Wikipedia talk:Community health initiative on English Wikipedia/Editing restrictions

Problems with current method of using editing restrictions

The Anti-harassment tools team can do a better job finding solutions if the problems are well defined. Please add to or elaborate on the existing problems with editing restrictions.

1. Editing restrictions can be easily violated which weakens the validity of sanctions if the edit goes unnoticed, or if noticed can require a time-consuming discussion.
2. Monitoring for violations requires a specific knowledge of the restriction and how to respond.
3. There is no technical barrier that reminds a user of their restrictions, which may lead to lapses in judgement when the temptation is too great.

Discussion of problems with current methods of using editing restrictions

• I don't deal that much with editing restrictions but my impression from what I've seen at ANI is that the biggest problems tend to be with cases that don't seem to be conducive to any sort of simple software solution. They tend to be cases where the person restricted doesn't feel that the restriction applies to whatever they are editing but others disagree. I mean I guess there are some more minor cases where the violation is disputed but almost no one, and the person claims to have forgotten about the restriction or just doesn't give any real reason, but how common are such cases? Nil Einne (talk) 17:52, 25 September 2017 (UTC)
  • Hi Nil Einne , At this point there is a lack of information about how well editing restrictions are working. The Anti-harassment tools team is doing some research on AN/I in order to better understand the types of cases brought there. Among other things, we're also looking to see if we can see patterns for why cases brought to AN/I are not resolved. We only have results from experimental/preliminary queries run for March and April 2017 so far. After we run more queries for the whole year, we should be able to give the community more information.
  • One area that we are exploring is improving the User Interaction History tool to make it easier to understand negative user interaction. Right now it is pretty time intensive to look through edits to see a problem. We're interested in understanding if and how improvements to the tool could assist with user conflicts. SPoore (WMF), Community Advocate, Community health initiative (talk) 22:12, 25 September 2017 (UTC)
• A particular restriction that pops up every now and then, but becoming increasingly common, is page bans, where an editor is restricted from editing or discussing a particular article. Currently there is no technical way to restrict an editor from editing a particular page so it relies heavily on other contributors keeping tabs on the sanctioned editor and reporting violations.
• In terms of specific restrictions from contributing to a particular sphere, there have been a few editors being restricted from editing in XFD's (MFD, AFD, CFD, etc), template space, as well as other administrative spaces (AN, ANI, AIV, ANEW, RFA). AFD bans, I would say XFD bans (qualitatively, from my observations of AN and ANI) are comparatively more common than the other administrative space restrictions, precisely because it is a forum where sides of a debate naturally form and acrimony can easily spill to other areas. Blackmane (talk) 03:12, 26 September 2017 (UTC)
  • Hi Blackmane, thanks for highlighting the types of bans and the namespace where they happen. This would definitely need to be taken into consideration when/if we look at tech solutions for user page bans or topic bans. SPoore (WMF), Community Advocate, Community health initiative (talk) 20:33, 27 September 2017 (UTC)

I don't think I'm even going as far as bans; I'm just talking about common or garden blocks. Consider the recent case of The Rambling Man; if you asked the key players in that feud on all sides of the debate "did you feel harassed", the answer would probably be "yes" from all of them. If you then asked them who was doing the harassing, they'd probably point fingers at each other. A more flexible number of options between doing nothing, locking the article to everybody, and blocking the user from everything, would allow more discretion to be used. Sure, we'll have to get policy drawn up at some point, but I would hope most people would be on board for that. Ritchie333 ^{(talk) (cont)} 15:14, 3 October 2017 (UTC)

I agree with the sentiment that we need more nuanced ways of blocking people. It's not just bans, as Ritchie said, the benefit is when someone gets blocked for causing a problem in a particular area and often it makes sense to have a narrowly tailored block to stave off the "but they contribute productively elsewhere!" issue so common to contentious blocks. Jo-Jo Eumerus (talk, contributions) 15:34, 3 October 2017 (UTC)

Jo-Jo Eumerus and Ritchie333, I understand what you mean and agree that it worth discussing. Using it as a substitute for full protection or a short user block, means that the article could stay open for editing by everyone else and the user could edit in other parts of encyclopedia. So, it might lead to a stabilizing the situation sooner. Less negative escalation, too, if it is not a total site block. SPoore (WMF), Community Advocate, Community health initiative (talk) 15:58, 3 October 2017 (UTC)

Scope appears to go beyond harassment

Outside of site bans, there are article bans, topic bans, and interaction bans. The only one of those three that seems relevant to this project's scope is WP:IBAN.

Would you please explain how the wide net you are casting is appropriate, and why this discussion should be about anything other than IBANs? Jytdog (talk) 04:09, 28 September 2017 (UTC)

Hi Jytdog,

The short answer is that the Anti-Harassment Tools team is part of a broad initiative to create a more welcoming environment on Wikimedia Foundation wikis. Since most harassment on wiki starts from content disputes, the AHT team identified early detection and improved community intervention in user content disputes as a key aspect of our work.

Additionally, it makes the most sense from the start to design a tool like user page block for broad and most inclusive use. This is especially true since it is common for several types of editing restrictions to be used simultaneously to each other on a case.

The AHT team encourages questions and input about the work that we are doing. You can talk with us on wiki, by email, or in an one to one interview. Let us know if you have any more questions or thoughts. SPoore (WMF), Community Advocate, Community health initiative (talk) 23:34, 28 September 2017 (UTC)

For more information

Read about the Anti-Harassment Tools team areas of focus as described in the grant request.

Or participate in the research on dispute resolution and harassment which will provide valuable information going forward for our team and the community to make decisions about tools and policy related to content disputes and harassment. SPoore (WMF), Community Advocate, Community health initiative (talk) 23:34, 28 September 2017 (UTC)

• Hey User:SPoore (WMF). So we are having a bunch of hassle with the Reading Team now. They started using a field from Wikidata called the "description" field to aid in navigation throughout the projects, and they just kind of slipped over and decided it would be great to show that field as the top line in en-WP articles in mobile view and in apps. We had an RfC in March and got them to take that out of the mobile view and we are now having a bunch of drama about the apps.

This arose from the WMF overstepping and moving into content decisions, an area where they have no right to be in the basic deal of this whole movement, which is that the volunteer communities generate content, and the WMF publishes it. They did this out a desire to help everybody, but they have created a whole shitload of problems that they didn't even understand. Wikidata has no BLP policy, no Verify policy, and there is no way to work out disputes between the en-WP and Wikidata communities which have entirely separate policies, guidelines, and governance. It just a mess. Not to mention the underlying governance issue of them overstepping in the first place. The folks who did this, had no conception of the boundary they were crossing, nor did they give any thought to getting consensus before they implemented it. They had no head-checking steps in their process to even stop and think about boundaries and consensus in affected communities and the like.

I see a similar thing going on here. Just like each editing community governs its own content (with some exceptions like COPYVIO), so too does each manage behavior issues that arise within it (with some exceptions like privacy).

I accept that the WMF has an interest in addressing harassment throughout the projects. But you are moving beyond that, and the rationale above is really stretching.

Just because you can overstep does not mean you should. The WMF has a lot of power - changing the platform we all work on, changes the experience and the interactions. And in my view by overstepping here this team is perhaps abusing that power. Please keep boundaries in mind and before you implement things, please please get consensus from communities that will be affected. Jytdog (talk) 03:30, 29 September 2017 (UTC)

Hi Jytdog, Here is a summary of why the Anti-Harassment Tools team is working on these features.

From 2015 Community Wishlist Survey/Moderation and admin tools#Enhanced per-user / per-article protection / blocking and Phabricator ticket T2674, the Anti-Harassment Tools team is aware that alternative types of remedies are used instead of full site blocks for some types of user misconduct. And over a number years, users have requested new features or improvement to software that would enhance the methods now used to enforce editing restrictions or other types of personal sanctions.

Editing restrictions address an user’s disruptive conduct, nothing to do with content decisions. The point of editing restrictions is to manage negative user conduct that has escalated to the point that it is interfering with day-to-day collaborative editing. The decision to enact an editing restriction on an user is done by the English Wikipedia editing community with no involvement from Wikimedia Foundation. Wikimedia Foundation involvement in this is to create tools for the community to use.

If we build tools that won’t be used, then we’ve wasted everyone’s time. So AHT team is committed to discussing everything we build with all stakeholders, both on wiki and off wiki. This is the purpose of this discussion page. :-) Additionally the team is available by email or user interview to hear input.

I'm off for the weekend, but will happy to answer any other questions on Monday. SPoore (WMF), Community Advocate, Community health initiative (talk) 23:54, 29 September 2017 (UTC)

i didn’t need the explanation of what various types of bans are for and you have not addressed the thing i am trying to call your attention to, which is that this is extending the scope of the work the anti-harassment team is doing into other behavior issues (not content issues). If you are going to define the scope of work that the "anti-harassment" team is doing to cover any problematic behavior then the name of the team and project is not appropriate. Thanks Jytdog (talk) 00:11, 30 September 2017 (UTC)

The idea should be that we discuss what to do, and how it is to be implemented here before implementation heads off into any wrong direction. Although a broader input should to be a good idea. Hopefully any negative impact is reduced, as it will be targeting harassment edits. Hopefully any features that are forthcoming for this will be configurable by en.wikipedia community in some form rather than an imposition. Graeme Bartlett (talk) 23:23, 3 October 2017 (UTC)

@Graeme Bartlett: That's our intention. If we design and build tools in a vaccuum, they may not solve any real problems and would be a waste of our developers' time. Configuration is also in the forefront of our mind — although we're focussing on English Wikipedia (due to its size) we want the tools we build to benefit as many wikis that want them, and no two wikis are the same. — Trevor Bolliger, WMF Product Manager 🗨 23:44, 3 October 2017 (UTC)

I'll repeat what I've said many times. Please build generic tools, not use case oriented features, especially not use cases oriented for English Wikipedia. A very useful tool is allowing a user to be banned from a page, with the option of also banning the user from the associated talk page. The typical additional field for that type of action is duration, and the existing duration field should be used. A form and a log entry are the only visible parts of the design, and the form will be seen by only a tiny percentage of users in a specific group to be determined by the community. Design done for the first iteration. That tool can be used by the community to achieve many different types of things, and it should be left to each community to determine how to implement that tool if at all. Then come back in a year and find what enhancements the community wants. John Vandenberg ^(chat) 09:07, 4 October 2017 (UTC)

@John Vandenberg: In many cases, building a multi-purpose generic tool is preferential to niche tools for the exact reasons you mentioned, but the risk is building a tool that does many things mediocrely and nothing excellent. The more this discussion progresses I grow more confident about the concept of adapting AbuseFilter to target the actions of specific users. — Trevor Bolliger, WMF Product Manager 🗨 18:49, 4 October 2017 (UTC)

Regex functionality

A quick technical suggestion: page warnings/bans/throttles/whatever should be optionally targetable via regular expressions (perhaps another checkbox on the form: "is this a regular expression?"). For instance, the above-mentioned XFD ban could be implemented in one action through the regex Wikipedia:.*?for\s(deletion|discussion). MER-C 13:22, 27 September 2017 (UTC)

Thanks for your suggestion. :-) Trevor (Anti-harassment Tools Team product manager) is following the discussion here. We bring ideas from these discussions to phabricator and to our team meeting. SPoore (WMF), Community Advocate, Community health initiative (talk) 20:42, 27 September 2017 (UTC)

@MER-C: That's a great idea. I can't comment on the technical feasibility, but that seems to be a clever way of addressing the shortcomings of single page blocking and the difficulty of topic blocks. Do you have any opinions if warnings, blocks, or throttles would be most effective? Our current thinking is just to build page (and now maybe Regex!) blocking. — Trevor Bolliger, WMF Product Manager 🗨 22:06, 27 September 2017 (UTC)

As written, only warnings and throttles are consistent with our current policy -- topic bans usually have the caveat of excluding obvious vandalism, and BLP violations. MER-C 06:31, 28 September 2017 (UTC)

Yes, I have highlighted that as an obstacle to using page and topic bans in the Pros and cons section. There would need to be a change in the way that the bans are written because a tech solution would not match the present sanctions. But there are alternative uses, too. Such as using user page block temporarily for several users who are edit warring on an otherwise stable article/policy page. This would be a substitute for temp Full Protections of the page. This case use would also need to be written into policy. There needs to be a community discussion about these potential policy changes. SPoore (WMF), Community Advocate, Community health initiative (talk) 16:27, 28 September 2017 (UTC)

These could be implemented using the edit filter functionality. Having the option to restrict the filter to one user; or perhaps one page would avoid the performance and management impact of thousands of edit filters. But it would allow the skills, interfaces and functionality to be reused. It would also allow warnings for good faith restricted people so that they can stop before saving. It would also allow for logging so that banned actions can be observed and not go unnoticed. Graeme Bartlett (talk) 23:12, 3 October 2017 (UTC)

This is a really good idea that we have not yet considered, thanks for suggesting it! I've read over every requested filter and many of them are rejected solely because Edit filter is not a per-user solution. And you're absolutely right about reusing the existing expertise of our Edit filter managers. — Trevor Bolliger, WMF Product Manager 🗨 23:44, 3 October 2017 (UTC)

Cluebot

For harassment in general, an AI bot something like cluebot could be used to detect harassment and possibly reverse it. Graeme Bartlett (talk) 23:14, 3 October 2017 (UTC)

I broke this suggestion into a new section so it wouldn't get lost in the Regex discussion. There are already some Edit filters in place for easily identifyable ("blatant") harassment, and we've discussed in the past that we could build a new Machine Learning scoring for harassing messages into ORES. Harassment is very contextual, so there are some severe limitations to these automated techniques but I agree they could certainly serve as a defensive layer against blatant harassment. — Trevor Bolliger, WMF Product Manager 🗨 23:44, 3 October 2017 (UTC)