You cannot police misinformation in a fair, reproducible, and representative way(twitter.com)
twitter.com
You cannot police misinformation in a fair, reproducible, and representative way
https://twitter.com/dnunan79/status/1491072159393148929
417 comments
"Widely available, authoritative sources" is very similar to Wikipedias definition for reliable sources, and as anyone who been active there knows, the process is neither scientific nor reproducible. You can vote on it, you can form a temporarily consensus, you can elect people to make a decision for the community, but at the end its about convincing people through discussions. Naturally there exist edge cases where the conclusion is close to 99.9% determined from the beginning, like with spam and malware, and at the other end there is the small sliver of "common knowledge" that people do not disagree about (ice cream is delicious, kittens are cute, bad things are wrong).
Remove the discussion and give the power to decide what is truth to a company and what you get is the consensus within that company. Naturally that won't be fair, reproducible or representative. Give the power to governments and you get similar results, just with more police and military, through with the slight bonus of people being at least being elected. Still a pretty pointless exercise unless discussion and transparency exist.
Remove the discussion and give the power to decide what is truth to a company and what you get is the consensus within that company. Naturally that won't be fair, reproducible or representative. Give the power to governments and you get similar results, just with more police and military, through with the slight bonus of people being at least being elected. Still a pretty pointless exercise unless discussion and transparency exist.
I find on Facebook these days I often have to use some substitute Unicode characters and obscure out words in pictures if I want my post to be seen by my own friends. Even if it's a scientific, reputable piece of content, all it takes is a word like "COVID" or "election" to trigger their censorship system and my post doesn't show up in their newsfeeds.
I'm tired of this censorship bullshit where companies play gatekeeper between me and my own friends. Even WeChat in China doesn't censor this much. Almost always if I post something on WeChat my friends see it in their feeds within seconds. But on Facebook, nope. Their algo stands in-between and decides whether or not it should be censored from each friend.
I'm tired of this censorship bullshit where companies play gatekeeper between me and my own friends. Even WeChat in China doesn't censor this much. Almost always if I post something on WeChat my friends see it in their feeds within seconds. But on Facebook, nope. Their algo stands in-between and decides whether or not it should be censored from each friend.
Just like the last couple of years have turned us all into amateur epidemiologists, the last handful of years have turned us all into amateur epistemologists. The very foundations of how we as a society collectively decide on what reality is, have come under an adversarial attack. So now we're living through the societal equivalent of realizing that we've been storing passwords in plaintext and have been hacked. We're trying to minimize the damage and fallout, and stop the damage from happening further, all the while trying (and failing) to invent a set of best practices to prevent this from happening again.
Each community decides what is "fair" in that community. Each community "represents" a slice of people. People leave communities that don't seem fair to them, that don't make them feel represented.
TL;DR: Tribal people have mores. That's how tribes, people, and mores work.
TL;DR: Tribal people have mores. That's how tribes, people, and mores work.
Plausibly true...but on large social media platforms, you could at least get slightly sane about how much scale matters. If I share videos / photos / captions of my great-grandsons enjoying a bubble bath with a few older in-laws...the Brain Police do NOT need to check what I say about COVID, or whether a 1-year-old boy's nipple can be seen through the bubbles. Vs. if some content is suddenly being viewed by thousands or millions of people - maybe that's where 99.9% of the fact-checkers and well-intentioned censors should be focusing.
And maybe show some humor about it. Say, tag the COVID-fiction stuff with "This content is cool and getting really popular, but it seems pretty weak in the Facts Dept. If you enjoy it, you might also want to check out Night of the Living Dead, An American Werewolf in London, Shadow of the Vampire, and The Andromeda Strain. Sign up for Horror Classics now, and get your first month free with discount code 'COVID'..."
And maybe show some humor about it. Say, tag the COVID-fiction stuff with "This content is cool and getting really popular, but it seems pretty weak in the Facts Dept. If you enjoy it, you might also want to check out Night of the Living Dead, An American Werewolf in London, Shadow of the Vampire, and The Andromeda Strain. Sign up for Horror Classics now, and get your first month free with discount code 'COVID'..."
This is well laid out, but it absolutely freaks me out that we even have to explain to very smart people why free expression of ideas is important. I know it's human nature to persecute, but I hate being reminded of it so often.
Bullshit.
In the beginning there was the word. Next there was the lie.
All throughout human history we’ve tried to figure out what actually is. The key component of that is figuring out what information we can trust.
We can never simply throw our hands in the air and say “it can’t be done” “it can’t be fair” or any othe cop out. We all do it every minute of every day. We can do it better, we must!
All throughout human history we’ve tried to figure out what actually is. The key component of that is figuring out what information we can trust.
We can never simply throw our hands in the air and say “it can’t be done” “it can’t be fair” or any othe cop out. We all do it every minute of every day. We can do it better, we must!
Right.
But IMO that doesn't mean you shouldn't try.
It's not perfect, yup, mistakes will be made. I don't buy into the idea that you thus shouldn't do it.
But IMO that doesn't mean you shouldn't try.
It's not perfect, yup, mistakes will be made. I don't buy into the idea that you thus shouldn't do it.
The same is true for gore, porn, or other types of undesirable content, yet hardly anyone wishes to post on platforms that make no effort to control those.
This should be fairly obvious? There have been myriad occasions where we have collectively discovered that certain "facts" are not, in fact, facts at all.
Given that the reality of reality is beyond the grasp of certainty, a platform owner still is able to make choices about what is and is not acceptable within their community. This is well understood in indie game community management; nurturing nurturing communities involves removing bad actors, and making _subjective_ calls about whether people are acting in good or bad faith.
Given that the reality of reality is beyond the grasp of certainty, a platform owner still is able to make choices about what is and is not acceptable within their community. This is well understood in indie game community management; nurturing nurturing communities involves removing bad actors, and making _subjective_ calls about whether people are acting in good or bad faith.
You're telling me that companies publishing information have to take an editorial stance on what they publish? I'm shocked!
More seriously, even though there's no "closed form" solution, it seems like each publisher can take a reasonable subset of what they can stomach. If "good information" is an amorphous, moving blob floating in a sea of "bad information" we can draw simple polygons that are more/less inclusive of "good information" while allowing more or less "bad information" to be included.
More seriously, even though there's no "closed form" solution, it seems like each publisher can take a reasonable subset of what they can stomach. If "good information" is an amorphous, moving blob floating in a sea of "bad information" we can draw simple polygons that are more/less inclusive of "good information" while allowing more or less "bad information" to be included.
From the comment section:
https://twitter.com/dnunan79/status/1491073564447244288/phot...
It is interesting, how the concept of misinformation is bound to time. Imagine 1 year ago, somebody would say:
"if you are vaccinated, you are still able to catch covid and spread covid", or "cloth masks don't work", or even the first "misinformation" from the whole pandemic "covid was leaked from a Lab in Wuhan"
All those 3 "misinformation" and many more were policed pretty hard until 2021.. Today 2022, they were either proved to be true or have high probability to be true.
https://twitter.com/dnunan79/status/1491073564447244288/phot...
It is interesting, how the concept of misinformation is bound to time. Imagine 1 year ago, somebody would say:
"if you are vaccinated, you are still able to catch covid and spread covid", or "cloth masks don't work", or even the first "misinformation" from the whole pandemic "covid was leaked from a Lab in Wuhan"
All those 3 "misinformation" and many more were policed pretty hard until 2021.. Today 2022, they were either proved to be true or have high probability to be true.
I was expecting a link to a paper that would be similar to results like in "On the impossibility of fairness" by Friedler et al.
This twitter thread seems defeatist and hand-wavy. The thread does have good intuitions in focusing on uncertainty and nuance. For example, the difference between asking questions and sealioning is hard to tell
Because of that, I agree that a perfect solution probably doesn't exist because of similar trade-offs like in the paper I mentioned, but doing more research at least lets us find out what we are trading-off and what we are willing to sacrifice.
This twitter thread seems defeatist and hand-wavy. The thread does have good intuitions in focusing on uncertainty and nuance. For example, the difference between asking questions and sealioning is hard to tell
Because of that, I agree that a perfect solution probably doesn't exist because of similar trade-offs like in the paper I mentioned, but doing more research at least lets us find out what we are trading-off and what we are willing to sacrifice.
It might be easier if we stopped considering twitter and instagram as suitable places to publish nuanced, serious stuff that was liable to misinterpretation and misrepresentation
Moderation/censorship/whatever you want to call it is a heavy-handed solution to a societal problem. The fact is that science isn't a linear process and the general consensus will flip all the time and this is normal. Coordinating a public health response while the science is still being conducted in real time necessitates some educated guesses which might be wrong. An effective response also requires adaptation in the face of new evidence.
All of this uncertainty is in direct opposition to how large groups of humans behave. There was a HN post recently using MSFT Azure as an example about why executive/C-level communication is so repetitive and basic; you simply can't announce something to a thousand people and expect them to understand context and nuance. The example there was that the execs zeroed in on new features above all else, resulting in tons of bugs. But trying to focus on both features and bugs was a lose-lose, people would find excuses to not do things using whichever focus was more convenient.
Executive Q&As at big companies are another example. I'm not sure why we still do them. It's simply not possible to have a deep dialogue in this environment. You wind up with disgruntled employees tossing up call-out questions, CEOs dodging any sort of concrete answer, and nobody walks away better informed. Let me be clear, I loved these at a smaller company where the CEO actually knew me personally, but they are a colossally cringe waste of time at a 2k+ person company.
The big platforms are also in direct opposition to one of the solutions to this problem, which is breaking down communication to small groups of people that actually know each other, and rationally discussing context and nuance is possible. Twitter is built around having the entire internet respond to a tweet. Of course this results in chaos.
Moderation is the only solution. It doesn't have to be balanced, as long as people stick around and keep using the service.
All of this uncertainty is in direct opposition to how large groups of humans behave. There was a HN post recently using MSFT Azure as an example about why executive/C-level communication is so repetitive and basic; you simply can't announce something to a thousand people and expect them to understand context and nuance. The example there was that the execs zeroed in on new features above all else, resulting in tons of bugs. But trying to focus on both features and bugs was a lose-lose, people would find excuses to not do things using whichever focus was more convenient.
Executive Q&As at big companies are another example. I'm not sure why we still do them. It's simply not possible to have a deep dialogue in this environment. You wind up with disgruntled employees tossing up call-out questions, CEOs dodging any sort of concrete answer, and nobody walks away better informed. Let me be clear, I loved these at a smaller company where the CEO actually knew me personally, but they are a colossally cringe waste of time at a 2k+ person company.
The big platforms are also in direct opposition to one of the solutions to this problem, which is breaking down communication to small groups of people that actually know each other, and rationally discussing context and nuance is possible. Twitter is built around having the entire internet respond to a tweet. Of course this results in chaos.
Moderation is the only solution. It doesn't have to be balanced, as long as people stick around and keep using the service.
Relevant: Brandolini's Law
"The amount of energy needed to refute misinformation is an order of magnitude larger than is needed to produce it."
It can take ~15 minutes to create a Twitter post containing misinformation, but a team of fact checkers can invest hours/days/weeks into the post. The fact checkers research the topic and write a refutation that's well-sourced and hopefully as objective as possible. Meanwhile, there are 1,000 more posts freshly minted with exciting new misinformation.
"The amount of energy needed to refute misinformation is an order of magnitude larger than is needed to produce it."
It can take ~15 minutes to create a Twitter post containing misinformation, but a team of fact checkers can invest hours/days/weeks into the post. The fact checkers research the topic and write a refutation that's well-sourced and hopefully as objective as possible. Meanwhile, there are 1,000 more posts freshly minted with exciting new misinformation.
You can't police society in a fair, reproducible, and representative way either, but how many people want to get rid of police entirely? Only a tiny fringe.
Sometimes the grossly imperfect solution is still an improvement on no solution at all.
"Don't bother policing misinformation" can't have methodological flaws the way moderating can, because it's just the absence of something. But ultimately what I think most people care about isn't soundness of methodology, but the results, and so the question you have to answer is: is a popular platform that makes no effort to control misinformation healthier for society than ones that do make that effort?
Sometimes the grossly imperfect solution is still an improvement on no solution at all.
"Don't bother policing misinformation" can't have methodological flaws the way moderating can, because it's just the absence of something. But ultimately what I think most people care about isn't soundness of methodology, but the results, and so the question you have to answer is: is a popular platform that makes no effort to control misinformation healthier for society than ones that do make that effort?
You have to revive a culture where truth is paramount. Unfortunately, I think a lot of cultures that value "truth" also tend to shoot down complexity in favor of simple, but wrong, "truths". For example, when it comes to complex topics like sex and gender, people tend to gravitate to either "There are only two sexes, a biological "fact" somewhere between, e.g. bimodal but overlapping distributions.
It's also good to remember that ideas like the heliocentric theory, the germ theory of disease, etc. would have been labeled as "misinformation" at one point in time. Now, they're the theories believed by the majority of people because that's where the best evidence points, which just goes to show that sometimes it takes hundreds of years for the best ideas to win. So I say let them fight.
The Cochrane review that was removed agrees with the authoritative sources that Instagram uses. This seems to be a case of misapplication of policy and doesn't support the OP's thesis that the policy itself is bad.
I would argue that the Cochrane Library itself should be considered an authoritative source, but again, that is a problem with application of policy rather than a problem with the policy itself.
I would argue that the Cochrane Library itself should be considered an authoritative source, but again, that is a problem with application of policy rather than a problem with the policy itself.
This reminds me of something out of one of Neal Stephenson's latest fiction novels, Fall; or Dodge in Hell [0].
The concept of the "web" devolving into a deluge of information ranging from completely false to purely factual that's almost unusable without a moderator. In-universe, moderated streams are available by subscription/hire, with inexpensive "AI" or shared streams with little personalization, all the way up to dedicated moderators that act more like a personal assistant constantly monitoring and updating your personal feed.
It sadly seems like a highly prescient view of what our modern-day internet is becoming. It's taking longer and longer to review multiple takes on a subject to get a full picture and failing to put in that due-diligence makes it hard to shake that feeling that you're just falling into a party-line decision on a subject as more "news" outlets become highly polarized.
Some days I just have to "run dark" on news to get a breather, but then everything seems all the more overwhelming when I get back online and see what new mess has erupted in the news cycle.
[0]: https://en.wikipedia.org/wiki/Fall;_or,_Dodge_in_Hell
The concept of the "web" devolving into a deluge of information ranging from completely false to purely factual that's almost unusable without a moderator. In-universe, moderated streams are available by subscription/hire, with inexpensive "AI" or shared streams with little personalization, all the way up to dedicated moderators that act more like a personal assistant constantly monitoring and updating your personal feed.
It sadly seems like a highly prescient view of what our modern-day internet is becoming. It's taking longer and longer to review multiple takes on a subject to get a full picture and failing to put in that due-diligence makes it hard to shake that feeling that you're just falling into a party-line decision on a subject as more "news" outlets become highly polarized.
Some days I just have to "run dark" on news to get a breather, but then everything seems all the more overwhelming when I get back online and see what new mess has erupted in the news cycle.
[0]: https://en.wikipedia.org/wiki/Fall;_or,_Dodge_in_Hell
We can't have a perfect voting system due to Arrow's Theorem, but elections are still preferable to alternatives.
Unrelated but how does one sign up "anonymously" to twitter these days? I tried one of my twilio phone numbers, one of which even worked with facebook but they're being rejected by twitter. Is there a way to sign up without a phone number?
It's called moderation.
Moderation has been working fine for public forums since the internet gained popularity.
The problem here is that no one can moderate a community of billions. That's the unfeasible scenario that Facebook, Twitter, etc. find themselves in.
Moderation has been working fine for public forums since the internet gained popularity.
The problem here is that no one can moderate a community of billions. That's the unfeasible scenario that Facebook, Twitter, etc. find themselves in.
A more accurate title would be "You cannot police Covid misinformation in a fair, reproducible, and representative way" You can't because it's relatively new, still being researched and there are varieties of scientific opinion about various things. The debate is ongoing.
You can police Holocaust misinformation in a fair, reproducible and representitive way
You can police Holocaust misinformation in a fair, reproducible and representitive way
Yeah I'm going to have to categorically disagree with that. You can totally police misinformation in a fair means by requiring that it presents credible evidence of weight commensurate with the impact of what it argues, and explain to those who are policed in a transparent manner why and how they can appeal it.
There is no one on earth who can form a credible argument that 5G spreads SARS because it's fucking lunacy. The evidence for ivermectin improving severe covid outcomes also clearly showed it was only noticeable in regions where untreated parasitic worm infections are common, so yeah, getting rid of multiple other infections led to better outcomes. Various counter-arguments against anthropogenic global warming (even the honest ones) have been thoroughly invalidated for decades.
It's like raising the required work to send spam or spray malware at random internet hosts. It doesn't fix everything, but it evens the playing field between trolls and agents of discord vs people having real discussions. We know this works because high-value discussion forums like the heavily moderated science subreddits do it and there is no mass hysteria -- and reddit should probably be compensating those mods, but that's another issue.
If your commercial platform does not put in the bare minimum content moderation to keep dangerous horseshit out of the common discourse where gullible and naïve people with no ability to serious discern credible statements from well-practiced lies, there should be penalties scaled to the revenue of the business.
Just because Facebook or whomever clearly didn't give a damn and phoned in some half-assed keyword flagging system doesn't mean it's impossible. It means Facebook doesn't care and needs to get fined until they do. If you can't link some credible evidence because of stupid limits on Twitter, maybe Twitter is just a terrible discussion platform.
There is no one on earth who can form a credible argument that 5G spreads SARS because it's fucking lunacy. The evidence for ivermectin improving severe covid outcomes also clearly showed it was only noticeable in regions where untreated parasitic worm infections are common, so yeah, getting rid of multiple other infections led to better outcomes. Various counter-arguments against anthropogenic global warming (even the honest ones) have been thoroughly invalidated for decades.
It's like raising the required work to send spam or spray malware at random internet hosts. It doesn't fix everything, but it evens the playing field between trolls and agents of discord vs people having real discussions. We know this works because high-value discussion forums like the heavily moderated science subreddits do it and there is no mass hysteria -- and reddit should probably be compensating those mods, but that's another issue.
If your commercial platform does not put in the bare minimum content moderation to keep dangerous horseshit out of the common discourse where gullible and naïve people with no ability to serious discern credible statements from well-practiced lies, there should be penalties scaled to the revenue of the business.
Just because Facebook or whomever clearly didn't give a damn and phoned in some half-assed keyword flagging system doesn't mean it's impossible. It means Facebook doesn't care and needs to get fined until they do. If you can't link some credible evidence because of stupid limits on Twitter, maybe Twitter is just a terrible discussion platform.
I think if you truly want to battle misinformation then you can't use the approach of dragging people kicking and screaming to The Truth. Rather, help people find it for themselves. Give people empowering tools to find the truth whatever it may be.
Three examples: Formal logic, statistics, and how to detect deception. All of these are general purpose. Teach people about fallacies, what an appropriate sample size is and make them play Diplomacy. Teach about historic hoaxes and how they were disproven, as well as about conspiracies, ones that were successful and others that were uncovered and failed. Tell the stories of corruption and coups.
For technology companies, build un-opinionated features that users love and work so well that people opt-in.
Three examples: Formal logic, statistics, and how to detect deception. All of these are general purpose. Teach people about fallacies, what an appropriate sample size is and make them play Diplomacy. Teach about historic hoaxes and how they were disproven, as well as about conspiracies, ones that were successful and others that were uncovered and failed. Tell the stories of corruption and coups.
For technology companies, build un-opinionated features that users love and work so well that people opt-in.
It will be interesting to watch people rediscover the value of freedom of speech. We aren't at the turnaround point yet though.
Always striving to collect and share the best and most compelling observations and evidence can go a long way.
People just don't trust authorities anymore, because authorities time and time again have misled them.
That's not to mention that it's gotten easier than ever to falsify evidence using technology that was never before available in history, like image editors, audio editors, video editors, and desktop publishing software.
Astroturfing and fake grassroots campaigns further undermine people's trust in what they hear and read.
At the same time people want certainty and simplicity, and when they find they live in a world where nothing can be trusted they double down and cling to whatever reflects back to them their most cherished beliefs and what vilifies those they deem evil. It might not be true, but at least it makes them feel good, and so they can pretend it must be true if anything is.
Moderation isn't going to change any of this.