@Gargron no problem

@Gargron though i do wonder why my instance doesnt show up in the lgbtq category on joinmastodon

@Laurelai did you put it in there on instances.social admin panel?

@Gargron Ah i forgot to tag it as lgbt, fixed now lol

@Riley hmmm did a germanism sneak into my speech? intransparent sounds very natural to me....

@gargron I have never heard it. A quick google tells me it's probably a germanism.

@clacke @Riley Already fixed

@gargron @clacke I would have gone with the Minotaur, but you know what you're doing. :P

@cathal @Gargron

> How do we harden ourselves against #Twitter / #Facebook-style manipulation without damaging the social fabric?

A proposal: https://instagov.com/wiki/Peer-to-Peer_Accountability_Enforcement (written before I first encountered Mastodon; I could add suggestions on how it might be implemented here)

@woozle @cathal @Gargron

@edsu can be interested I guess.

@woozle @Gargron Interesting, but it strikes me as the kind of solution that could end up worsening the problem, in a very context-dependent way. e.g., if the implementation was "anyone I follow can be pseudotrusted to rate others", then that's just a broken model at the outset. If it's "I mark people as trustworthy manually", what about people who put themselves forward as trustworthy critics, then impose their biases to silence undesireable opinions? &etc..

@cathal @Gargron

Trust would be explicitly set, not implied by followership.

People are welcome to put themselves forward as trustworthy, but that doesn't mean you have to trust them. I think most people are *less* inclined to trust someone who seems to be trying to convince people to trust them.

@woozle @cathal @Gargron The whole delete FB gathering of data was built on people trusting an app... that then sent the data to Cambridge Analytica...

@shellkr
@woozle @Gargron And, Mastodon exposes most user behaviour in a way that makes sneaky methods like that unnecessary: a mostly straightforward crawler can collect most user interactions without trouble. That's ~fine, as all that is meant to be public. But in the wrong hands, full datasets like this can give undeserved power to predict and manipulate.

@woozle @Gargron Another, concrete case: neofascists have a myth that Irish people were subjected to chattel slavery. I once engaged a very active debunker in what I felt was honest discussion; that while economically-forced indenture isn't chattel, it's still a "lite" form of slavery. He immediately assumed I was a neofash and blocked me. Under the P2P-AE model, his knee-jerk response could become a propagating label of "fascist" that I never earned.

@cathal @Gargron

Only if a lot of people trust him.

...and even if they do, then it would have to be sufficient to overcome the existing trust people have in you.

...plus presumably the debate would be on record, and people could judge for themselves. (Not sure how well that fits into the Mastodon model, though... but then, Masto really isn't designed for debate.)

@woozle I don't think a platform being designed or not for debate has ever stopped people using it for debate! :)
I'm not against networks-of-trust in principle, by the way. I just think that implementation has to be super sensitive, and realistic, towards real-life humans. Who are generally a lot dumber than nerds give them credit for, when it comes to "sheeplish" behaviour. Unfortunately, plenty of people _do_ trust people who ask for trust: they're called Politicians!

@cathal And it may be that P2PAE isn't the right model for Mastodon either -- but I thought I'd put it forward as an idea that's at least been baked for awhile.

@woozle Well, the more one tries to find objective sources of "trust", the more despairing one becomes. Webs of trust combined with ways to trace a discussion back to primary, verifiable data are really the only "trust" we can have. So it's inevitable that we must use them. But done wrong, they can backfire spectacularly.. see also, Youtube declaring that they'll send hoards of conspiracy theorists to Wikipedia articles. Good intentions perhaps, but idea is undercooked.

@cathal I don't know about you, but I have a pretty clear idea of who I trust and who I don't... and verifiable data is only part of the picture.

@woozle But, "trust" doesn't translate through "who" to "what", right? I "trust" X family member, but I don't "trust" their opinion on vaccines, gluten, genetic engineering, or autism. I "trust" that they'll act in my best interests according to their world-view, but I wouldn't let them screen my toot-stream for me.

@cathal I agree that a finer-grained system with user-maintained categories would be more accurate... but also would involve a lot more work from each user, and might drive people away from using the system at all... and makes it harder to explain the basic concept to anyone in the first place.

I had planned to add ratings categories later, as an option.

@woozle Exactly; it's not that the problem can't be solved.. it's that any technical solution that captures the problem accurately requires the sort of education that would probably make people largely immune to fake-news anyway. I'm increasingly left worrying that there _are_ no technical solutions to factual-trust, and that we're left with the improbable; educating our way out of mass-credulity.

@cathal So the thing is to find a solution that's *good enough* to reduce the problem to something endothermic.

Mastodon already does this to a large extent -- the mechanisms of federation can be viewed as a form of delegated trust, in that you trust your instance sysop to make good decisions about other individuals and instances.

Gargron is arguing that we will need more, as Mastodon becomes more of a target, and I agree -- but no solution will ever be perfect.

@woozle That could also be viewed as a mechanism of network trust fragmentation, as I'm sure plenty of discussion has already trodden. But just on the subject of _profiling_ users.. the decentralised nature of Mastodon makes that far easier than with Twitter, for example. Making a user on lots of open servers and watching the public timeline, or scraping small and un-rate-limited networks to get a look-in at the user ecosystem (e.g., harvesting all the users mentioned).

@woozle I do appreciate the basis though; that we only need to make it more expensive than it's worth. That's really the goal of all anti-mass-surveillance measures; make it too costly to bother with. And perhaps we can do that, especially for specific topics of discussion; narrow domains of politics, vaccines, genetic engineering, global warming. I can imagine ways to auto-intervene in those common sources of misinfo, and help provide unbiased information.

@cathal @gargron Thank you for this.

Could we please have a way to block bots by default, rather than individually, one by one? It's not a solution, but it would drive up the cost of a psy-ops campaign.

@jannamark And how will we distinguish the bots from actual humans? :)

@cathal

I do not pretend to be an expert in bot detection, but there are some out there.

@jannamark There are, but they range from solutions that impose mass-surveillance (invasively probing visitors' browsers, storing that info, correlating devices), to those that simply ban things that don't look like browsers (and screw all the outliers, like blind readers), to sketchy solutions that rely on referer / request history information (again, surveillance..).. nothing's perfect. Some of these measures, in careful doses, are useful. But in the main, you can't tell people from bots.

@cathal I'm going to think about this a bit, because, overall, I found I could tell which were bots and trolls yelling at me on twitter and facebook, and which were people with different opinions, who wanted to engage.

Also, I agree/understand that there's a difference between the ham-fisted but successful efforts of a Trump or Brexit campaign, and a subtle gas-lighting or whisper campaign more commonly associated with the psychological manipulation now relegated to bots.

@jannamark That should be a testament to how amazing your brain is, because computers would find that sort of distinction extremely hard, in a realistic war-on-bots scenario. Twitter's bot problem has many cases that would be trivial to fix, but they never bothered; that's not the kind I'm concerned with, though. They basically didn't fight back, so the attackers never bothered with sophistication. Usernames like "frank92754719" are laziness, not a required bot feature. :)

@cathal @gargron CA only managed this on Facebook because Facebook looked the other way. It would be tremendously difficult to scrape the kind of data Facebook handed them without someone noticing.

I've seen a scraping operation identified and stopped on here in real time. An admin noticed, mentioned it, people shared the toot, and a movement to identify and block them formed.

They would have to run a fake node and manually federate with everyone slowly, and they would always be behind.

@Riley @Gargron I'm aware of that case, and I'm telling you, professionally, that that kind of amateur scraping is not what I'm talking about. For starters, one does not even need to be logged in to get Masto data. And the proliferation of small servers makes it trivial to enumerate users on larger servers, to prepare a deeper crawl. Update frequency prediction can be used to make it drastically more efficient, if needed. Standard proxies and crawl strategy: go. It's all fairly straightforward.

@Riley @Gargron And, yes; fake nodes are a problem that I don't think is fully solved by using manual bans. Because, setting up and artificially populating realistic-looking nodes is not going to be hard, either; only a small minority of the fake users need to be run by bot-farmers to disseminate fake news and propaganda. And replacing the nodes would be cheap. Federation? Easy; just make "authentic" accounts on public mastos, and follow some quiet/harmless bots to get the instances on the net.

@Gargron @Riley With careful separation of concerns it should be possible to create instances that quietly act as 'bridges' but don't engage in obviously abusive behaviour, and use more ephemeral/disposible servers through these bridges. Does mastodon currently track the providence of messages through the federation network? This could help to map out these networks, if/when they arise.

@Riley (I should add, by the way, that Scrapinghub doesn't engage in illegal activity and wouldn't, I think, engage in mass-fakery to collect data. We only do legal scraping, and we're not a bot-farm. But through working in this line of business, you do get a good eye for what a less scrupulous operator could do, and for how much effort.)

@cathal @Gargron Great, it's about time we start fighting back against data harvesting, too. Also time to decide how to prevent data we produce to be harvested on the other servers of the fediverse. What if we started with discussing forbidding the indexing of remote toots by search engines ?

@npze @Gargron Well, this is something I've wondered about for a while; can toots contain meta-preferences that follow them through the fediverse, indicating things like "Please only show this toot to your logged in users, who follow me", or "Please don't let search engines index this", or "Please don't retain this for longer than X". These would all have to be "please"; there's no way to force it. But some of them can be externally verified by other servers, to check for compliance..

@cathal @Gargron If you think that the manipulation started with Twitter and Facebook, you have no hope of resisting it. Our feelings and beliefs have always been manipulated, and the defense has already been the same: learn to distinguish truth from fantasy. As long as people are willing to blindly repeat shit without putting any effort into confirming it for themselves, they will be vulnerable to manipulation.

@Gargron @cathal Before people even had email there were chain faxes going around spreading fake news about things like gang initiation rituals where they'd drive around with their headlights off and shoot the first person who flashed their lights at them.

@cathal @Gargron And before chain faxes there were chain letters and mass paper mail. The mass paper mails even went so far as to look like newspaper clippings to make them seem more news-like.

@Gargron @cathal My great grandfather used to send my father that kind of crap all the time.

@seanl Agreed, that misinformation isn't new; I don't buy into the narrative that "Fake News" is this new threat. But this starts to look a little like the gun issue, after a while. Sure, someone can kill using a breadknife, so why not guns? Because they're more efficient, that's why. And automated profiling of credulous targets, and automated creation of content for specific credulous demographics, is vastly more efficient than faxing bullshit to a bunch of offices. Dangerously so.

@cathal This is what Bruce Sterling's book Distraction is about, by the way. Not specifically about social media but about the use of data and psychology to figure out how best to influence people.

I think part of the problem is that the left believes that its own message is somehow not propaganda, that what they're selling is "truth" and what the right is selling is "lies." We need to accept that truth and data don't change minds; good stories (i.e. propaganda) do.

@cathal As long as we continue to delude ourselves into thinking that our own beliefs are self-evident because they're "true" and that we don't need to use propaganda techniques, we have no chance against people who realize that everything's propaganda.

@cathal There's a story on Michael Crichton's Wikipedia page about how he wiped the floor with a couple of climate scientists in an Intelligence Squared debate. They said the audience preferred his flashy style to their dry statistics. This is a room full of people who self-select to believe they're convinced by truth and data. And they weren't.

@cathal There's another way to look at the problem, too, which is to focus on innoculating the people around us rather than trying to think about everyone at once. We all have that family member or friend who's a little bit too credulous about what they share. Don't just ignore them or smile and shake your head when they do that. Teach them how to fact-check. I was able to successfully do that with my mom, who used to think deodorant caused breast cancer back in the mass email days.

@cathal Now my mom is the one posting links to Snopes articles on other people's Facebook shares.

@seanl Hey, nice work and good job by your mum! :) As far as the fact/false, right/left.. I don't think political affiliation per-se has anything to do with the tactic, though it _is_ true that it's more widely used by the American/UK far-right right now because of Mercer's influence. And, facts _do_ exist in the world, just not in all domains. And many of the ambiguous domains become the most hotly debated! e.g., There are no "facts" about when human-ness begins, so the abortion debate is toxic

@seanl So if I see value-based stuff about Abortion that doesn't ring true for me, I don't automatically label it as "false" unless it makes directly refutable statements. But if I see people trying to gerrymander the meaning of "homeless", it's mostly unambiguous whether someone has a home or not and we have good figures for that in most places. So I can call bullshit without relying on value-based judgements. Facts do have political dimension, when they exist at all.

@cathal @Gargron I think it needs to be easier (i.e., built in) for admins to review and communicate about silenced instances, to facilitate addressing common concerns. Individual admins still must make the decisions which instances to silence, but more information needs to be readily available to make timely, informed decisions.

@cathal @Gargron I think there are two problems here. People using advanced tools to create fake accounts and manipulate users opinions on a social network, and building a social network that's literal monetization model is based on this. The former is a hard problem, the latter is just insane.

@Gargron @cathal The optimistic part of me believes that on a network like mastodon, influence isn't directly sold to advertisers so the structure of the network will grow flatter with less of a focus around celebrity (in the loose sense of the word) accounts. Therefor, the cost of influencing goes up. I guess that is the defense right? Encourage a network structure where there are relatively few large influencers (weakpoints in a way) to drive up the cost of influence.