Wikipedia talk:Autoconfirmed article creation trial/Archive 2

Trial duration
Hello everyone, I’m User:Nettrom -- if we haven’t talked before, I’m the researcher working with the Community Tech team on the upcoming ACTRIAL experiment. We’re preparing for the start of the trial -- I’m currently working on the study’s research questions and hypotheses, which you can see here, and developing the instrumentation that we’ll all use to evaluate the impact of the trial.

I’d like to revisit the discussion about how long the trial should last. The circumstances have changed significantly since the autoconfirmed trial was proposed back in 2011. We have more sources of data available and tools for analyzing that data, and the WMF is putting research and development resources towards the execution and analysis of the trial (e.g. the redesign of the landing page for non-confirmed users who attempt to create an article).

A key question to ask about the trial is: what is its purpose? Based on my reading of the two proposals from 2011, the ACTRIAL proposal and the duration proposal, the main purpose is to run an experiment to gather data about what effect restricting page creation will have so that the community can have an informed discussion about whether to implement it permanently or not. It would be preferable to complete the trial and its analysis as quickly as possible.

At the same time, I know that people have brought up important concerns about limiting the trial duration. Two questions that I have seen brought up are the challenges of handling seasonal effects, and that we need to study the longer term effects on editor retention and New Page Patrol.

When it comes to seasonal effects, they will always affect the study, regardless of when it is run and for how long. It is a reason why studies on Wikipedia might use a short timespan (ref: this paper) or sample across time (ref: this paper). There is several years worth of data that we can use for comparisons in order to control for seasonal variations.

When it comes to editor retention, running a longer trial would actually be counterproductive to our goals. The non-autoconfirmed users who begin editing during the trial period will form a cohort that we can follow over time, to measure their retention. Generating a massive cohort beyond the point of statistical significance won’t make the data more reliable; it just slows down the analysis.

We are of course also interested in understanding more about how the trial affects quality assurance processes like New Page Patrol. We have proposed several hypotheses about what the specific effects will be for NPP (and also AfC). Some of them concern the workload of NPPers (e.g. the number of review actions taken, and whether a small group of NPPers do most of the reviews), but also the quality of article entering the patrol queue. Similarly as for the previously mentioned concerns, the availability of historical data should allow for comparisons, effectively reducing the need for a longer trial.

In summary, there are several reasons for why a six month trial should not be necessary, and I would like to discuss that. Looking forward to learning more about what other community members think about this! Regards, Nettrom (talk) 20:41, 7 August 2017 (UTC)
 * Hi, . I'm certainly opposed to a short duration. I think there are things other than quantitative analysis that the community will want to judge through their hands on experience working with the trial: the data is important but so is the observations that reviewers and community members make, and I don't think a month as Danny had suggested will be long enough for most people to make a judgement on this. The consensus was for six months in 2011. I don't see any reason not to have it that long other than the funding concerns Danny had over your position. had suggested splitting the baby and doing it for three months in the past, which I could possibly get behind if this is something that the WMF is going to insist on. The editor retention data analysis shouldn't have that significant an impact for the basic data we need for a followup RfC on whether or not to make this trial permanent, and we should be able to use sample groups from previous time frames as well there even if the cohort is larger. TonyBallioni (talk) 20:54, 7 August 2017 (UTC)
 * Hi TonyBallioni, thanks for commenting on this! When it comes to quantifiable data and tests to be run on those, one issue that is likely to come up with a longer trial (and the larger cohorts that come with it) is that we'll find minuscule but statistically significant differences. Given that the trial is intended for a community discussion, I would not be surprised if those kind of results reduces the productivity of that discussion because it's difficult to interpret whether those kinds of results are meaningful or not.
 * I'm curious to learn more about the non-quantitative analysis that you mention! I've read a fairly large amount of historical info on ACTRIAL, but haven't really seen it discussed much. For example, the statistics discussion only describes quantitative data, and our hypotheses and discussion on meta haven't brought it up. From my researcher point of view, I'm definitely interested in other types of data, but that also means we'd have to figure out ways of gathering it. For example, I see talk page discussions about what goes on during the trial as useful, but I'd also be very interested in hearing from those who don't participate in those, meaning it would require surveys or interviews.
 * You wrote: I don't think a month as Danny had suggested will be long enough for most people to make a judgement on this Could you explain a bit about what makes you draw that conclusion? I'm not sure who "most people" are in this context, and why they're a key group. Cheers, Nettrom (talk) 17:03, 8 August 2017 (UTC)
 * Thanks for the response ! I do appreciate it. In terms of data: I was always under the impression that a larger sample size with statistically significant differences was typically a good thing and more accurate: a smaller sample size might be easier to discuss, but what we would be discussing would also probably be less meaningful.In terms of qualitative things, there are several: first, the quality of articles. We can measure deletions statistically, but whether or not ACTRIAL improves the raw product coming in beyond that is going to be difficult to measure using a number. Next, whether there is a boost in morale for the editing community: I think seeing the impacts on morale would be better judged with a longer time frame, and are just as likely to be accurately measured through comments in an RfC as a survey instrument.Finally, there is the volunteer hands on experience that should be used to compliment the data: to use an example from political campaigning, all the data in the world can tell you who you need to talk to or what you need to do, but it isn't likely to be effective in achieving the goals in terms of votes at the ballot box unless you take into consideration the input of people in the communities that it measures who have been working and living there over time. Our reviewer base is very dedicated and probably knows the quality of the content coming into the encyclopedia better than virtually anyone. While the data is very important, getting their take on explaining the numbers is as well. I think it would take more time than a month for most people to notice trends that the data might point out and to form opinions as to what causes them. A longer trial period is needed so that the human factor has time to process the changes that are happening: both positive and negative. TonyBallioni (talk) 17:22, 8 August 2017 (UTC)
 * You're most welcome, TonyBallioni, and thank you for your responses, I appreciate the discussion! Your impression about sample sizes is correct, you want a sample size that is large enough to achieve statistical significance (determining the sample size is often done with a power analysis (see also sample size determination), but you might already know this). What I'm referring to is what can happen when you increase your sample size further. As it increases, you'll see statistical significance for increasingly smaller effects. It might then become a judgement call of whether the effect you're seeing is really meaningful, and I've seen plenty of papers that ignore what might appear to be statistically significant results just because their sample size is huge and the effect is small. I'm somewhat worried that we'll end up with those kind of judgement calls with a longer trial, and I'm not sure how they'll affect the post-trial discussion.Thanks for bringing up article quality, one of my favourite subjects! I won't talk much about it, I'll just mention that we have hypotheses about how it'll be affected. When it comes to boost in morale, I see your point about how that might take time to change, but I also haven't seen anyone talk about how that is a part of the trial until now. The focus has instead been on for example restricting the influx of lowest-quality content (e.g. spam) in the article namespace. I'm not against making qualitative aspects a part of the trial, but then we should be discussing how we go about measuring those (on the research page on meta). With regards to surveys versus RfCs, I think we'll just have to agree to disagree on that :)I liked your example from political campaigning, and I want to make sure that the "boots on the ground" are a part of this trial as well. That's why my hypotheses are up for questioning, why I'm publishing work logs whenever I make progress that I think others might be interested in, and hope that we can have good discussions about our findings as more data comes in. The way I interpret your last point it seems to suggest that there should be a longer time post-trial for analysis, reflection, and discussion, not necessarily that the trial itself needs to run longer. Cheers, Nettrom (talk) 01:15, 9 August 2017 (UTC)
 * Re: the timetable, which is what we are discussing here, I actually disagree that there needs to be a longer time period for analysis afterwards. What I was trying to convey is that assessing the impact of the trial by our volunteers based on their experience will take longer than a month to get the feel here and form opinions while the trial is going on. I think the other response here give very practical reasons why a longer trial is warranted, but in terms of the followup to the trial/the research, I don't think the community will have enough time to understand the implications of the change with only a month. I'm certainly not doubting the value of data (I use it frequently in the workplace), but do want reviewers and other community members to have enough time to observe the impact and implications of ACTRIAL themselves as a lens through which to view the data. Hope that is clear. I'll comment on meta tomorrow about the other research things you mentioned. TonyBallioni (talk) 01:49, 9 August 2017 (UTC)
 * Thanks for explaining that TonyBallioni, I see what you mean now! Cheers, Nettrom (talk) 18:07, 9 August 2017 (UTC)


 * Hi . Thanks for your work on this. In terms of purpose and trial duration, one point I thought was worthwhile from earlier discussions was that a too-short trial risks giving us an inflated impression of the positive effects of an AC limitation. Introducing a new limitation may be accompanied by a learning curve in which some users (e.g. paid editors) adapt to game the system (e.g. we may see a marked rise in accounts opening one day and then editing four days later), so I hope the trial will run long enough to allow any unavoidable regression to the mean, so to speak, to take place, and then see if the AC limit still improves things or if it only meaningfully change cruft, backlogs, etc. at first, then things backslide. I don't have a firm number in mind for how long it'd take to see that though; would be interested in others' guesses. Innisfree987 (talk) 21:55, 7 August 2017 (UTC)
 * Hi Innisfree987, thanks for commenting on this! You bring up a good point about learning curves. Based on what I've read about some of the paid editing incidents, I would expect them to adapt rather quickly, because reaching autoconfirmed status isn't that big of a hurdle. We've got a fairly extensive set of hypotheses up on the research page on meta and they cover for example number of created accounts, how many and how quickly accounts reach autoconfirmed status, and how many drafts are submitted to AfC. Through those, we should be able to capture changes in new account behaviour in a fairly short time. Cheers, Nettrom (talk) 17:03, 8 August 2017 (UTC)
 * Agree with Innisfree987 here. We cannot know in advance how long it may take for less sophisticated COI editors to "get wise" to the change, and for advice pages on the internet on the lines of "how to get yourself/your business on WP" to be adapted to help such editors get round the new barrier. Since we can't be sure it seems better not to limit the trial period on the basis of someone's "I would expect" Noyster (talk),  11:24, 9 August 2017 (UTC)


 * I haven't yet seen a compelling argument for a short duration and I've seen many that indicate that a longer duration is better. I think that a significant pushback from the community would be experienced if anything was implemented shorter than the 6 months decided by RfC. Without a very strong and compelling reason not to, I see no reason to deviate from the RfC's 6 month figure. —  InsertCleverPhraseHere  21:40, 8 August 2017 (UTC)
 * I've seen enough at NPP, AfC, and in handling drafts to believe no new editor should be creating new pages. If ACTRIAL reduces the rate of new mainspace page creations (as we expect) we'll need most of the 6 months just to plow through the NPP backlog already existing. Legacypac (talk) 21:49, 8 August 2017 (UTC)
 * The community set the trial duration at six moths. After giving it careful thought (again) I cannot find any compelling argument for making it shorter. What I do see is that it's taking an interminable time to get this trial set up and my latest information is that it will begin sometime in September. This trial is a community driven project and requires the participation of dedicated volunteer users. If we are to retain their interest, we should do so without redebating elements of it that have already been signed an sealed.
 * I notice also that states above:
 *  ...there is the volunteer hands on experience that should be used to compliment the data: to use an example from political campaigning, all the data in the world can tell you who you need to talk to or what you need to do, but it isn't likely to be effective in achieving the goals in terms of votes at the ballot box unless you take into consideration the input of people in the communities that it measures who have been working and living there over time. Our reviewer base is very dedicated and probably knows the quality of the content coming into the encyclopedia better than virtually anyone. While the data is very important, getting their take on explaining the numbers is as well. I think it would take more time than a month for most people to notice trends that the data might point out and to form opinions as to what causes them. A longer trial period is needed so that the human factor has time to process the changes that are happening: both positive and negative.
 * I cannot express strongly enough the importance of empirical findings. I am sure that no one within the WMF has the hands-on experience of the regular users who have been doing NPP for years (both for patrolling the pages and for monitoring the performance of the system), I am therefore at a loss to understand why each time I mention this, it is glossed over. ACTRIAL is about protecting the very fabric of Wikipedia and retaining the interest of its users. It is not about data and stats. Data and stats are just a means to an end.
 * adds a further important observation which I have also made several times in the past. We do need to reduce the backlog and based on the current performance of the New Page Reviewers, it will take all of six months to clear it, if not longer. Kudpung กุดผึ้ง (talk) 23:16, 8 August 2017 (UTC)


 * ACTRIAL deserves six months to a year. A year of the program would allow our reviewers to catch up on the backlog, begin to assess how the quality of new articles change, and perhaps identify how long it took paid editors to design new end-runs around the system. A year has definite upside and I see no real downside. The claim that "when it comes to editor retention, running a longer trial would actually be counterproductive to our goals" is without evidence. A year of new editors against what has been a sixteen-year project in Wikipedia isn't a massive cohort beyond measure. After all, you should probably value the retention of the beleaguered reviewers and other long-term volunteers that wrote this encyclopedia at least a much as you'd want to retain potential new editors that have done nothing for the project so far. Chris Troutman  ( talk ) 23:59, 8 August 2017 (UTC)
 * I'd like to see if we can measure some of the factors that people have brought up in this discussion. We're looking at this as an experiment, where we can make hypotheses and test outcomes. Right now, there are several measures on the Research:ACTRIAL page that will help us understand the New Page Patrol experience:
 * number of patrol actions
 * number of active patrollers
 * distribution of review work across patrollers
 * size of the backlog
 * quality of articles entering the queue
 * survival rate of new articles by autoconfirmed users
 * If folks can think of other measures we can use, that would be really helpful. As Chris says, the research is meant to measure both the impact on New Page Patrollers as well as on new editors, and other people and processes.
 * There's one factor that's only a problem on our side: we have a three-month contract with Morten -- to figure out hypotheses and data sources, to run the experiment, and to do the analysis and participate in the discussion. That's our problem and not yours; you're not responsible for our funding. But it means that Morten, Ryan and I need to figure out the best way to support the trial using the resources that we have.
 * That might mean doing some analysis mid-trial; we don't need to have ACTRIAL "turned off" in order to analyze the data and discuss what's happened so far. The goal for us is to have data that we can all talk about -- New Page Patrollers, WMF/Community Tech people, and other interested community members. We can have that discussion in multiple stages, if we need to, and the trial can still be running while we look at how things are going so far. The trial duration isn't a hill that I personally feel like dying on. -- DannyH (WMF) (talk) 15:13, 9 August 2017 (UTC)
 * I think there is going to be a lot of talk during the trial anyway, and that's what I would expect. However, I don't see any harm in getting Morten to set up the formulae for extracting the stats then coming back for a few hours at the end of the trial to extract his analysis. Perhaps you could even ask . Kudpung กุดผึ้ง (talk) 15:33, 9 August 2017 (UTC)


 * At least six months is the minimum time that would be meaningful for the community, particularly the small portion of the community that deals with new pages. A statistician may be satisified with, say, one month, but satisfying statisticians is not the purpose of the trial. Satisfying the WMF is also not relevant, unless the WMF has plans for maintaining the integrity of the encyclopedia. The exact length can be revisited after two or three months if circumstances show that would be warranted. Johnuniq (talk) 01:01, 10 August 2017 (UTC)
 * While a one month trial may be enough to measure an immediate impact on new accounts, it's going to take significantly longer for the full effects to percolate through the system. How long is it going to take for NewPagePatrol to clear the backlog? And how about ArticlesForCreation adapting to new submissions and processing them out to mainspace? It's certainly reasonable to look at the data along the way, and we can always consider revising the plan based on data and community experience. But we should probably presume the original six months. Alsee (talk) 04:06, 10 August 2017 (UTC)
 * I'm opposed to doing a trial of below three months, as we need a minimum of one month to answer the questions that the research on this is trying to answer, plus the time for the news to get around. I am also opposed to doing a trial with a time above 6 months, as a time above that might have real, lasting consequences on Wikipedia is stuff goes badly. RileyBugz 会話 投稿記録  21:03, 10 August 2017 (UTC)
 * I'm happy with the current plan (which appears to be for a 3-6 month trial starting in September). My personal recommendation would have been to start a 7-day trial this Thursday, August 17, with a longer trial starting in September if there are no unexpected issues. Power~enwiki (talk) 22:55, 13 August 2017 (UTC)
 * I'm not sure what the proper length of the ACTRIAL should be. But committing to a trial of six months in advance seems unwise and unnecessary. I'd like to advise caution and an openness to being guided by the data.
 * My main concern is that we don't yet know whether restricting access to article creation damages recruitment and retention of productive new editors. This is one of the hypotheses  the trial seeks to test.  If it doesn't cause damage, then restricting access may become permanent. But if restricting access does cause damage, then restricting it for a longer time will cause more damage. That may not be “a compelling argument for a short duration,” to quote, but I hope it's a compelling argument for caution.
 * If can get meaningful results on this and other questions after a month or two, then that will give us good information about whether it's safe to continue.   speaks about “the purpose of the trial.” I know that some members of NPP feel that they need a longer break to catch up and build morale. But the primary purpose of the trial is to determine what is best not only for NPP but for the health of the wiki as a whole.  So with that as our goal,  let's look at the data as we go along and make good decisions based on what we find. JMatazzoni (WMF) (talk) 23:17, 16 August 2017 (UTC)
 * User:JMatazzoni (WMF) you are walking on extremely thin ice here.  the comment just prior to yours, away below here is much better.  If you are not aware that the the team at the WMF working on this in the past generated mountains of distrust (earned bad faith) through its terrible behavior thwarting this trial in the past, be aware of it.  Even hinting that WMF might shut this down unilaterally early is a very bad move.  Jytdog (talk) 23:37, 16 August 2017 (UTC)
 * and : My post below that Jytdog linked to is the latest info from the Community Tech team. Things are getting a little confused on this page, because there's three threads (Trial duration, Start date and Check list) that are all touching on similar themes. JMatazzoni is saying essentially the same thing as I did, in a more anxious tone. :) Community Tech is committed to running this trial in good faith with the community. See below. -- DannyH (WMF) (talk) 00:10, 17 August 2017 (UTC)
 * "anxiety" was the exact line of concern trolling that was used to thwart this effort years ago. There is no room for that anymore - we are not interested in the "anxiety" of WMF employees.  I did read your note below and that is much more correct.  Jytdog (talk) 00:22, 17 August 2017 (UTC)
 * , sorry -- I didn't know that the word "anxious" had historical baggage. I agree that WMF employees are responsible for our own emotional responses. -- DannyH (WMF) (talk) 00:25, 17 August 2017 (UTC)

My original post below today was because I was trying to get a firm commitment from the WMF on a time frame: there has not been a single community member here that has objected to six months, that is what the original consensus was for, and no one except WMF have expressed any trepidation at the length. There are two important reasons that I think having a public commitment to a specific time frame are important: An indeterminate trial period with results ideally after a month but no intentions on shutting it down if we want it longer is not what the community agreed upon 6 years ago, and it is not what anyone who has been involved in the conversation who is not a WMF employee (whether in a volunteer or official capacity) has suggested would be a good idea. A commitment to respecting these wishes by agreeing publicly to a set end date is important for restoring the trust of the community. TonyBallioni (talk) 00:55, 17 August 2017 (UTC)
 * 1) The minority of the community that opposes restricting creation to AC accounts has expressed fear that this will be like Jimmy Wales' fiat on a temporary basis that IPs couldn't create new articles: we are still currently undergoing that trial over a decade later. I've said very publicly by anyone who has challenged me on this that I will be the first to oppose this becoming indefinite without an RfC. Having a firm date makes it clear to those people that those of us who have staked our reputations on getting ACTRIAL up and running that we are also representing their views to the Foundation. I take this very seriously, because I would expect if I were in their shoes, I would want the proponents to consider my concerns as well.
 * 2) As mentioned above, there was great distrust created within the community after programmers shot down a community mandated initiative. I have probably been the most WMF friendly of any of the people who have commented in these discussions over the past few months, but I also wasn't active when that happened. I see many colleagues whom I respect and whom i don't consider to be paranoid express frustration anytime we get anything that looks like vacillation from the WMF: that's because they have been here through the last 6 years of the WMF blunders in community relations. A set timeframe, with the understanding that if the Wiki starts falling apart we can pull the plug, relieves these fears because it means that there it is a public record of when it will be happening.
 * Okay, that makes sense. As I said earlier, this is not a thing that I want to fight about, and I understand that the WMF has lost credibility overall. We can commit to six months for this trial. We're going to use the first month's data to look at how the change is affecting the wiki, and after that first month, we're going to publish our findings and the data we've collected, for discussion with the larger Wikipedia community. That data analysis and discussion will happen while the trial is still running.
 * There are two reasons why we decided to help run this experiment. First, we're genuinely curious about what's going to happen, and we want to use what we learn to inform future work on the new editor and page creation/curation experiences. Second, we want to take a step towards repairing some of the damage that's been done to the WMF/community relationship. We're taking both of those goals very seriously. -- DannyH (WMF) (talk) 16:45, 17 August 2017 (UTC)
 * , thanks for your response and your willingness to engage with the community to rebuild trust here. It is very much appreciated :) TonyBallioni (talk) 00:26, 18 August 2017 (UTC)

Start date

 * the conversation above seems trending towards a longer trial. I know Kaldari said that we were thinking of going live at some point in September. Since we're almost half way through August, I think we need to set the actual start date soon for two reasons: first, it will allow us to give people advance notice of the actual date of the trial. Second, it will put pressure on those of us who want to be involved with the research planning to get a finalized set of research questions (and I'm speaking in terms of the community here more than the WMF. I have interest in these, but haven't had the time to provide some of the feedback I said I would above. I suspect some others might be similar.) TonyBallioni (talk) 14:33, 11 August 2017 (UTC)
 * Also pinging in case they have thoughts on this. I think I pinged everyone who commented in the duration section above, but if I missed someone, I apologize. TonyBallioni (talk) 14:37, 11 August 2017 (UTC)
 * I support immediate implementation. I don't know who needs what time to prepare. Assuming the interested parties can get their act together, I'd like to see implementation yesterday, if not sooner. Chris Troutman  ( talk ) 14:56, 11 August 2017 (UTC)
 * I suggest Tuesday, Sept. 12, as that should give enough time to formulate the research questions and such and it gives everybody one day to fix anything before we start (assuming that they work 6 days, taking Sunday off). RileyBugz 会話 投稿記録  15:15, 11 August 2017 (UTC)
 * I think that as far as the community is concerned, we are pretty much on our toes. Most of it was prepared by Scottywong and me already. There isn't really much else to be done - we already figured out the list of required minor GUI changes and someone with admin rights will need to make them. As pointed out, to keep the results as an accurate comparison, not much should be changed apart from integrating  that splash  page somewhere. The actual resetting  of the permissions in the MediWiki software is straight forward enough, failing  which  we can always do  it locally  with  one of the filters that have been suggested. As I'm still very committed to avoiding discouraging bona fidae good faith  editors from using the wizard (which is the other 50% of this exercise), I'll finish redacting  the wizard's TL;DR info pages and ask  for copyedit/feedback.  I think a roll out for Monday 4 September would be fine (some of us volunteers hold down daytime jobs, or run the local snooker club during our retirement). Kudpung กุดผึ้ง (talk) 15:55, 11 August 2017 (UTC)
 * Kudpung กุดผึ้ง,, I'm concerned with the test-philosophy expressed above. I understand the interest in limiting changes, to try and get a clean test of a single thing. However I believe the purpose of this test is to explore this option as a new approach. If ACTRIAL turns out to be successful and permanent, naturally we will do everything we can to adapt and improve the system. I believe the intent and purpose here is to try this approach, and make any and all changes that will help new users under this system. A new landing page should be considered a minimum starting point, not an end point. My view is that we should pretend this is permanent, and we do whatever we would normally do to make it work as well as possible. Alsee (talk) 23:00, 11 August 2017 (UTC)
 * , I think it's important to keep focus on this as an experiment. We're all going to be using the results of this trial to talk and make decisions about whether these changes should be permanent or not. If we pretend that it's permanent, and throw more elements in as the trial is going on, then it would be harder for us to tease out what changes are responsible for the outcome. For example, if we initially restrict article creation for nonautoconfirmed users and then two weeks later introduce a brand-new creation workflow, and the trial is successful by lots of measures -- is the success attributable to the restriction, or the new workflow?
 * To be realistic, I'm not sure if any solution is going to be universally successful -- this is going to impact a number of different groups and processes, and life is complicated. We want to understand that impact as well as we can, and decide together whether it's a trade-off that we're comfortable with. So it's possible that we'll conclude "Everything worked, let's do it all!" -- but in my opinion, it's more likely to be mixed, and we want to understand what was responsible for the change. -- DannyH (WMF) (talk) 17:21, 12 August 2017 (UTC)


 * I'm happy with the WMF's collaboration on this project, and I think the community can deal flexibly with starting date. I would invite, and give a lot of deference to, the WMF's preparation timetable. Alsee (talk) 22:31, 11 August 2017 (UTC)
 * Some time in the first couple weeks of September, with at least 2 weeks notice given of the exact start date so that users get the chance to get up to speed on what it means (some users only edit on the weekend, so potentially might miss on any messages about ACTRIAL for up to a week). Given this, I think that the specific date should be chosen in the next week from now. I also think that this should be advertised widely two weeks prior to implementation to users once we have a specific date, on the Watchlist page, various noticeboards, and as a talk page advert sent out to anyone involved in NPP, AfC, major wikiprojects, the teahouse, etc. —  InsertCleverPhraseHere (or here)  23:20, 11 August 2017 (UTC)
 * We have draft messages that will go out to all admins and NPR rights holders. I don't think a watchlist notice would be appropriate since by the time it goes out, we will be done with the majority of the discussion and watchkist notices are typically reserved for major discussions: same with CENT, though that might make more sense. Posting notice in highly visible places makes sense. My only concern with sending out too many user talk messages as a mass message is that many people might fall on multiple lists. Basically we want to make this as known as possible without being disruptive. As an aside, the current discussion at VPR gives me the impression that many people think ACTRIAL is permensnf policy that's already in place, so I suspect some people will be shocked by the message :) TonyBallioni (talk) 23:37, 11 August 2017 (UTC)
 * I am of the view that some-day(preferably 4th) in the first week of September after Kudpung finishes up tidying the wizard pages will be desirable.And I am against a watch list message! Will cut unnecessary drama by a mile.:---Don't we need to inform folks actively involved in editathons etc.Or are they already covered in NPP+Admins group? Winged Blades Godric 01:39, 12 August 2017 (UTC)
 * , yes: I think letting the Women in Red/Art+Feminism people know is important as well as probably post notices to all of the local Wikimedia chapters/local Wikimedian groups talk pages (and any where else that might host an offline event). The point I was trying to make is that many people are on a bunch of different mailing lists: we don't want someone ending up with three notices on their talk page. Sending a mass message to a limited number of lists with a large number of users I think would be sufficient. From there if you post it widely enough on project talk pages it ideally get the message out without people getting duplicate messages. TonyBallioni (talk) 02:03, 12 August 2017 (UTC)


 * Personally, I think ACTRIAL should be rolled out  with  the minimum  amount  of fuss and publicity as possible. It  only  directly  affects new users who  won't  be aware of the change anyway, and it doesn't change any of the bells and whistles of anyone's maintenance workflow. Too much publicity will prepare people for it and give a false reading to the stats. Depending on the results of the trial, any wide reaching changes to policy and/or software that might be necessary can be notified then. Kudpung กุดผึ้ง (talk) 02:09, 12 August 2017 (UTC)
 * Roll out the changes right now or ASAP. The more people we notify the faster the word gets out to COI editors that create new accounts everytime, and allows them more time to figure out workarounds. The leaders in NPP and AfC are well aware already. The work flow for the typical NPP and AfC reviewer is exactly the same. We already know what happens when uninvolved editors notice ACTRIAL and decide to stir up trouble based on incomplete understanding. This impacts two types of editros - COI spammers and Good Faith but inexperienced newbies. For all the existing editors doing NPP this is only a positive. Legacypac (talk) 08:08, 12 August 2017 (UTC)

Hi, sorry that Kaldari and I have been slow to respond here: we're at Wikimania this week, and it's been very busy. We need to complete work on the simplified version of ArticleCreationWorkflow, leading to the landing page. We also need to make sure that all of our data collection is instrumented, and that we're getting good data. We'd like to propose starting the experiment on the afternoon of Thurs, Sept 7th. (If we need to do any last-minute fixes, we'll be able to deploy them on Wed to reach EN Wikipedia on Thursday.) The redirect/landing page extension needs a security review, which is a little unpredictable, because it's out of our hands -- the Security team has to review it. We expect to have that finished by Sept 7th, but I'm letting you know that's a possibility.

I know it's tough to wait; we want to get this started as soon as we can. We'll be giving regular updates on how things are going; we'll have a firmer idea when we talk to the Security team. What do you all think? -- DannyH (WMF) (talk) 15:14, 12 August 2017 (UTC)
 * September 7th sounds good for a launch date. At what point would this be locked in so that we can let people know in the appropriate forums. Thanks, TonyBallioni (talk) 18:16, 14 August 2017 (UTC)
 * Sounds good. —  InsertCleverPhraseHere (or here)  19:24, 14 August 2017 (UTC)
 * Yep, works for me. DrStrauss (locked out on wikibreak, will confirm identity when I get back) 77.99.24.149 (talk) 10:53, 15 August 2017 (UTC)

Supporting edit-a-thons and similar events
Before we launch ACTRIAL, we need to make sure we have proper guidance in place for event organizers. We also need to figure out how event organizers should handle event participants who want to create new articles. The solution we come up with doesn't have to be a permanent solution, just a solution that will work during the trial. Some options include: Thoughts? Kaldari (talk) 23:55, 18 July 2017 (UTC)
 * 1) Let event participants follow the Article Wizard workflow (along with everyone else) and create the articles in Draft namespace.
 * 2) Create a new   user group that includes the   right (to create new accounts without hitting the rate limit) and a new   right that allows them to grant confirmed status to new accounts. This would be a temporary (expiring) user group, similar to Account creators.
 * 3) Give them a back-door mechanism (for example, a special URL) that they can use to bypass the ACTRIAL restrictions.
 * 4) Direct them to use Requests for permissions/Confirmed and change the guidelines there to support granting confirmed status to event participants for the duration of ACTRIAL.
 * Looping in here per his above comments on working with editathons. My preferred option is option 2. I'm also not entirely comfortable with the idea of creating a new user right without consensus to do so and think it might be worth adding in the followup RfC if it becomes seen as needed. If others think it would be non-controversial I wouldn't oppose it. Option 4 I think is a possibility, but we'd need to have a place to alert admins of upcoming events that would need confirmed status to be granted in advance. I don't think either option 1 or 3 is good. TonyBallioni (talk) 00:28, 19 July 2017 (UTC)


 * Hi.
 * Possible, and seems to me to be the logical approach. We need to go through the Wizard though anyway and start reducing the walls of text it forces the users through. (I've already started doing this). The editathon facilitators do not have to be an admin to approve and move those those drafts mainspace. That said, having facilitated on editathons, the main objective is to teach people how to edit and/or create articles. The channel they use to finally get their creations to mainspace is secondary and I never met an editathon participant who was worried about having to wait 4 days. They were all ore concerned about learning about the policies and guidelines and how to use the interface. We could use some input here from  regarding his wide UK experience. In view of the IEP disaster, I don't think any special exceptions should be allowed for Educational Programmes. However, Alex is a sysop AND an employee so can do what he likes. The Wiki Education Foundation by contrast, is still run by the people who refused to listen to the volunteers and hence caused the IEP fiasco and the loss of around $2mio it incurred - that kind of money could have bought all the software we need, and more.
 * involves unbundling (even temporarily) an admin right, creating new user groups and all the paperwork, and it would have to be approved by community consensus - lead time approx 8 - 10 weeks.
 * could probably be too easily hacked and abused by determined spammers, especially paid advocates and SEO experts.
 * Possible. Requires an admin to be on duty (either on or off-Wiki) during the event. Easy enough to find an admin who (like me, for example) is online practically 24/7. On the other hand, processing at WP:PERM isn't always as prompt as one might imagine. Conclusion: be sure an admin agrees to be available beforehand.


 * Please let's not allow these issues to become an excuse for delaying the actual development of ACTRIAL coding requirements. Kudpung กุดผึ้ง (talk) 00:43, 19 July 2017 (UTC)
 * Thanks for sharing your thoughts. I don't know if you saw the announcement at Wikipedia talk:The future of NPP and AfC, but we've already started coding the backend for the new (simplified) ArticleCreationWorkflow. That's the reason I'm asking about edit-a-thon support actually, since options 2 and 3 require some software support. Better to figure out the requirements now while it's being developed than try to tack it on afterwards. Kaldari (talk) 01:36, 19 July 2017 (UTC)
 * My post above was an evaluation. My opinion is that #4 is the easiest and least bureaucratic method. A new tradition has been begun of sending  all  admins a monthly  (or as needed) newsletter which  is a welcome  development. I think its editor-in-chief is . We  also  have the New Page Reviewer newsletter which  I  started, and which   has kindly  taken under his wing. It  can warn patrollers to  be on  their mettle for  up coming  events. This will  also  help  avoid a repeat  of the IEP debacle. Kudpung กุดผึ้ง (talk) 02:13, 19 July 2017 (UTC)
 * 4 is certainly the easiest. 2 is my ideal, but I see that as something that can be added as part of the followup RfC or something separate. TonyBallioni (talk) 02:16, 19 July 2017 (UTC)
 * I agree 4 is the easiest, and I engage at least temporarily to be available remotely to do this in reasonable time zones for me.(EDT). But there is no solution except 1 for events that don't organize themselves properly, & anyone has the right to run their own event on editing WP.  DGG ( talk ) 04:43, 19 July 2017 (UTC)


 * Don't have strong opinions either for or against it.I have seen volunteers working on these aspects(GLAM et al.) speak in diametrically opp. tones about the user-mindset, user-patience to wait for 4 days etc.On hindsight, trash content generated from these events are quite low-volume and could be prob. tackled at NPP.So, slight skew towards oppose.But per, there is no solution except 1 for events that don't organize themselves properly.
 * IMHO, is the best option.Dis-bundle the  to those holding the   subject to certain restrictions.But the downside is that the process to dis-bundle(holding RFCs...) is time-consuming.
 * Dead-against this proposal.Too prone to be hacked and I, at any eventuality, don't expect to see video-guides springing up on Youtube teaching spammers to generate URLs that bypass the AC Requirement.
 * Is certainly the easiest.But admin-availability and the severe backlog of PERM-boards at certain span of days look concerning. Winged Blades Godric 08:51, 19 July 2017 (UTC)
 * For 'time consuming,' read 'years' :D &mdash; fortuna  velut luna  09:03, 19 July 2017 (UTC)
 * Tried to be optimistic....And some things are better left un-said! Winged Blades Godric 09:27, 19 July 2017 (UTC)
 * ...or unreadable ;) &mdash;  fortuna  velut luna  09:40, 19 July 2017 (UTC)

As I was pinged, I'll try to set out my thoughts on the potential impact of these suggestions on events that I run: TL;DR: I don't expect ACTRIAL to have any noticeable impact on events I'm involved with. I know that other trainers will have their own way of doing things, and I'm not asking anyone to change the way they work, but I do believe that it's possible to adopt currently available training methodologies to reduce the impact of ACTRIAL to almost nothing.
 * Because I prefer to have first-time editors (the vast majority of those I train) work within their own user space to begin with, I have rarely come across a problem arising from their non-confirmed status. Anyone working in their sandbox won't need special measures because of ACTRIAL.
 * Because I encourage first-time editors to improve existing articles rather than creating new ones (because of the greater skill-set required for the latter), it is unusual for me to come across new editors who can create a sustainable article within the confines of an event. However, some events specifically try to create new articles. In those less common circumstances, I still encourage new editors to work in their sandboxes and I simply move the sandbox to mainspace for them when we both agree that it's ready (sometimes days after the event because we keep in touch). It would be useful as an event organiser to have the extended mover flag, but it's not really necessary.

As for the suggested options: Hope all that helps. --RexxS (talk) 16:28, 19 July 2017 (UTC)
 * 1) AW + Draft space is fine, and would work with anyone who is comfortable guiding new editors through Article Wizard and using Draft space. There's a chance that AfC reviewers may be helpful with a paricular article in the longer run, but I'm wary of imposing an extra workload on volunteers who already find difficulty in keeping up with current demands.
 * 2) Being able to grant autoconfirmed would come in handy once in a while (especially for killing off those annoying captchas that confuse new users as soon as they add a reference. FFS, we want to encourage them to use references, not put them off). However, I'm not sure I could justify "dead" time taken out of an event while I went round giving autoconfirmed to the new editors (they never tell you in advance what their user names are).
 * 3) A backdoor is a non-starter for what should be obvious reasons.
 * 4) Either they go to PERM during the event, thus killing time at the event that would be usefully spent on something other than learning about Wikipedia bureaucracy, or they do it before the event (thus killing their own time that would be usefully spent on something other than learning about Wikipedia bureaucracy), and to be honest, it's hard enough getting them to register a username and remember the password, so I really don't want to give them another hurdle to get over just to take part in a Wiki-event.


 * Hi . I concur with most of what says. Most edit-a-thons we're seeing/supporting have newbies either improving existing articles, or writing new articles in their sandboxes first with autoconfirmed users moving them to mainspace when ready. It's not ideal, but it works, and I don't think ACTRIAL will change that (?). Many of Wikipedia's existing/formal systems are so complex we've all developed other workarounds, which means this trial may not impact edit-a-thon organizers anyway - that sounds odd, but there you go. You might check with Art+Feminism and AfroCROWD organizers to confirm with them too, if someone hasn't already reached out to them? In terms of options 1-4:


 * 1) Doable in theory, but in practice not at all likely to use (the workarounds I've mentioned above are what we've seen more organizers using). I don't know any edit-a-thon organizers using AW+Draft space at events, and personally I've never found AW to be super useful. Though if Kudpung makes it much much much better, I suppose we'd be willing to give it a try w/ newbies at some point.
 * 2) Agree this could come in handy at times.
 * 3) Don't need this.
 * 4) Not a bad option but seems unlikely to get much use. It's simpler to focus on having advanced editors move stuff from sandbox to mainspace until newbies are autoconfirmed than it is to get a newbie to learn even more special wikibureaucracy in order to request a temporary override.
 * Cheers, Siko (talk) 21:41, 19 July 2017 (UTC)
 * This is a complex issue, but let me see if I can distill the consensus so far:
 * Use Article Wizard: Lukewarm support, but concern that the current Article Wizard has too many "walls of text" and needs to be improved.
 * Create new user right: Moderate support, but concern that it will take too long to be implemented before ACTRIAL.
 * Use a back-door: Unanimously opposed.
 * Use Requests for permissions: Not much support. Concern that admins won't be available/responsive enough.
 * Use user sandboxes: Suggested by RexxS and Siko. Seems like a viable option.
 * These options aren't mutually exclusive, so how about we move ahead with 1, 2, and 5? In other words, on the article creation landing page, we would give people 3 options, similar to Template:Newarticletext-unconfirmed: User sandbox, Article Wizard, or leave (and maybe a link to the Teahouse for people that want more help). At the same time, we could start an RfC about a new "event coordinator" user group, but move ahead with ACTRIAL in the meantime (i.e. we would not delay ACTRIAL due to the RfC). What do you think? Kaldari (talk) 00:47, 20 July 2017 (UTC)
 * I've started a new RfC for a simplified version of proposal #2 to see what the community thinks of that idea. We are also planning on moving ahead with proposals #1 and #5 on the landing page. Kaldari (talk) 21:34, 4 August 2017 (UTC)
 * I've withdrawn the previous RfC and created a new RfC based on the feedback from the first one. Kaldari (talk) 22:07, 18 August 2017 (UTC)

Potential costs to community health
We need to be very cognizant of and try somehow to measure how this affects community health - e.g. as page creation is one of the key steps to becoming an active editor we should watch how this barrier influences editor acquisition and retention. *I propose we randomize which non-autoconfirmed users are blocked from creating articles so that we have a control group.*

One thing I'd like to know is 6 years on whether any observable impact of the 2011 trial can be found. I expect it's pretty confounded and swamped by the larger trend of editor decline. --The Cunctator (talk) 13:32, 1 August 2017 (UTC)
 * The consensus is for limiting creation to autoconfirmed accounts. Even if there were a consensus to randomize which accounts could and couldn't create articles, I don't think the current means of implementing the trial have the technical means to make the distinction.
 * I've reverted your edits because I think the summary in the lede made sense. I have made some tweaks as well. The 2011 trial is what is being implemented. This is not a new idea and documenting the history of it for people who are new to the conversation is important.
 * TonyBallioni (talk) 14:00, 1 August 2017 (UTC)
 * , to answer your first question: "One thing I'd like to know is 6 years on whether any observable impact of the 2011 trial can be found." There wasn't a trial that was run in 2011 -- it was proposed and discussed, but the trial didn't actually happen, so there's no impact that we could analyze.
 * About measuring the impact on community health, there's a Research page on Meta where we're putting together the hypotheses and measures that we'll use to see the impact of the trial. It's still a work in progress, so it would be great to hear any thoughts about whether that list addresses the questions that you feel are important. -- DannyH (WMF) (talk) 14:42, 1 August 2017 (UTC)
 * DannyH (WMF) thanks for the helpful feedback! I was confused by how the project summary was written and tried to edit for clarity but TonyBallioni reverted, dismissing my concerns. I'm still a bit confused by the history. I've added some thoughts at meta. --The Cunctator (talk) 17:24, 1 August 2017 (UTC)


 * about the claim that "page creation is one of the key steps to becoming an active editor".  I didn't create my first article until i had many thousands of edits - but hey maybe i am a freak.  is there any data that actually supports that claim? Jytdog (talk) 01:32, 19 August 2017 (UTC)
 * I don't know where the stats are, but I think it's around 20% of new editors create a new article. Given, 80% of those get deleted or redirected, you can extrapolate that approximately 4% of new editors create an article that stays in main space past 90 days. TonyBallioni (talk) 01:38, 19 August 2017 (UTC)
 * If we accept the premise that Wikipedia has caused the unit of analysis to be "number of articles" and that new contributors think new articles are the thing to do in writing their preferred narrative, then yes, stopping editors from writing their new articles will stifle them. We chase these editors off, anyhow, by sending their creations to CSD. I think ACTRIAL will just cut to the chase and stop editors automatically before we have to manually. Besides, new editors will still write drafts, game the system into writing new articles, hijack redirects, and edit existing articles. Our community health is probably threatened a lot more by the toxic culture that causes good editors to quit in frustration. Not everyone is here to contribute, so those editors won't be lost. Chris Troutman  ( talk ) 01:57, 19 August 2017 (UTC)
 * Thanks Tony that goes somewhat to the data question I asked. I found that claim just startling.  Chris I struggle with the language of "stopping" or "shutting the door" - there is nothing unreasonable in having new editors go through AfC instead of creating directly.  Most sane organizations have orientation procedures for new people and I see nothing even a little outlandish in directing new people to AfC.   My sense is that truly new editors will find being walked through their first article to be as "natural" as taking drivers ed before you get a drivers license, or going through orientation on the first day of a job.  A lot will depend on how this is communicated.  There should be nothing like "gee sorry we are running a trial" but rather "Hey, welcome to WP.  Figuring this place out is hard, so for your first few articles we provide guidance and review before articles you create are published.  Welcome to new editor orientation!""   Totally positive and totally normal.  I want to say more about this elsewhere . Jytdog (talk) 02:08, 19 August 2017 (UTC)
 * User:Chris troutman about your edit note question here, i agree, and i spoke to that at mw here. Jytdog (talk) 03:07, 19 August 2017 (UTC)

Check list
September 4 is only 16 working days from now. Could someone make the checklist of things to be done?
 * An admin to make the GUI changes?
 * A template editor (if the changes require one)
 * See the list of tasks at the top of this talk page.
 * Another page that will need to be updated is User access levels (no changes needed) and there may be more, such as Special:ListGroupRights (under User: strike "Create pages (which are not discussion pages) (createpage)" -it's a Special Page a regular editor can't change. Kudpung กุดผึ้ง (talk) 11:12, 12 August 2017 (UTC)
 * Note: I've modified Kudpung's 4th point based on my read of the pages Legacypac (talk) 19:53, 21 August 2017 (UTC)


 * I'm centralising a checklist at Autoconfirmed article creation trial/Checklist, along with a countdown to September 9th September 7thEdited -- There'sNoTime (to explain) 11:10, 15 August 2017 (UTC)
 * some help centralising the list would be appreciated -- There'sNoTime (to explain) 11:22, 15 August 2017 (UTC)
 * , just saw this post: thanks for your work doing that. Quick question, above I saw September 7 as the start date, not September 9, was there another post I missed? TonyBallioni (talk) 20:48, 16 August 2017 (UTC)
 * Kudpung mentions the 4th, I said 9th and I set the timer as 7th... *confusion* -- There'sNoTime (to explain) 20:50, 16 August 2017 (UTC)
 * The 7th is what said above. Danny, can you confirm that we are still on track for a September 7th rollout? TonyBallioni (talk) 20:52, 16 August 2017 (UTC)
 * I'm pretty sure it's still the 7th, the only confusion is mine  -- There'sNoTime (to explain) 20:53, 16 August 2017 (UTC)
 * That's fine: I was actually coming here to ping Danny anyway because I was just explaining to someone how a non-AC account could create an article without going through AfC (we have users who think ACTRIAL has been old policy for a long time...), and gave the September 7 date as likely, but wanted to confirm it for the future. We also need to confirm the length it will be running: the consensus above seems pretty strongly in favour of 6 months, which was the original consensus. TonyBallioni (talk) 20:55, 16 August 2017 (UTC)
 * , the thing that would keep us from Sept 7th is how long the security review takes. The Security team looks at all the code changes that could create vulnerabilities that hackers could exploit. (Here's a description of our security review process.) The code for the new extension should be merged today, and then we can file a request for the security review. We expect Sept 7th will be plenty of time. I'll update you when I have some more information, and if I don't find out anything useful, I'll post on Friday anyway.
 * For the trial duration, I wrote above (on August 9th) about how things look from Community Tech's side. We've got funding for three months of Morten's time, which gives us enough time to collect one month's worth of data, and then do the analysis, report back to the community and have a discussion about the findings. Whether the trial is "switched on" or "switched off" after a month doesn't really affect the data analysis and discussions, so the trial can continue for however many months the community agrees on.
 * What I'm hoping for is that the findings after one month are compelling enough to have a good discussion with the entire community -- including our team, the New Page Patrollers, and everyone else in the community who's interested in quality control, new user experience and page creation workflows. If the results show unequivocally that restricting non-AC page creation is a huge improvement, then it'll probably stay on forever; if they show unequivocally that it's a huge disaster, then we may all agree to turn it off. If it turns out the experiment needs to run longer before we get any interesting findings, then CommTech will have to figure out how to support continued data collection and analysis. So the final duration of the trial could be anywhere from a couple months to forever, depending on what the data shows and what we all decide to do about it.
 * So for us, there's no reason to argue right now about one month vs six months; we'll figure that out when we've got findings. What I can tell you is that Ryan and I have no intentions of turning off the extension based solely on an arbitrary date of our choosing. In other words, we won't suddenly say "Okay, it's two months and that's long enough," without everyone talking about the findings and coming to consensus. -- DannyH (WMF) (talk) 22:21, 16 August 2017 (UTC)

I'll just reiterate that this is strictly a community project backed by technical and statistical support from the Foundation. The consensus which was for a 6 month trial, followed by a further one month without the trial cannot be changed. This was stressed and insisted upon during the debate. This also gives us one additional month without the trial again to see just what happens when the editing restriction is dropped. This provides a double fail safe which IMO the WMF, who are still strongly opposed in principle to this restriction, would appreciate. Kudpung กุดผึ้ง (talk) 03:42, 17 August 2017 (UTC)
 * See my answer above to Tony's request for clarification. -- DannyH (WMF) (talk) 16:46, 17 August 2017 (UTC)
 * I have,, but it didn't seem clear to me that the WMF has understood that this trial will run for a full six months, then be switched off for four weeks. Kudpung กุดผึ้ง (talk) 00:23, 18 August 2017 (UTC)

I said just above that I'd give an update on Friday. We've currently got two ArticleCreationWorkflow tickets that we're working on -- they're T172085 and T173605, if you feel like following along. When they're done, we'll ask for the security review. We're still on track for Sept 7th; I'll update when I've got more info. Let me know if you have any questions. -- DannyH (WMF) (talk) 01:20, 19 August 2017 (UTC)
 * Thanks. My fault - it's already halfway through Saturday morning here. Kudpung กุดผึ้ง (talk) 01:42, 19 August 2017 (UTC)
 * Here's another update -- the ArticleCreationWorkflow code has been submitted for the security review. We're going to try to encourage that process to go as quickly as it can, but the security team is overbooked and we don't have any direct control over it. We're shooting for the September 7th start date. I'll let you know when I've got more info. -- DannyH (WMF) (talk) 19:06, 21 August 2017 (UTC)
 * Who is the security team exactly?  The 7th is only 14 days away. I'm thinking of going to Ladbrokes.Kudpung กุดผึ้ง (talk)
 * You can see the Security team on the Staff and contractors page. They have to check out new extensions before they're deployed, to make sure there isn't a security risk that would create vulnerabilities that hackers could exploit. Here's a description of our security review process. You can see our request for the security review on this Phabricator ticket: T173619. -- DannyH (WMF) (talk) 15:07, 22 August 2017 (UTC)