Automated Spam filters - Boon or Bane


Ryan sent me a message yesterday clarifying that some recent moderation I had complained about was an aggressive automatic filter at work, and not a moderator action.

As I have written to him, I am putting my response to this and to the spam question in general in this thread so others with value to add to the finding of a simpler and better way of handling these irritants can contribute here.

IMO:

Automated spam filters don't seem to work well in two ways.

One, we still see spam here that any decently designed filter should have caught; spam that is blatant, from first time posters. Automation often doesn't seem to catch such spam.

On the other hand, when for example a spam filter blocks a post that I have written as part of an ongoing conversation or argument in a thread for no valid reason, all it achieves is messing up of the thread by disrupting the flow of the argument. And I say no valid reason because after some hours or days, the dumped post is reinstated by moderators with not one word changed therein; but the damage to the conversation flow is not eliminated by correcting this error of automation.

Why not then eliminate both sets of problems by throwing away the automation, and subjecting the first few posts of every new member to manual moderation? It could be just the first five posts - or perhaps even just the first one. And for such posters whose first post does not pass moderation, the next one as well, until credibility is proved.

That leaves the matter of objectionable posts/language by posters, regardless of their post count. In one forum I know, this is largely dealt with based on member reporting of such posts to moderators. Every post has a little "report" button that can be used to do this conveniently. Although this sounds like citizens ratting out their fellows in a typical totalitarian state, it can work well enough because moderators will still be the ones deciding if the reporting deserves a response.

I doubt that changing to such a system will substantially change moderator work loads, but will certainly improve the state of what is there in the various threads today.

14 replies

I would suggest that the hours of moderation of this board would not benefit any "first time poster" while they waited for the next moderator to show up and authorize the post. The effect that I imagine is that we'd have a much smaller community, not getting assistance in a timely fashion from other users.

Which doesn't counter your concern, it's just an alternate viewpoint. Ideally, the automated filter would be smarter than what is in use currently, and catching posts inappropriately. But every AI system I've ever been involved with has had flaws.
Only moderators can speak to how much time the existing resource would take to clear every first post; I have no idea. But if Sonos wants this to be the fast response resource, they need to ensure that this isn't more than a few hours, 24/7. Alternately, something similar needs to be done to ensure that the official response centres are better resourced to meet the ASAP assistance needed requirement.

On the other hand, I also don't know how much time will be saved on scrapping the AI by moderators then not having to go through all the flagged by AI posts to see if they deserve to be passed instead. And how much time would be saved by not having clean up the spam that has slipped past the AI.

My suggestion to scrap the AI is because for over a year now it has been seen that AI isn't doing the job well enough.
As far as I can trace them, there were only 3 first time posters in the last 24 hours, asking for a problem fix. Again, moderators would know what is a typical number, but it does not seem like a flood.
Userlevel 7
Badge +21
Something else I noted with this auto-spam filtering.

I posted in a thread yesterday. I subsequently received an e-mail notifying me of a reply to the thread. But when I went into the thread there was no reply. I realised that the response had gone into moderation but the notification e-mail had still been sent.

I'm frankly not surprised that the Insided software is so poorly written that the notification mails are sent even before or if a new post is actually allowed. Totally cock-eyed
Automated moderation is the wave of the future unfortunately, and it still needs a ton of work. A few boards I frequent use it and one of them we were able to prove to the moderators the problems that Kumar laid out above. Their comment is always 'It is using AI. It is still learning.' It is a cost cutting measure so there aren't mods working overtime.

There is no reason why a first time poster putting a phone number in the subject of their first threads here should make it past that automated moderator. That code is just poorly written, and the fact that the problem has persisted for months shows that Insided either doesn't care or doesn't have a clue how to program.

I looked into upgrading the forum software for my company. I passed because of the AI infiltration. I don't need the extra headaches of needing to explain to our users about all of the above ad nauseum. The current software lies firmly planted in 2009 and still functions perfectly with nearly no maintenance. And we take turns moderating.
It is a cost cutting measure so there aren't mods working overtime.

Aren't they having to do this to take care of the frequent false positives and false negatives anyway? So why not dump the AI and do things right first time via - guessing here of course - much the same effort?
Because the promises have been made and the feature oversold.
Sure; but after over a year of being found out, isn't it time to call a spade a spade and do something different than the same old same old...if the interest of a good board is being kept front and centre?
Many IT folks that I’ve dealt with are loathe to go back on a system that they’ve invested sugnificant time and money in.

Not all of them, mind you. And frankly, it’s often as a result of pressure from management, who is substantially less happy to admit failure.

But sure, I’d say it is time to “fish or cut bait”, as they say.
Userlevel 7
Badge +21
Not a fix but when I see a post caught in the spam filter I post a note that the post is spam trapped and if possible add an edited version (no links or such) to provide as much info as possible until the original is released.

It would be nice if the spam filters could prevent posting of links or phone numbers for new users, say for 24 hours. Using a post counter just gets you a number of worthless posts to bump them past the spam filter limit.
Even without links or phone numbers, editing causes a comment to vanish into thin air. Seriously, it's tedious.
Or even a post in reply that contains nothing except a pasted link to a Sonos sticky. More unnecessary work for Sonos Staff.
Every other post of mine in the last few days is going for moderation and some are never seen again. The message I get is that Sonos is telling me to shut up, so I will.
Every other post of mine in the last few days is going for moderation and some are never seen again. The message I get is that Sonos is telling me to shut up, so I will.
And this one went for automatic moderation as well. Ridiculous.

Reply