@Dugnom if you mean the paytm ones just now, the spam answer was posted by a blacklisted user which brought the weight up into autoflag land, whereas the question only has a weight of 2
the number is blacklisted now but even then the weight would be insufficient for an autoflag
the (2) and the (174) in the metasmoke report line is the weight
individual users can set up individual autoflag limits but the recommendation is to keep them above 200-ish, though some users have specific rules for specific sites which go somewhat below that
if you click through to the metasmoke report page for that post, the red flag next to the subject indicates that it was autoflagged, and you can click on it to see details metasmoke.erwaysoftware.com/post/499715/flag_logs
ok, more questions :D why do I not show up? Was it already deleted when I flagged? (From my POV it looked like I was the first one flagging, based on 0 downvotes) metasmoke.erwaysoftware.com/post/499714/flag_logs
@Dugnom it should show up if you have a metasmoke account and flagged in a way which notifies metasmoke when you flag but I'm guessing either or both of those are not true
so clicking "flag" on the post does not notify metamsoke but using something like FIRE uses the metasmoke API, or clicking "Spam flag" near the bottom of the metasmoke post page obviously does
Additionally, from an early RFC document, "Concern: False positives are very dangerous. Mitigation: The system will be built to limit the number of automatically issued flags. Additionally, it will not flag until other criteria have been met" -- which I think probably morphed into the recommended flag weight threshold
@tripleee I am clicking the flag on the post. I read into FIRE a few times, but as I'm only a sporadic user I've never gotten used to it. I don't see the "Spam flag" button near the bottom.
@JeffSchaller But why such high accuracy. I recall reading somewhere that the average accuracy for spam flags on SO is 85% or so. Why does the autoflags have to be, at a minimum, 60 times more accurate than the average human?
@Starship I'm not a founder of the project, but I can certainly imagine that if had built a system that was casting red flags on behalf of other users, I would recommend high standards/accuracy as well. Otherwise it's going to give the project a bad look.
@JeffSchaller False positives really aren't that dangerous though. No harm is done unless the post is actually deleted (which wotn happen without an actual human involved). And frankly posts can also be undeleted as well so...
@tripleee Well stuff that are autoflagged can be deleted faster. So being able to autoflags a larger portion of the spam would allow spam to be deleted faster. I think getting spam deleted faster is a pretty good reason to change something
if anything we might be even more brutal in enforcing higher thresholds just because it's an unhappy outcome for everyone involved when we get autoflag FPs
@JeffSchaller I know you didn't decide it. But by agreeing to autoflags you do agree there will be FPs. I've autoflagged 1 FP before and I don't see any need for apologies...I've also manually flagged more than 1 FP
@JeffSchaller my point is I don't see why you need an apology for "We had something click a button but because it was wrong it actually had no effect on you"
@JeffSchaller they only come with penalties if the post is deleted, not if they are declined though.
@JeffSchaller would you not say that more benefit is done through 100 posts correctly autoflagged and deleted much quicker than 1 post incorrectly autoflagged and not deleted?
@tripleee That pattern looks like it's already caught by Potentially bad keyword in answer and Potentially bad keyword in body Append -force to the command word(s) if you really want to add the pattern you provided.
@tripleee no, I don't see that, and have never seen that (thank you for the image). This might have to do with my (non-existing) privileges here? I just linked my stackexchange account and registerend nowhere else, and never tried to get more privileges
@Starship The autoflagging threshold is not something I'm terribly worried about. I'm personally boggled that SE hasn't integrated more of this into its system, to prevent spam from being posted in the first place. Catching it afterwards is already "too late"
@tripleee the repeated phrases at the end (after machine translation, anyway) tipped the scales for me
@JeffSchaller Well the issue is that preventing something from being posted wrongly is much more problematic than wrongly flagged it and it not getting deleted
huh? I only see the text in what Google Translate identifies as Batak Toba and translates into "My phone is very slow and difficult to load Full memory, high intelligence, low RAM, background programs and programs running in the background, start automatic programs, increase information processing speed in mobile phones, best background management apps and programs."
@Starship it's SE's system; if they want to block something that looks like spam, it's up to them. No free speech rules here. But that's neither here nor there. We have a post-posting system in CHQ, and if you want lower autoflagging thresholds, I guess you can set that up.
I don't have a strong opinion on the finer points of the thresholds & accuracy levels, except that I'm in general support of high standards. Maybe a GitHub Issue would be a better place to make the case for/against a specific change?
@CodyGray Correct. Just “pixcap”, no URL (at least half of the posts; the other half contained “pixcap.com” or even an actual link). Also nothing else that I would expect to help identification.
> Would not be caught as a post, title or username.
> Potentially bad ip for hostname in body, potentially bad ip for hostname in title ---------- Potentially bad ip for hostname in body - theonepiecemangaonline.com suspicious IP address 104.21.62.171 Potentially bad ip for hostname in title - theonepiecemangaonline.com suspicious IP address 104.21.62.171
@Fastnlight That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer Append -force to the command word(s) if you really want to add the pattern you provided.
