Automatically detect edit abuse and exclude from badges/flag for mod review

Question

Flem has 5 pages of 1 character edits (including an edit to one of my posts), apparently in a massive attempt to earn the Copy Editor badge.

Should Copy Editor, etc., ignore 1 character whitespace-only edits?

---

If we still want to count minor edits (which may be useful), could we at least detect edit abuse via some secret algorithm? ~20+ one character whitespace-only edits in the course of an hour should at least raise a flag of some sort.

---

If whitespace character are too hard to count, can we at least assume that 20 non-tag edits in 20 minutes (less than a minute per edit) is a good signal of abuse? The specific algorithm should be secret, obviously, but I think these sorts of edit abuses shouldn't be too difficult to detect.

Answer 1

I made a one character change earlier today that was critical. The post had a handful of foreign language words in it and was specifically a discussion of the language structures used. A one character typo in one of these had the unfortunate effect of saying something embarrassing. I was about to leave and didn't have time to review the post to find other fixes to make (although I'm sure there were plenty of possibilities) but I could use the 60 seconds I had to fix the issue that jumped out at me.

Other examples I remember seeing are fixing a closure in code incompletely pasted into somebodies answer (note if this was a question that would be pointed out in an answer but for an answer that's a valid reason to edit)

One cannot programatically determine whether a 1 character change is an improvement or not. Blocking them would cause unwanted hassle for trusted users. Which brings us to another point.

One character edits are already blocked for low reputation users. I believe six is the minimum. This helps new users understand that editing has consequences (such as bumping posts) and should not be used for trivial purposes. Once they learn that, then they are trusted with both smaller edits and bigger ones without review. The problem here is that you've got a trusted user who should know better (they've earned those latter privileges) who is gaming the system.

If they are that deliberate at gaming the system, do you think a character limit change would stop them? If one didn't work why couldn't they just put     in somewhere and presto. They could still do it en-masse and collect their badge. If the limit was 100, it still wouldn't stop them.

The solution is to fix the problem: they need to be warned by a mod message or somehow blocked from the privileged until they are willing to stop abusing it.

If you want to do detection, you need to detect and flag a pattern of abuse, not any one infraction of a filter that might not actually be an infraction. While one or two or a handful of single character or whitespace only edits might not be an issue, a series of them is a clear pattern of abuse. A series or high percentage of such edits might be something automatically detectable.

Answer 2

Can we at least assume that 20 non-tag edits in 20 minutes (less than a minute per edit) is a good signal of abuse? The specific algorithm should be secret, obviously, but I think these sorts of edit abuses shouldn't be too difficult to detect.

Instead of an algorithm which can always be outsmarted by someone who is determined enough, I propose something else. I propose doing something that already works well today - letting the community judge whether edits are valid.

My proposal is that edits that are made by high rep users be reviewed (similarly to how low rep user submissions are reviewed). There is one difference: frequency of review. Edit submissions by high rep users would only be reviewed randomly, and a very low percentage of the time (e.g. 3% - 5%, exact percentages to be determined).

By doing random auditing like this, it allows the community to catch abuse. If an editor fails too many audits, the mods (or the system) can take a look at the situation and act accordingly.