Prevent protecting questions under certain circumstances

Question

I've found a trend that I'd like to discuss. There are users that have reached 15K+ reputation that are participating in a question and then protecting it so that low reputation users can't answer. This wouldn't be a problem if the question was attracting poor, low quality or spam answers from these low reputation users. However, in these cases, that's not what's happening.

The Protect Questions privilege has the following guideline:

Questions should be protected when they are garnering lots of views and newbies are adding "me too!", "thanks!" and possibly even spam non-answers.

Some recent examples to consider:

• Can vstest.console filter tests to run by providing traits? This Q&A are by the same user. There are no deleted answers on the question. (14 views)
• Can vstest.console filter tests by all matched traits? This Q&A are by the same user. There is a second answer that is not the highest of quality, but the user has 200+ reputation and wouldn't be affected by the protection anyway. (23 views)
• What is the difference between getPageLabels and getPageLabelFormats? This question was protected by the author of the only answer. There are no competing answers. (36 views)
• Move LineSeparator to the bottom of the page This question was protected by the author of the only answer. There are no competing answers. (37 views)
• How to create a list without indentation? This question was protected by the author of the only answer. There are no competing answers. (50 views)
• Arrays.sort (with Comparator) - same or different thread? This question was protected by the author of the only answer. There are no competing answers. (64 views)
• How to correctly convert MIDI ticks to milliseconds? This question was protected by the author of the question (not the answer). There is only one answer. (140 views)

Proposal

I'd like to propose a couple small changes to how a user can protect a question:

The question should not be able to be protected if all of the following are true

• Has less a low number of views. We use 100 views to determine when close votes start expiring. I think 100 would work here too.
• None of the posts have garnered low quality or spam flags
• None of the answers have a score less than 0
• The user protecting the question is a participant

If any of those criteria are no longer true, then the participants should be able to protect it normally. I worked with ChrisF to spot check the flags on the above posts. There are not any low quality or spam flags.

This means that the final question in my example would be able to be protected by both participants.

If you haven't answered the question, you would be able to protect a question as normal. The idea is that the author of the question or answers can't protect if if the other three criteria are true

Motivation

The idea is to prevent locking out low reputation users from these questions. They are not gathering the types of answers mentioned in the guidelines that require a post to be protected. Instead, it looks like a conflict of interest. The user protecting the post is one of the users that can benefit the most from preventing competing answers.

Answer 1

We recently implemented a change to protected questions based on the suggestion by Robert with a few modifications. 15k users will still be able to protect but in order to do so the question must have at least one answer by a new user aka a user with < 10 rep on the site (which is the required rep limit to answer protected questions).

If the question does not have an answer meeting this requirement, then the protect option will not be available. This should minimize some of the protecting of questions that do not need it.

Note: Moderators will be able to protect a question at any time regardless of answers by a new user.

Answer 2

This one has bothered me for awhile. I see a lot of questions being protected that show none of the problems that protect was designed to avoid (i.e. misplaced answers by low-rep users).

Preemptively protecting a question "just in case…" is essentially barring a certain class of users from participation where it is completely unnecessary. New users should NOT be barred from participation… unless there is cause.

So this feature-request can be made much much simpler with one straightforward premise:

Do not allow protect if #answers from low-rep users < 2

I understand your concerns about avoiding a (potential) conflict of interest, but look at your examples; they would all have been be avoided if we implemented this one simple rule.

I'm sure you can find the rare exception of abuse of this feature by a high-rep user, but those instances are easily caught if folks simply understand what the feature is for explicitly. My (simplified) version makes that clear by enforcing the use case of when it should be applied. It does assume good faith by our high-rep participants — I would hate to bar high-rep user from fixing threads they participate in — but this also avoids the blatant misuse of this feature when it is applied outside the purpose for which it was designed.

This restriction will make that clear.

Answer 3

(I am basing the answer on Brad’s comment where he says the mod’s don’t have access to tools to investigate this)

We don’t need yet more automated complex rules to stop high rep users doing bad things, we need the mod to take action against them, after issuing warnings.

But before moderators can take action, they must be able to investigate. Therefore the moderators must be able to see all posts a given user has protected, ideally in a report that also lists if they have provided an answer to that post.

If a filter could be run to find all users that have protected more than N posts, that they have also answered, it could allow the moderators to search for problem users doing this. Also including number of deleted answers and comments in the reports may help.

Answer 4

Near as I can tell, the "Protect" feature was based somewhat loosely on a Wikipedia feature called "Semi-protect" that restricts edits from new users without a nominal history of positive contributions.

As usual, Wikipedia has a set of rough guidelines for its use:

• All or almost all of the vandalism is coming from unregistered users.
• Unregistered editors should be making very few quality contributions to the article compared to the amount of vandalism coming from unregistered editors. The negative effects of semi-protection on discouraging positive contributions should be more of a concern than the positive effect of decreasing vandalism.
• There are regularly many new vandals, therefore it would be a huge unending task to notify and warn all the vandals individually.
• [...] The higher the percentage of vandal edits, the greater the need for protection.
• Consider a lower threshold for protection for articles on living people as vandalism is potentially more damaging in these cases.

There are a few sites that occasionally host questions about living people, but generally this isn't a big problem on Stack Exchange; outside of meta questions, questions here are not potentially damaging to some 3rd-party's reputation. But the rest of these criteria are pretty solid - in fact, we've already baked the first two into the system for automatically protecting questions.

That pretty much just leaves those extreme edge-cases where a post is being vandalized by new users but the system can't reliably detect it. As usual, we've erred on the side of letting real humans identify these cases, but now it's been demonstrated that they're not applying any reasonable criteria...

...So let's educate them!

As bluefeet noted in the comments, we're considering removing the option to protect from questions that haven't received at least one answer from a new user. That's a bit of a no-brainer - by definition, you can't have new users vandalizing a question with their answers if none of them have posted any answers. Wikipedia's guidelines have much more complicated rules for evaluating this need, involving percentages of bad answers to good ones and historical data on the average amounts of vandalism - I elided these with ellipses in my quote, because I think something like that would be overkill here; I'd rather make it easy to protect when there's a chance it's needed, and just as easy to unprotect when no longer necessary.

With this change in place, our opportunity for education arises: rather than removing the "protect" option, cause it to fail with an illuminating message:

Now we're killing two birds with one stone: eliminating a chunk of completely unwarranted protections, and educating the folks with this privilege on its proper use for the remaining cases where it might be used.

Answer 5

This is all predicated on "we don't have the tools to investigate this."

Quite simply, we do. Most people are probably not choosing to look at them, or act on them, but those are different problems.

A quick walk over to the 10k tools brings us to: 10k tools > stats which provides a section:

And now you can click on those links and see why. This implies that people are actually using the tools that they are given. No additional code is needed. The "what was done recently" is already there along with all the needed things for human or mod oversight.

If there's a problem with 15ks protecting questions against low rep users, then that is something to flag for the mods to deal with.

No automated protections are needed. 10ks and mods just have to use the tools that they are already given.

If this is a 0.1% of protection being abused as described in the OP's question (that number is a guess), then what is needed is better auditioning of the tools and its use rather than a hard code prevention.

---

Lets look at what protection is intended to accomplish and see if this matches up with how the suggested restrictions would promote or hinder this.

Protection is intended to prevent new users to the site from posting answers in questions that are gathering poor quality answers. Many new users tend to follow the trend of other answers to the question. If there are poor quality answers, the new user is also more likely to post a poor quality answers.

Protection tries to prevent these poor quality answers by ensuring that the user has at least a minimum amount of familiarity with the site and its norms to be able to post a good answer. This is especially common on sites that are a bit more on the discussiony side and you get answers such as "I can't comment, but here's what I think." Or if a question that is borderline is getting activity, the new user chimes in with their two cents.

Preventing the post with protection serves two goals:

• It prevents the new user from having a worse user experience than being prevented posting - getting down voted and having the post deleted.
• It reduces the need for the 20k users to delete vote answers. And in the situation where the post gets up voted and attracts mod attention to clean up the dozen answers that somehow have an up vote and the mods then go through and wholesale clean all the poor quality answers out. This type of moderation doesn't scale well.

Thus, protection is a tool to reduce the amount of community moderation needed, or mods needing to come in and deal with all the answers. It prevents users from posting and thereby having a very poor user experience on the site.

If any restrictions on when a post is to be protected is put into place, one needs to carefully examine how these two goals of protection can be accomplished using other tools and the impact of the new user experience in situations where it should have been protected and now instead has their post deleted (either because its negatively scored now, or a mod came by and deleted a positively scored post and took away their points).

My contention is that the frequency of protection of a question being abused is less than 1 in 1000 protected questions. What needs to be done is making it easier to find these situations where the protection mechanic is abused rather than reducing the community's ability to protect itself and the need for people to explain the "why this answer marked down" comments on poor quality questions that are gathering poor quality answers from users who don't know better.

This might need to be coupled with better visibility of the 10k tools both to 10k users and moderators who apparently don't look at them frequently enough as the tools to prevent the sort of abuse described in the question already exist.

---

In my reviewing of recently protected questions, I found this: Dealing with failed sprints and deadlines which was protected by a diamond mod (which I assume isn't at issue), but rather I want to point out the various circumstances of the question:

• It was asked less than 24 hours ago (thus the mod protect rather than trusted user)
• As I write this, it has 5618 views.
• It has 8 answers.
• The least repped user answering the question has 552 rep.

If this wasn't protected by a mod, I would have been looking at doing it myself (as it shows up as #1 post on the 'most viewed 2 days' stats section).

The key consideration in this is that this sort of protecting shouldn't be restricted (given the time frame for a trusted user to protect it). Proposals that run counter to the ability for the community to handle such situations. The community needs to be able to protect questions where excess popularity can lead to diminished new user experiences.