YES or NO question: Will Narrative ever penalize downvoting circles?

Service: Narrative

Everyone knows that at least one large downvoting circle exists that includes at least one high-reputation member, if not more than one. Those of us who are often the targets of mass downvotes know it. Other people who comment on our posts are beginning to feel it, too, because COMMENTS on our posts, or comments we make on other people's posts, are getting downvoted, and it cannot be easily argued that these are low quality. Heck, even people who aren't often targets of the downvoters have noticed it from browsing posts that interest them, as it has become a topic of discussion on the Narrative Discord channel started by Blogger Krunal, which currently boasts more than 100 users, according to the last announcement I posted with the @everyone tag.

The downvoters are discouraging people from posting and creating distrust in the platform. Surely, you can see why this is a bad thing. But every time someone brings up being downvoted, the stock reply seems to be, "people can vote however they want." So will this form of abuse ever be dealt with, or is the community expected to tolerate being overrun by bullies who only upvote their friends and downvote anyone they determine to be enemies, regardless of actual post quality?

Will Narrative ever do something to stop downvoting circles from sabotaging other users? YES or NO?

The lack of an official response in a timely manner WILL be interpreted by the community as a NO. That is why I made this question public instead of private, because it affects all of us.

Original Post

Activity Stream

I agree there is a problem, but don’t think banning multiple accounts is the answer. IMO the best approach is to increase the ‘spread’ between high and low quality members and the influence they have on each other. Christina is a pillar in this community and she should therefore not be heavily penalized by even a large number of sketchy accounts. Likewise, she should be able to impact ‘lesser’ accounts greatly.

Gord posted:

I agree there is a problem, but don’t think banning multiple accounts is the answer. IMO the best approach is to increase the ‘spread’ between high and low quality members and the influence they have on each other. Christina is a pillar in this community and she should therefore not be heavily penalized by even a large number of sketchy accounts. Likewise, she should be able to impact ‘lesser’ accounts greatly.

@Gord - I agree, and I think the Team is trying this.  The influence of rep on voting weights is not linear at all, but seems to be quite exponential.

While this could be pushed further and further, we also could run into the problem of high rep individuals becoming too powerful.

Not only is that bad for our democratic aspirations, but It is quite apparent that at least one high rep individual is engaged in abusive downvoting.  Further empowering abusive high rep votes can invalidate all our efforts, since one high rep vote is worth so many low rep ones.  

I think the only real solutions involve the Narrative Company announcing that Narrators will soon be able to lodge complaints for abusive downvoting, and such complaints will allow the Tribunal to view the votes for a post, and gain access to relevant voting data of accounts involved in the downvoting.

(EDIT: I've posted this as a Community Support site suggestion, here )

There should be a statute of limitations on this, of something like 6 months - in any event, long enough so that the functionality arrives in time for people to lodge complaints about violations that occurred this month.  The knowledge that abuse carried out now will have consequences in the near future will have an immediate and beneficial effect.

To reduce frivolous complaints and to fund the work this creates, there should be a NRVE fee involved.  If the Tribunal finds there has been abusive downvoting, the abusers could then be made to cover the fee, and to pay penalties on top of the fee, in the form of docked earnings.  Significant reputation penalties should be possible as well.

The complaint process should involve an opportunity for the downvoters to explain their downvotes - these explanations will aid in a determination.

Within the course of a complaint, it should be part of the scope to determine whether there might be voting circle downvoting involved.  Once abuse is found to have been committed by several accounts on one post, if it is found that these same accounts have acted in concert, abusively, elsewhere, a new level of more severe penalties should be considered.

These measures will be effective because abusive downvoting only works if the voters have enough rep to do damage.  The mere threat of removing the rep that gives them their power, will stop most of the activity, and will effectively neutralise the power of those who persevere.  Add to this the penalties that limit rewards for infractions, and you have a serious incentive to vote honestly.

 

I think that ANY voting circle should be prevented and detected. Not easy and all possible, of course, but when multi-accounts and abusive patterns are clearly there, it should be done, someway.

Reputation is a key factor for balancing things, here, but it's also the first factor to clearly need tuning, avoiding actors who just bomb the platform - maybe with also the use of multi-accounts or voting circles - to gain high rep. The Quality analysis factor has now clearly NOTHING to do with quality metrics, and Conduct has also a weird and useless behavior.

But also some basic pattern recognition should be there, and many other tools.

All of this may be already on the roadmap, but some of it should have already been in place. It's not, and we need at least some intervention and a clear position that clear abusive voting is not tolerated and WILL be fought.

Surely there are algorithms that can be deployed to learn and spot abusive voting - just as there should be algorithms in place to spot multiple accounts from a single user - none of these is fool-proof, but something is better than nothing. There seems to be retaliatory voting going on as well as "gangs". I have eased back from posting, because I am waiting to see how all this plays out. I like Narrative as a platform, and I am willing to hang in for the end of Beta to see what this evolves into, I would hope that the team doing the development can see how this affects the morale of the Users and the platform's success.

Hi all,

We are very aware of your concerns and have been tracking your feedback internally. We have also spent time analyzing the data, and at this point, there is insufficient evidence that this is a substantial or widespread problem affecting users and posts on the platform (both in terms of content quality and user reputation). We are constantly evaluating our reputation and quality analysis algorithms, and they are sure to be tweaked and updated in the future, particularly as part of the Reputation and Plagiarism Improvements from the Roadmap.

If there are specific posts or comments where you think downvote abuse is occurring, we'd be happy to investigate a few more closely to see if there are particular patterns that we haven't yet caught in our analysis. You can submit those here or in a separate private support ticket, if you prefer. We won't be able to provide any specifics of any underlying post ratings, but specific examples may be useful in our future algorithm tweaks.

Thanks for your support!

Brian

Brian Lenz posted:

Hi all,

We are very aware of your concerns and have been tracking your feedback internally. We have also spent time analyzing the data, and at this point, there is insufficient evidence that this is a substantial or widespread problem affecting users and posts on the platform (both in terms of content quality and user reputation). We are constantly evaluating our reputation and quality analysis algorithms, and they are sure to be tweaked and updated in the future, particularly as part of the Reputation and Plagiarism Improvements from the Roadmap.

If there are specific posts or comments where you think downvote abuse is occurring, we'd be happy to investigate a few more closely to see if there are particular patterns that we haven't yet caught in our analysis. You can submit those here or in a separate private support ticket, if you prefer. We won't be able to provide any specifics of any underlying post ratings, but specific examples may be useful in our future algorithm tweaks.

Thanks for your support!

Brian

Hello @Brian Lenz - good to hear your current thinking.

I will submit a private support ticket that will provide you with more data.

This surge in new Community concern is happening for a valid reason - the phenomenon has made a leap and is now much more acute.  I think it would be good to consider most of us have a track record of saying the downvoting is not a serious problem, despite having been the targets of abusive downvoting before on several occasions.  In the past these attacks have resulted in good quality posts perhaps having 95% quality instead of 100%, and the attacks have been more sporadic and have not spread as much the comments. 

The pattern has shifted quite radically to being obviously caused by probably multiple high rep voters (the quality percentage hit is MUCH bigger), and the attacks are sometimes attacking almost all the commenters on a target post, as well.

This not only dissuades the poster from creating content, but it has a serious chilling effect other Narrators wanting to interact with their content.  This is measurable.  After all, is it worth receiving potent downvotes merely because you have something innocuous to add to an innocuous conversation?  Probably not.

You may not have seen these patterns yet, but if you have, I would be very concerned that you don't see this as a problem.

 

This is definitely one of those moments where as developers, stepping back and seriously asking whether algorithms are going to solve the problem, is in order.

High rep bad actors cannot be solved by changing voting weights.  Sad, but true, and ignoring this will not make it go away.  It will only give more time for the trend to accentuate, as more and more people become high rep.  The damage done by a high rep bad actor is disproportionate because they can abusively downvote a very high number of individuals, with no repercussions to themselves.  That's what you have to crack, and your message here, despite being appreciated, sends exactly the wrong message.  It sends the message that this network-negative behaviour will not result in any penalties for the bad actors.

Please consider the following suggestion:

https://community.narrative.or...-future-measures-now

The only solution to high rep bad actors is to send the message you will be taking steps to effectively demote their high reputations when these actions are detected.  No deterrent, and the problem will continue to grow.

As a secondary consideration, there will be no sense of Justice as long as the economic impact these actions have result in economic penalties for the bad actors.  If you don't do this, you are not only failing the Narrators you should be valuing the most - you are also failing the entire content economy.  After all, these few individuals are eating up your dev time, and hurting the economy you are trying to build.  

You cannot teach better behaviour when bad behaviour has no consequences.

 

Malkazoid posted:

This not only dissuades the poster from creating content, but it has a serious chilling effect other Narrators wanting to interact with their content.  This is measurable.  After all, is it worth receiving potent downvotes merely because you have something innocuous to add to an innocuous conversation?  Probably not.

...

 

As a secondary consideration, there will be no sense of Justice as long as the economic impact these actions have result in economic penalties for the bad actors.  If you don't do this, you are not only failing the Narrators you should be valuing the most - you are also failing the entire content economy.  After all, these few individuals are eating up your dev time, and hurting the economy you are trying to build.  

You cannot teach better behaviour when bad behaviour has no consequences.

Yes. This is already happening. I have an entire body of work I want to eventually contribute to the platform, editing and updating it to be as relevant as it was when I wrote it up to 15 years ago. I'm not contributing anything newly written exclusive for Narrative (unless it's ABOUT Narrative) because I don't want my hard work to be for nothing as it gets downvoted unfairly. 

Other HIGH QUALITY NARRATORS are also holding back on posting content they care about here. Low quality content is being disproportionately rewarded because there are more people willing to upvote each other (regardless of quality) to get THEIR cut of the content economy rewards than there are people who care about quality.

I am very disappointed in the response from @Brian Lenz. You may not be able to see it, but there's a list of members you could look at post with quality below 85% (because some people were kind enough to see the even lower quality ratings the posts had and upvoted when they couldn't understand why anyone would find it low quality) to find the malicious downvotes on the posts and occasionally the comments. It's not a "widespread" problem because it's mainly high rep users who have been targeted, and the percentage of the user base with high rep is not very large. It will get larger though, and the problem will grow like a cancer if left unchecked. 

I'll be working with other users who want to put together a private ticket to show what's been happening. 

Hi all,

I realize that my response may not have been what you are looking for, but facts are facts. There is absolutely zero evidence of any targeted downvote abuse. Users may downvote whatever they want. I think in a platform like Narrative we have to accept that we're not always going to please everyone. Posts are going to get downvotes. Posts in the 80-85% quality range are very good posts! Let's not lose sight of that. The platform considers anything 80% and higher to be high quality, so I see very little harm occurring here.

Further, there is no "economic harm" caused by these downvotes, so I want to make sure that is understood. The only time a post's rewards can be impacted by downvotes is if the downvotes take it down below 25% quality. At that point, it is low quality and is no longer eligible for rewards. That doesn't apply to any of the posts we are talking about here, all of which have a lot of upvotes and a minority of downvotes.

I know it's difficult that you don't have verifiable data to back up my statements about the lack of downvoting patterns, so you're just going to have to take our word on it. This is one side effect of protecting the rater's privacy: you can't see exactly who is upvoting or downvoting individual posts.

Lastly, I'll just mention that the reputation system already has checks and balances in place to detect and deal with bad actors, including those who try to use the downvote as a tool to hurt other users.

We'll continue to monitor the situation, but this is the latest info I can provide after a few private reports of specific posts for us to investigate. No need to send in further reports at this point; we are confident the statements above reflect reality.

Regards,

Brian

@Former Member, no, that's not what I said.

My statement is factual: unless a post is below 25% quality, then downvotes will not affect rewards. Rewards are based on other activity (not downvotes) that comprise a content score (and ranking). In your example, the 26% rated post and a 100% rated post will be scored independently based on the content scoring criteria. You can read more about it here:

https://spec.narrative.org/doc...content-creators-60-

Hi Brian,

I'd like to offer a specific example if I may?

https://www.narrative.org/post...ust-100k-challenge-2

The vast majority of my posts get between 6 and 8 votes at 100%. Occasionally I'll get a few more votes, but it's usually for something I've put a bit of effort into.

The above post is really just an update, nothing special, but it somehow got 22 votes and is at 78% which is super unusual for me. Quite a few of the posts around the time that this post was posted also got a lot more downvotes than usual.

I agree with you that the overall percentage of downvoting circles would be really low... but I think if you compare this post with my others, you might see a blip in the data.

I'm not stressed about these votes at all, I just offer it up as an example.

Thanks!

 

 

Thanks, @AussieNinja! That's one of the posts that we've investigated, so the findings I reported apply to it 

@Former Member, the content scoring and ranking formula is intentionally unpublished. The spec states that activity affecting content scoring/ranking is "comments, views, likes/dislikes, and more."

Are you saying that the quality rating number is not part of the equation affecting the AMOUNT of earnings unless it's below 25%?

Downvotes do not take any points away from the post at all unless below 25%?

Correct! If two posts have identical comments, views, likes/dislikes, they will earn the same amount. Keep in mind that activity is tracked on a monthly basis, as rewards are on or about the 6th of each month.

@Brian Lenz - the info above is a revelation that will certainly reassure many people.  That was not at all clear to me, and I think to many others.  I do understand you are not publishing too much about the inner workings of the system, but I guess this was the right time to share that aspect!

Thanks!

Two concerns remain for me - I don't know how many people use the filter to remove low quality content from their feed, but you seemed to be saying in another conversation, that a post at 65% would not show up in someone's feed if they are using that filter?

If even just one fifth of users use that filter, abusive downvotes causing a post not to appear on that many feeds would be something to be concerned about IMO.  As you know, humans are exquisitely attuned to fairness, so considerable frustration will be bred whenever a Narrator works hard to post, then knows her post isn't appearing in many people's feeds because someone is being a jerk.

Which brings me to the last concern... from these responses, it sounds like the Team is actually ok with letting abusive downvotes happen - that it is just something to expect.  On other sites where downvotes have no affect on visibility at all, that's less of a problem.  But on Narrative, where everyone is primed to see the space as a content economy, the perception is always going to be that any malicious sabotage of somebody's hard work should not be something the platform tolerates.

The Community has proposed a large number of methods to improve this situation, so it is frustrating to learn months and months later, that actually, it isn't even viewed as problem by the Company.  Pretty big breakdown in communication!

Some of the methods:

- request a short typed reason for a downvote, that moderators can flag as disingenuous if they see it

- at least state in the AUP that abusing the downvote categories is a violation of acceptable use...  really not sure why you can't at least frame the behaviour as being bad?  You've mentioned to me elsewhere that the system is already affecting people's reputation if they downvote abusively, so why not make this knowledge public so it serves as the deterrent it should be?  Seems like the whole point...  Deterrents don't work if people don't know they are active, and the behaviour they are meant to act against has not even been defined as a transgression?

@Malkazoid

I don't know how many people use the filter to remove low quality content from their feed, but you seemed to be saying in another conversation, that a post at 65% would not show up in someone's feed if they are using that filter?

Low quality content is anything under 25%. A 65% quality post is medium quality and will be unaffected by the default post filter, which is Exclude Low Quality. There is an alternative option for High Quality Only, and that will only include posts with an 85% quality rating or higher.

from these responses, it sounds like the Team is actually ok with letting abusive downvotes happen - that it is just something to expect.

My perspective is: downvotes will happen. If we wanted there to be no downvotes, we wouldn't have the option. People can downvote for whatever reason they want.

Now, the reputation engine will also analyze behaviors and take appropriate action. Bad actors very likely may see reputation hits that result from abusive downvotes.

- at least state in the AUP that abusing the downvote categories is a violation of acceptable use... really not sure why you can't at least frame the behaviour as being bad? You've mentioned to me elsewhere that the system is already affecting people's reputation if they downvote abusively, so why not make this knowledge public so it serves as the deterrent it should be? Seems like the whole point... Deterrents don't work if people don't know they are active, and the behaviour they are meant to act against has not even been defined as a transgression?

The difference is that downvoting multiple posts is not an Acceptable Use Policy violation. Somebody might upvote every one of your posts, and they also might downvote every one of your posts. Neither should be an AUP violation. Of course, the reputation engine will analyze your behavior on the platform and adjust your reputation accordingly. The AUP doesn't need to list every negative action that might affect your reputation since reputation doesn't have anything to do with acceptable use. Once the reputation algorithm is published, it will be very clear what you can do to improve your reputation and what you can do to hurt your reputation.

Hope this helps...

@Brian Lenz- certainly, when the formula is eventually published, a lot of the current friction will dissipate.

"People can downvote for whatever reason they want"...

That's the problem - that's a terrible message to send!

You're a representative of the company, saying people can downvote maliciously if they want to, and that the rep algorithms might tweak their rep accordingly.

Well, in this tech era, people push the limits.  They'll think "I bet those algorithms won't catch me if I downvote maliciously infrequently, or using a pattern that is difficult to detect".  So they'll try it.  They'll do a couple of malicious downvotes, and see if their rep budges.  If they see no change, they'll do it again.  The lax message you send invites boundaries to be tested.

It even invites concerted attacks on the system.  I will describe one to you in a private ticket, that could flood the platform with 200 high rep downvotes per day, be up and running in a month, and cost perhaps $3000 per month to operate.  Cheap way to set back the competition, no?

-----

There seems to be an idealism surrounding protecting the anonymity of voters, and I agree we should do that when a voter is voting honestly.  When a voter votes fraudulently, on Narrative, that is like vandalism.  Protecting the anonymity of a vandal/fraudster is not the same thing as protecting the anonymity of a voter.

But you don't even need to dive into who voted.

Moderators will have very little work to do: I'm the default moderator for a bunch of niches right now, and it takes me just a few minutes per day.  When moderators get elected, there will be at least two moderators per niche!  They're going to be begging for more work.

Why not make downvoters type a sentence explaining their downvote, and let moderators flag explanations that seem disingenuous.  Their success rate will blow your algorithms out of the water - no offence.

The flagged downvotes could go to the Tribunal, who could make a decision, with or without knowing the identity of the downvoter (in other words this can still work even if you maintain 100% anonymity of voters, even the vandals).  Abusive downvotes could result in a rep hit, and successive ones, could result in bigger rep hits, and even financial docking.  Always with an email sent to the abuser letting them know they got caught.

The moral of the story?  People will know - if I downvote, I'd better be able to say why, and it better make sense and line up with the downvote reason I chose from the list.

Algorithms are great, but the dev team is small, with a massive feature list to develop.  Why not use the manpower you have on hand - the Moderators who on paper sound like they will be twiddling their thumbs 95% of the time...

And even if you hate this idea, at least recognise that the Narrative Company should have a moral stance here.  Saying you can act any way you like with your downvotes and maybe you'll suffer a rep hit, is like saying you can steal and lie and cheat and defraud, and maybe you'll get caught.  It is the truth, but it is a truth devoid of any moral backbone.  Societies work best when there is a shared and taught ethos of what behaviours are positive, and what behaviours are negative.   When people in positions of leadership say: "do as you will (inferred: you might well get away with it)"... that has negative consequences.

 

Question Answered

This action was taken by David Dreezer.
To follow up on this question, please click here.
×
×
×
×
×