy to understand what the bias actually is.
Currently, we largely sidestep bias with various, low clarity arguments:
Official
Fair and Balanced
Objective (this actually takes a number of us working together over a sustained time to do. You can be 1000 percent sure it does not ever happen in a cable news cycle. Ever.)
Size, with 3 and up to few letter big players attempting to be reputable because reasons.
You get the idea.
Clarity is powerful. It implies accuracy, helps people to understand bias and when bias is different from stated intent or claims of authority, or publication of record, note, stature.
Did I mention there is always bias?
Always?
A great example in the US might be coverage from the labor point of view versus big business point of view.
Larger, established organizations rarely publish or broadcast from the labor point of view. Low clarity material makes understanding that as well as what is fact and what is opinion very hard to discern.
Small orgs, indie media, often does more from the labor point of view, and lower clarity beings the same difficulty.
Many more examples abound!
Low clarity facilitates endless meta too:
Who is objective? (Nobody in reality, but they all say "the other people" are "a problem" or "are untrustworthy" somehow.
This is all expensive and useless.
I could go on.
Greater clarity. It will help. By nature, it is more accurate. It makes bias more easily understood.
And on that point, bias is OK! It takes too many of us too long to be objective. We can benefit from well understood bias and seek multiple points of view to help us be informed and form our own opinions.
Clarity requires disclosure.
It will also force truth in branding.
"News you can trust" means nothing when it is an unclear mess of fact, opinion, bias all mashed together to get a reaction, is it?
Nope.
I submit a push for clarity does the most good.
I say that, because we can't solve the thinking for people problem. We can Empower them to think better, and we can give them more clear material to think with, and we can foster discussion that encourages common ground, but we can't actually think for them.
And just as an example, when I was a little kid in primary school we had a media class.
That class covered the basic types of propaganda, with advertising as a vehicle to demonstrate all of them. It had a topic on bias, identifying the point of view from which the piece was written and why that matters.
It even talked about clarity. In any given piece, are fact and opinion well differentiated?
Where they aren't, that piece isn't very reputable, or useful, or the minimum you need to do is seek more information if you're intrigued at all.
We read articles from the labor point of view, we read articles written from overseas, ones from the local newspaper, from big business, and we talked about them and we found the common facts and we found out how to think for ourselves.
That was seventh grade. In a small backward town no less!
I am regularly shocked at how poor our media is today, and how ill-equipped so many of us are to deal with it.
All of these rule based schemes avoid both the clarity and trust problems inherent in all media.
Secondly, they sidestep the intended discourse!
Not only are we supposed to be thinking for ourselves, but we are to be understanding one another well enough for policy discussions to make better overall sense!
All of this is why we used to teach critical thinking in primary education.
Today, we do not do that very well and look at the mess!
Software won't fix this. Humans doing the work to improve will.