34

Proposal:

If a <2k user's past N suggested edits are all approved ("good edit streak"), make their following suggested edits only require one approval.

For rejections, I can't think of a reason to change the number of rejections required to reject the suggested edit (decreasing or increasing), but this is worth discussion here. Perhaps if there is one rejection but not yet two, the number of required approvals could fallback to two (without immediately breaking the streak).

If we want to press harder on encouraging suggestors to try to learn from improvements others make on top of their edits, we could consider making looks-ok-&-improve reviews also reset the streak counter. There is also this dedicated feature-request on the subject: "Help users discover and learn from the results of their edit suggestions".

I'm not sure what a good value of N is. In my head, the first number I thought of was 10, but there was nothing scientific about that.

Related FAQ post: How do suggested edits work?

Rationale

(in forward order of causation, and reverse order of size of groups that benefit)

  • Reduce pressure on the queue for users that have shown to be trustworthy
    But don't trust them too much either (streak mechanism)

  • Let trustworthy editors edit at a faster rate
    and incentivize keen suggestors to learn to improve when their edits are rejected
    (well, if they edit more, then it somewhat "nullifies" the previous mentioned benefit, but...)

  • Community gets good edits faster

Motivation

Experience reports of staff from the 2022 community-a-thon report that the edit suggestion process is slow on the edit-suggestor side. (See Aaron Bertrand's report, and John M. Wright's report). I also felt that way before I reached 2k rep.

Other related Proposals

  • To solve the opposite problem of too many poor edit suggestions, we added the 5-pending-suggestions-at-a-time limit. This proposal shouldn't undermine that: The streak is reset when poor edits get rejected. There could be benefit for loosening that limit of 5 for suggestors with a good streak (but I smell feature creep in this post).

  • Revert the 2017 the top bar change: show 2k users the size of the suggested edits review queue at all times. Rationale: showing it everywhere encourages giving the queue attention.

  • Get rid of the 2k rep threshold for the edit privilege and allow/make reviewers vote on whether user is trusted enough to give full edit privileges. I heard of this from @KevinB. No link handy (sorry).

  • There have been two previous feature requests for a review bypass mechanism based on percentage of approved suggested edits. I am not proposing a total review bypass, and I'm proposing a mechanism that can "change its mind" about trust quickly, which should be less susceptible to "gain-trust-then-be-evil" abuse.

Spotlighted Discussion

  • Sonic says:

    A bit of an alternate perspective: one major reason why I continued to edit anonymously was because I wanted my edits to be reviewed for accuracy. If this were implemented back then, I wouldn't like it.

    My response: We could let users opt-out of the "streak benefit", and give a short explanation of why someone might want to ("I'm a bit unsure. Please give me two peer-reviews"). I'd suggest such an opt-out to be changeable per-suggested-edit (instead of a global setting accessible only in the user profile). That way simple edits like formatting, spelling, and grammar can still be easily opt-in.

2
  • 5
    Also, with regard to timing and full queues: prior to changes to the top bar that took place in 2017, edit reviews used to be way faster and I'd never run into a full queue on Stack Overflow or here or any other site. During my anonymous editing spree, there was a clear divide in the time it took for my edits to be reviewed; it changed from the order of minutes prior to the change to the order of hours after it. I think the best way to resolve those issues is to revert those changes. Nov 16, 2022 at 7:22
  • 8
    Before the top bar changes, on Stack Overflow, the edit queue size was only 200, 40% its size today, and edits required three reviews there instead of two. As a result of the capacity being reached immediately after the top bar changes, the number of required reviews was reduced to two and the queue size was bumped up to 500, which worked to stem the problem for a time, but failed to address its root cause, as it's come to a head once again. Nov 16, 2022 at 7:25

1 Answer 1

4

This is a good idea that gets to the heart of what most discussions about revamping privileges land on: assigning privileges based directly on site use/experience, rather than indirectly via reputation.

I think the streak value n should be significantly higher than 10. It should be some number greater than what a single reviewer could approve in a single day or two, to limit potential fraud (e.g. a second account with reviewer privileges). To that end, I think 50 would be a good n value, with the streak count getting reset back to 0 after a single rejection (not a single reject review, but a singly fully rejected edit). A post owner roll-back should also probably qualify as a rejection.

On a related note, I think after 500 suggested edits, the number of approvals needed would also be worth reducing to 1, regardless of any streaks. At that point you're likely halfway to the ability to edit unilaterally anyway (given that edits grant +2 reputation, meaning 1000 reputation gained from suggested edits, and the threshold for unilateral edits is 2000 reputation), and it would help organically "speed up" your suggestions as you earn more experience with suggestions.

5
  • 3
    50 seems like a lot to me. maybe just because with the current rate of suggestion handling on SO, you can hardly get in one edit suggestion per day. I don't quite get why tie it to how much a reviewer can review in one day, as opposed to how many edits the suggestor can submit per day (which IIRC is a much lower value- 5)? Feb 22 at 21:22
  • as for your 500 suggested edits thing, boy do I have the perfect link for you Feb 22 at 21:23
  • @starball it's tied to how much a reviewer can review because it's not only a check against a lucky streak but also as a check against fraud/account rings. It's also much easier to find 10 poor questions and superficially improve them than 50. The requirement is to have multiple people reviewing your edits until you've reached 2K rep. I think 50 edits in a row (so, if you do nothing but suggest edits, and get them all approved, that'd be at 101 reputation) is actually low for a perk that reduces by half the amount of oversight you need before unilaterally making changes to the site.
    – TylerH
    Feb 22 at 21:32
  • 1
    I feel like there are different (better?) solutions to that problem? apply similar logic from voting fraud detection to edit reviewing. If one edit suggestor's reviews are being primarily reviewed by a specific, small percentage of the current active pool of reviewers, then that should auto-flag for mod attention or something. Feb 22 at 21:38
  • @starball That sort of method requires active work. Ideally system guards against fraud are passive and can run unattended. Both would be ideal, but at the very least, the first one to be adopted should be the passive one.
    – TylerH
    Feb 22 at 21:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .