Reply Spam
The AVC comments has been experiencing a wave of comment spam that has largely been replies to legit comments. I appreciate the community for flagging it and our moderators for nuking it. Keeping the comments free from spam is important to me but not an easy chore.
One idea I have, which I don’t even know if is possible in Disqus, is the idea of limiting replies to longstanding community members who are registered with Disqus and have high reputation scores.
What this will do is eliminate the reply spam, but will also make it impossible for new commenters to reply. They will still be able to leave a comment.
I think the pros may outweigh the cons.
Thoughts?
Comments (Archived):
Is reputation the number of upvotes? Is it some equation of upvotes to comments?
And unfortunately it can easily be gamed once it becomes ‘a thing’.
Well sure, but so is getting your comment viewed, liked, replied to in the current setup. For example, if I needed maximum exposure on this blog I would make sure to be in the first three comments chronologically, be a first response to a few commentors that people tend to always upvote, and reply directly to the highest upvoted comments.The upvote is a blunt tool. Why upvote something? A different but thoughtful perspective? Hoping for a quid-pro-quo? Hoping to get a response? Express agreement about a point?
Overall, earning the right to reply seems like a net good.
The ideal solution is for Disqus to tweak their spam algorithms such that killing spam is not a daily exercise (as I am the one who sees them and kills most of them lately). That said, since we’re not there….In a way, you’re hedging towards a reputation-based kind of commenting, and there’s nothing wrong with that. For it to happen, Disqus needs to be able to flag that a given commenter is placing their first comment. The flip side is that the spammers could start placing new comments too, and we’ll need to see what it yields.
Agreed. This is a Disqus issue at the end of the day.
Does Disqus have a version of KYC? It may sound too extreme for them but it seems to work really well. Marketplaces (and their payments partners) do a good job of letting people take a first action without friction (critical in my opinion), and then at some point (trigger or time-based), requiring the person to go through KYC. If they don’t pass KYC, content comes down. If they pass, they’re good to go unless they’re flagged in the future. Some communities may be fine with comments from anyone and some may want only people who have gone through KYC.
William, it’s like one of the Canadian Mounties having our back when you take them out. Happy hunting.
I do not think you should limit replies to reputation scores Fred. Better to be open to all.(w.r.t spam – couldn’t spammers initiate a new comment instead of reply ?)If you are open to more feedback, don’t find that collapsing replies feature implemented a short while back very appealing. Not a biggie…but on a couple of occasions, found that I had missed a comment because it was under the collapsed replies link. Also, and maybe more importantly, the conversational aspect of a thread seems reduced by that feature.
fwiw, i strongly agree with your points, Girish.
Opening the link is not a major effort! Think of it as maybe the surprise of the day.
I just did, else would not have seen your comment !More important I think is that the conversational aspect gets reduced…not the same as before.
I’m guessing there’s a division of types of people and how they read, those who are happy to just skim – and those who will read through everything (at least the beginning of sentences) to see if there’s anything interesting.
I appreciate your feedback but I love it.
Ok :).
yes, and I am Nigerian Prince and offer you opportunity for safe investment in no risk high reward opportunity. Please just send me your social security number and bank account number so I know where to send the money. good yes?
Collapsing no problem for me, FWIW G.Fred’s idea sounds like something Apple would dream up. A cut off nose to spite face move.
It’s unfortunate as it seems that the acquires of Disqus platform haven’t cared about improving it, however that it stopped improving and UX degrading years prior however.
Siding with Girish in this beef, particularly w.r.t. to:>(w.r.t spam – couldn’t spammers initiate a new comment instead of reply ?)
Definitely wisdom, well stated wisdom.Tough, essentially impossible, to compete with that with algorithms or even applied math.
At Seeking Alpha, we treat all new commenters as suspect. Their first 3 comments are moderated by humans. After that, if the first 3 were legit, we approve them.It’s overhead, but well worth it IMO.
That’s a good process too.
Nice. Great site too
I think limiting based on reputation could prevent the proliferation of new, potentially valuable ideas through the community. But, I’m biased as I likely have a low reputation score and have some new ideas I’d love the community to consider – so at times might be considered a spammer.Greg KieserFounder of Supersystemic.ly LLC https://www.supersystemic.ly/Author of Dear Machine: https://www.amazon.com/gp/p…
This is a valuable bias and exactly the type of commenter I would hate to see inhibited.However… are you more likely to comment (which would not be inhibited) or to reply to a comment (which would be inhibited)?
Thanks! I tend to do both.
Many other communities (like LinkedIN) require “approval” of the new participant before they are allowed to comment. I am not sure how many new people actually join? Maybe that would be easier.I support this (except the “reputation score” which i don’t fully understand. I am against systems that allow other users to effectively silence other users they do not like.
Yup, I like pre-approvals too. I just explained a bit more in another comment.
I thought the proprietor of this joint was an ‘open systems are messy but awesome’ type.
Well, by default we are all bona fide approved by DISQUS when we register there before we get into avc. So, if DISQUS is failing in letting spammers register despite their captcha, something has to be done that’s better than whack a mole on the spammers.
Or we could just do our part training AVs…Tbh, your suggested approach surprised me given your view on incumbents and fostering creative destruction. I guess it comes down to where you set the bar for new entrants and how you protect against censoring unpopular views.
https://uploads.disquscdn.c…Would the new repliers be able to tag someone they are replying to in their comment? @
Had the same thought. But this is one of those features on the “hidden menu” and not as well-known by non-Disqus-veterans — and you have to know the person’s actual Disqus i.d. which sometimes differs from the name that appears. In the end, a potential barrier — which I know has pros and cons.
Maybe some kind of probation time frame to determine if real.
Organic growth or genetic engineering? Though hardly Dolly the sheep like, often tweaks intended to do good have an adverse effect. Collapsing comments, for example, limits interaction and engagement as it requires a bit more work and time to scroll down, with little if any benefit. The comments are often the best part of the journey, even w/ the occasional snark. If you fielded a readership study, I think you’d find an awfully high satisfaction rating for AVC among readers. Imperfection is like a scar, it too is a form of beauty.
Add a downvote counter threshold. Once the community downvote exceeds a threshold, remove it.
Personally I read all of your posts, but I usually reserve commenting to topics where I feel I have something of value to add to the dialog. I’ll also screen comments to see if points I wanted to make have been made. In the latter circumstance I don’t post.I think limiting posts by seniority or reputation may artificially limit the free exchange of ideas and prevent new perspectives from being shared.
I’ll also screen comments to see if points I wanted to make have been made.Ha — I’m guessing that if we all did this the number of comments would significantly decrease!When Fred posts from New York, sometimes by the time I actually see the post on the West Coast, getting through the comments is another 30+ minutes of reading. So admittedly, sometimes will comment first and then make my way through comments during short breaks.Can usually tell when he is on the West Coast or in Utah by the timing of the post. Sometimes by the content. Guessing today is a ski day. 😉
i’m relatively active in a few communities … mostly metal fabrication, woodworking, maker related topics. the level of engagement and quality of participants / reply content isn’t far off from the what you see here on AVC.a few of them use reputation scores to limit access to features like listing items for sale … however, the primary way you establish your reputation is by participation, aka, posting.not quite sure how you’d invert that approach … sit quietly and listen for a while doesn’t feel like it translates well in a digital space.
>a few of them use reputation scores to limit access to features like listing items for sale … however, the primary way you establish your reputation is by participation, aka, posting.Good ideas.>not quite sure how you’d invert that approach … sit quietly and listen for a while doesn’t feel like it translates well in a digital space.a.k.a. lurking, which is a known behavior (and term) on mailing lists and groups dating back to Usenet.The system (e.g. Disqus) could identify lurkers by comparing date of signup with date of first comment – e.g. exceeds a threshhold of some days or weeks, at least for systems that require signup in order to comment. So maybe the problem for that approach shifts to requiring signup, i.e. not allowing anonymous comments.
A first cut guess is that (A) the press wants eyeballs; (B) in pursuit of eyeballs they go for exaggerations, excitement, shock value, scandal, lies, fake news, sensationalism, etc., (C) people who go for (B) are fools; (D) there are a LOT of fools out there or at least the newsies believe that; (E) for fools, a good information source is booooooorrrring and they leave; (F) spammers are also fools and they, too, commonly leave.So, rotten food attracts flies, but good food doesn’t!Or a blog post on honoring pro-rata rights just won’t attract the Hollywood scandal crowd spammers!That’s a guess. I’m trying to understand, if only because such understanding could be important for the positioning, marketing, publicity, etc. of my startup as it goes live. Or, AVC has, say, call it, an audience. The comments at some NBC site will also have an audience. The audiences may be quite different. I need to understand and be clear on what audience I’m going after and how to appeal to them.
The best rhetoric is availed by those who formulate arguments in the absence of evidence.For the rest of us if Disqus have a public API we could be informed:Something like this would give us the data:commentators <- “https://disqus.com/api/search?blog=avc+fields=commentators, commentCount,replyCount,spamReportCount,reputation”A few histograms and a bit of modelling and a datascientist worth their salt could come up with a “goodPlayer” score using a generalised linear model manifold.perhaps a 3-k-means analysis could then classify these into c(allowPostingOnly, allowReponses, block).From it would be trivial to set a target thresholds and let the system search to learn the balance of blocking required for a particular blog preference.The key advantage is a network effect where the more someone comments (wherever they comment) all blogs benefit.
How detrimental is the spam and what are the risks involved?If AVC is a product, then the overwhelming response has been that comments are one of the most valuable features. Therefore, the quality of the comments is crucial, but so is continual cultivation of new voices.Personally, I’d rather deal with the annoyance than inhibit new people from engaging. New commenters are part of the lifeblood of AVC.Some information we don’t have:How often does a new person initiate a comment vs. reply?Would adding a new barrier or point of resistance inhibit commenting overall?How much of a deterrent is spam to bona fide commenting or reading the comments?
Also, is it *literal* spam—ie people just trying to put links in with no relevance to the initial comment? Or is it sometimes difficult to tell whether it is spam or not? If it is literal spam, then it seems like their must be a tech solution that doesn’t inhibit new, good faith commenters regardless of a reputation score. If it is people who are pretending to reply in good faith but trying to ‘sneak in’ some spammy elements, then it is a slightly harder problem to solve—and some way of vetting them up front might make sense. It reminds me of the post on Robo-Calls — spammers are getting more intelligent about how to use every communication portal available to get through. But if it was solved for email, it should be able to be solved in other places.
Literal spam. I have lost count of the number of times after commenting on AVC I have been offered an opportunity to earn money from home or something like that. On slower billing months it’s tempting.I agree — seems there should be some way to address this — but this is outside my knowledge area of what is and is not doable. One thing that could flag possible spam is multiple dollar signs in the reply or other telltale signals.
Exactly. You could also disallow links—especially for first time commenters.
BINGO!WINNER WINNER CHICKEN DINNER!!
One of my clients does this on their community forum and it works well! I was recruiting for them and could not post a link on the forum at first, but still supported the practice. (Cryptocurrency so they get a very *interesting* and *lively* group of commenters.)
That’s what the Disqus algos look for, but sometimes they fail.
Who cares, if it limits a new commenter its a bad idea, IMO.
I saw the notification for this comment appear on my computer screen and didn’t see the commenter name right away but knew exactly who it was. :)))
Recognize the British lion by his paw!
How does Stack Overflow do it? I can’t remember the last time I saw a spam comment there.
There is missing information here.Seems like a poll (or polls) directed toward lurkers and new and infrequent commenters is in order.
You’re talking about giving power to people who demonstrated credibility. Stackoverflow took that approach and worked.
My view is that the AVC material is significantly better than the StackOverflow material. I go to AVC everyday and come back the next day; I go to StackOverflow when I have a tricky technical question about Windows, and there I usually don’t like what I get. At times, once I did discover a good answer and posted that back to StackOverflow, somehow it didn’t get treated well.Generally better than StackOverflow are the essentially curated answers on Google: A recent amazing one is getting a Brother B/W printer to shut up about refusing to print. Apparently the printer has a page counter that has to be reset back to 0. Even installing a new ink cartridge doesn’t do that. And an old ink cartridge still with plenty of ink can get the page counter to its maximum.So, the solution is to place the printer on a maple cutting board on a 400 pound block of ice, stand on your head, and follow:To reset the ‘Toner Life End’ / ‘Replace Toner’ message on Brother HL monochrome (black and white) printers:Open the front cover and leave it open while completing the following steps.Turn the printer off.Hold the ‘go’ button (or ‘start’ button) while turning the printer on. All panel lights should be on.Release the ‘go’ button (or ‘start’ button).Press the ‘go’ button (or ‘start’ button) 2 times.Pause. All panel lights should be on.Press the ‘go’ button (or ‘start’ button) 5 times. The toner light should be off (the error may be flashing). The paper light should be on or flashing.Close cover. The ready light should be the only light on. Cycle power once more.Now can print again.Might not need the block of ice!Sorry about StackOverflow.As I wrote here earlier today, there is a surprising dichotomy: Some things are really good but by fortuitous circumstances or almost by accident, and some things are really bad that try deliberately to be good.
Spam ‘n’ chips. Almost as good as fish ‘n’ chips.If your idea is technically possible i would say give it a go as a short experiment and then evaluate what happened.
Maybe I don’t recognize most spam but I haven’t noticed any more than usual. May just need time to place more filters through Disqus.I tend to agree with Donna’s thoughts here; “I’d rather deal with the annoyance than inhibit new people from engaging”.
You don’t see them because we kill them soon enough 🙂
Thank you. Can’t believe you still have time to do this. Thankless, behind the scenes service to the AVC community.
It literally takes 2 minutes because they come in a flurry at the same time. Typically same spammer hits 6-8 posts, so in a few clicks they are squished.
😉
Ah, some irony!!! I saw a lot of lines indicating deleted comments. And for those I saw a lot of good responses, apparently to user Guest.For the irony, I can’t evaluate how bad the comments of Guest were because I couldn’t read them because they were deleted!Between deleted comments and having to click to get other comments shown, say in conversations with several comments, AVC is less fun than it was, say, a year ago. I’d rather read fast and scroll than click to expose comments one at a time and then read fast and scroll. Can scroll really fast by pressing on the center mouse wheel!The only comments I did see that were obviously spam were for making the big bucks in miracle, easy to do while the baby sleeps, working at home, if only send a little money now for the kit with the secret sauce.AVC is one of the jewels on the Internet, totally gives the lie to essentially all the rest of the media — a nearly open door to Fred’s Place where somehow the conversations render nearly all the mainstream media comparatively and even absolutely lower than pond scum, on paper can’t compete with Charmin and on the Internet useless for wrapping dead fish heads.There’s also some good stuff on Hacker News, yes, on computing and startups but also from a lot of Ph.D.’s in pure/applied math, physics, some relatively mathematical parts of computer science, and electronic engineering — essentially wide open but often quite good. Who would’a thunk? And JLM’s BRC site routinely, one blog post at a time, writes championship quality essays on current, hot topics — national security, foreign policy, the US economy, US politics, and more. For more there is the James Simons Quanta Magazine.So, right away, the best of Hacker News and Quanta Magazine make, e.g., the best of Nova look from weak down to silly. JLM’s essays make the NYT, front page and Op Ed, look like the screaming of a drunk maniac. AVC commonly beats all of Bloomberg News, The Economist, and Financial Times.On Twitter I follow just POTUS Trump, FLOTUS Trump, and Speaker Gingrich — whatever they say, I want to read or hear it, and Twitter lets me do that. For explaining what the heck is going on in Congress, I’m eager to listen to Gingrich; i.e., he seems to have some really good understanding of Congress! If there is more good on Twitter, then I haven’t found it or seen links to it. I’ve never seen anything very good on Facebook or seen links to anything very good there.To be more clear, for me, the sources ABC, CBS, CNN, MSNBC, NBC, NYT, WaPo, Boston Globe, LAT, Politico, Huffpost, altogether, are less good than any one of the sources I like and mentioned above. Why? ABC, etc. work really hard, have huge staffs, spend big bucks, all for the purpose of creating pond scum; the pond scum is what they want; they are good at creating pond scum and do it 24 x 7. What isn’t there is news, objective information, thoughtful material, or good reporting from solid sources. E.g., I’m not sure how much I care personally about pro-rata rights, but I do know that at least I should know the basics. On ABC, …, Bloomberg, etc., I likely won’t see that or if I do it will be less objective than at AVC.I REALLY like classical music, and I REALLY, deeply, profoundly, bitterly HATE pop music. Well, the sources I’ve mentioned here that I like are to me like classical music, and ABC, …, pop music. GOD in of all the universe I HATE ABC, …, and pop music. JLM recently had an essay on some Rodin sculpture — REALLY nice.Beyond just my tastes, this dichotomy is astounding: Who would’a thunk? There’s plenty of evidence here of something astounding and potentially powerful, earth shaking.For Fred’s Place, I would urge it to be close to what it was a year ago.
Thanks for pointing out Quanta Magazine. I wasn’t aware of it. Liked the two articles I read after reading your post.
Maybe Disqus can add an option auto-flag all comments of first-timers? Your moderators maybe have an easier time since it is no longer time sensitive because the spam lives in their queue?
Yup, similar to the SeekingAlpha process.
If I am not mistaken, reputation scoring is available within disqus, and it is driven primarily by spamability of the commentator + information like ip address. It might be worth it to get in touch about the issue with disqus itself
“Reputation”? Is that something like what the NYT tries for?Reputation is heavily from social considerations, thinking, information, credibility, etc., i.e., is close to that thingy so tough to touch, meaning.My view is that a reputation score is doomed to be forever weak on both reliability and validity.
It appears to be literally meant to work the similarly way your spam filter works. The difference is that people abandon sites regularly, since like their funny fanfic site doesn’t pick up an audience. This leaves a ton of long tail sites where spam isn’t moderated.The best analogy is in google apps when it asks if you want to send an email to someone with a different domain.Tightening up would literally mean that instead of nuking spam, you’d check once to twice a day for people who got caught in the spam filter, the same way you should do with your emails, in which most likely no one will be caught unless they decided to do something like post 15 links. Newness isn’t usually a trigger. Having a crazy ip address and posting nonsense but not being moderated across all disqus comments usually is
IMHO it seems like fred needs a hosted CMS service. This space has been an immense opportunity for over a decade I think.
Here’s another idea: commenting by pre-authorization.1) Start by restricting commenting to existing commenters, ie pre-approve anyone who has already commented. They are clear.2) Initiate a simple process where newcomers need to register on Disqus, and “apply” to comment. The process is routed to the moderators/Fred and we quickly one-click approve. All that’s needed to screen the spammers is usually a legit email address. All spammers have weird emails and names such that they are easily recognizable. One could go an extra step by automatically validating emails (there are services that do that and they reject spammers or emails with no history).So in essence, it’s like an opt-in commenting, with grand-fathering of existing commenters. That way, the spammers will be shut out from entry.Obviously, Disqus needs to implement these workflows, but I don’t suspect they are too big to tackle, and I’m sure it would be a welcomed feature to most other blog owners who face similar issues. In hindsight, Disqus should be screening and rejecting spammers registrations with fake identities anyways, then we wouldn’t have the issue of seeing them sneak in.
Just posted something similar before reading yours. I like it.Don’t mess with grandfathers. 🙂
Perhaps, Akismet (from Automatic) can still be used with Disqus:https://wptavern.com/disqus…Or, the other (radical) option… in the spirit of “back to independent blogging” – remove Disqus and switch to WP comments with Akismet. Hopefully Disqus comments would be importable as WP comments… or you may already have syncing enabled.
What is behind the push to manage the comments lately?So what if some spammers get through? Also, so what if ppl argue about political topics?You don’t have to police every little thing. There is beauty in messiness.I think a huge part of the divide in this country right now is that people are not engaging with ideas from the ‘opposition’. I think it’s a good thing when ideas clash, when ppl argue, when new commenters post.Incidentally some of my non spam comments were flagged as spam recently on this blog.
I have nothing to do with the company, but came across it today not long after I read this post. Commento is a FOSS comment system that takes aim at Disqus. Perhaps something to look into if you wanna keep comments around.
I believe the people should be able to control what they see. Why not let them filter comments and replies using the reputation score? Then they can decide whether they want to see content from only the most reputable people or search manually through all the comments. At least this should lessen the impact of spam.However, if you use the reputation score it is important to ask, “how did this person gain their reputation?” and “Do you trust this rating mechanism?” I think reputation should be crowd-sourced and recursively dependent on the reputations of those who performed the rating.
Here’s an idea, though not be technically possible out of box. Limiting number of responses tof new people or people with low ratings (say 2 per month or something like that) could have a large impact.Trolls often get obsessed with a particular post. If they are limited to two responses, they will wind up in a position of trolling but then being refuted unanswered on same post.Perhaps low impact to regular commenters. And high benefit controlling trolls.
Disqus could generate tokens that interested people would request to the site owner or moderator. Could be by email. To post for the first time some type of registration provided by Disqus would be neccesary.BTW, I’ve seen a few deletions of posts posted by regulars. I don’t know if this is collateral damage of the new policy enforcing.I also miss LE and Philip.
I read every single post of yours and a huge fan but never usually comment.This one is important, “limiting replies to high reputation scores” will lead to Internet records data bias, young people who are more organized on the internet may never get to make a point on your blogs, which also means limiting the conversations because hey you didn’t join Disqus and were born late.I understand the problem statement here for limiting spams, the best way is by adding transparency i.e. if a comment is made with a link or to sell something it’s highlighted and can be filtered as per a user’s preference. Their are a lot of NLP solutions also to this problem, it’s about the right use case implementation.
Sensing how deeply disturbing the process of policing comments are, I hope a better solution surfaces…directionally, an exclusionary policy (form of censorship) just doesn’t feel right. The value of information exchanged reaches zenith, when the platform encourages more user participation and a broader spectrum of viewpoints.
That would work if you want a closed community with no new followers (i.e. no community). I don’t know your ultimate objective. If you just want to “preach” without any new comments, why not just disable comments altogether? (Why would you think long-term followers know more or have a better ability to add something valuable to the community? Or maybe you don’t see it as a community?) It seems one way or the other to me. A community or no community.
Sounds like a problem for Machine learning!
I’m getting worn down by ‘Comments’ AVC posts. Not because I think the topic is unworthy or because these decisions are easy, but because I’m not seeing a principles-first approach to decisions (and implementations), and we’re going in circles.This is not crypto. It is not a National Park or a ‘public good.’ Its policies are centrally-owned (by Fred). It is not ‘for everybody.’ Fred uses the analogy to a bar, and at bars I frequent, they ‘reserve the right to refuse service to anyone.’The person with whom I want Fred to speak is Nick Denton. Folks think of Nick as publishing Gawker (true), but what I think Nick was really into — and may be working on now — is a better commenting system (hearsay, unsure). Regardless, to comment within Gawker was an earned privilege, based on merit (or snark, or both).Why the constipation in this decision? Make those with permission to comment earn the right! As with immigration, a ‘clear path to citizenship’ is overdue. For those who mean it and want it and who meet the non-discriminatory criteria to join, let’s *welcome them.*And if Disqus lacks the means to do what I say above, I sense Fred has the means to employ a moderator, or perhaps a ‘Public Editor’ is selected (like The New York Times).But let’s get principles, recommendations and consensus done.
I think the commenting platform can solve this:Default view: Everyone with high reputation scoreFull view: EveryoneNew users see Full viewKind of similar to how people do Popular and Everything in social systems.Both views have ability to flag spam and upvote and downvote comments
>I am making more than $30,000 a moCareful there, your comment could be marked as spam – per Donna’s comment about comments containing $$$ (or ,000) 🙂 /jkActually the above thing points to an issue that AI has: inferring meaning of human speech or text when it is ambiguous.Couple of classic examples:1: Time flies like an arrow. Fruit flies like a banana.2: One superpower built a supercomputer to machine-translate the intercepted communications of another superpower. When demo day arrived, with the generals watching, the operator fed in this input to the big box:”The spirit is willing but the flesh is weak.”After a round of translation from source language to target language and back, the result was:”The vodka is good but the meat is rotten.”https://www.google.co.in/se…