Page MenuHomePhabricator

Change Growth features parameters on Portuguese Wikipedia
Closed, ResolvedPublic

Description

Since we have been very successful with the new platform, and we have managed to attract a good number of mentors, would it be possible to increase the wgGEHelpPanelHelpDeskTitle for ptwiki to 100%?

In addition, we recently decided to move some pages to the "Ajuda" (help) domain, including our help desk. I would like to request the adjustment of wgGEHelpPanelReadingModeNamespaces for ptwiki, from "Wikipédia:Tire suas dúvidas" to "Ajuda:Tire suas dúvidas".

Event Timeline

would it be possible to increase the wgGEHelpPanelHelpDeskTitle for ptwiki to 100%?

I'm assuming you mean wgGEHomepageNewAccountEnablePercentage, which is currently set to 80%? Setting it to 100% is possible, but having 20% of new users not get the homepage is deliberate, so as to create a control group for the experiment. I would like @MMiller_WMF and @nettrom_WMF to sign off on going to 100% (which would eliminate the control group).

I would like to request the adjustment of wgGEHelpPanelReadingModeNamespaces for ptwiki, from "Wikipédia:Tire suas dúvidas" to "Ajuda:Tire suas dúvidas".

This appears to refer to wgGEHelpPanelHelpDeskTitle. I can change that, but right now Ajuda:Tire suas dúvidas doesn't exist yet. When do you plan on moving this page? Right now, our software doesn't deal well with the helpdesk being a redirect, so we'd either have to fix that bug, or coordinate so that you move the page at the same time as I change the configuration setting.

@Trizek-WMF can help sort this out and explain how the Growth team is thinking about this.

Setting it to 100% is possible, but having 20% of new users not get the homepage is deliberate, so as to create a control group for the experiment.

Oh, OK. Sorry for the confusion. I had seen that euwiki was set to 100 and I had assumed it was sort of customizable. So, I don't see a problem with staying at 80%, but is there a plan of how long the tests will stay? After that, the number could be raise to 100%.

I can change that, but right now Ajuda:Tire suas dúvidas doesn't exist yet. When do you plan on moving this page?

In fact, it should have already been moved (along with Wikipedia:Informe um erro), but I didn't because I predicted that an error like the one mentioned in T265404 would occur.

Right now, our software doesn't deal well with the helpdesk being a redirect, so we'd either have to fix that bug, or coordinate so that you move the page at the same time as I change the configuration setting.

I planned to coordinate that, making the move at the same time when the change is deployed.

Setting it to 100% is possible, but having 20% of new users not get the homepage is deliberate, so as to create a control group for the experiment.

Oh, OK. Sorry for the confusion. I had seen that euwiki was set to 100 and I had assumed it was sort of customizable. So, I don't see a problem with staying at 80%, but is there a plan of how long the tests will stay? After that, the number could be raise to 100%.

I'll leave this to @Trizek-WMF to answer.

I can change that, but right now Ajuda:Tire suas dúvidas doesn't exist yet. When do you plan on moving this page?

In fact, it should have already been moved (along with Wikipedia:Informe um erro), but I didn't because I predicted that an error like the one mentioned in T265404 would occur.

Right now, our software doesn't deal well with the helpdesk being a redirect, so we'd either have to fix that bug, or coordinate so that you move the page at the same time as I change the configuration setting.

I planned to coordinate that, making the move at the same time when the change is deployed.

If you're around now-ish, there's a deployment window coming up in a few minutes where we can make this change. I've also PMed you on IRC to see if you're around. If not, no worries, we can schedule another time. There are deployment windows every weekday (except Fridays) at 18:00 UTC and 23:00 UTC.

Change 633843 had a related patch set uploaded (by Catrope; owner: Catrope):
[operations/mediawiki-config@master] Rename GrowthExperiments helpdesk on ptwiki

https://gerrit.wikimedia.org/r/633843

Change 633843 merged by jenkins-bot:
[operations/mediawiki-config@master] Rename GrowthExperiments helpdesk on ptwiki

https://gerrit.wikimedia.org/r/633843

Mentioned in SAL (#wikimedia-operations) [2020-10-13T23:18:39Z] <catrope@deploy1001> Synchronized wmf-config/InitialiseSettings.php: Rename GrowthExperiments help desk on ptwiki (T265214) (duration: 01m 04s)

I ended up reaching @Albertoleoncio on IRC, and we successfully moved the help desk together.

@Albertoleoncio, I need a clarification about why you want to move to 100%.

At the moment, the features are available for 80% of new accounts. As Catrope explained in a previous message, we keep 20% of new accounts without the tools so that we can check if the tools have any bad impact compared to our control group (the 20%).

Basque Wikipedia is the only case with a 100% activation. We did it because they are a very small wiki and people create accounts on this wiki only because they attend editing workshops. We agreed to deploy the features to 100% of newcomers so that newcomers don't have to explore their preferences during workshop time.

At the moment, since our tools are really changing (they are prototypes), we need to keep a control group for a wiki as big as Portuguese.

We haven't explored yet when we would be able to switch from 80% to 100%, but your input is definitely welcomed there so that we can think about it and take a decision at some point (this decision requires time and various inputs though). Is it something your community discussed about? If so, can you share a link to the conversation? Thanks! :)

@Albertoleoncio, I need a clarification about why you want to move to 100%.

Shortly: Because the new mentoring system miraculously solved an old problem we had :D

Long version: I don't know if you have noticed, but we already had a mentoring system before the implementation of the Growth Experiments. Although it was well structured, it was very inefficient due to the lack of mentors interested in adopting new users. Even though this problem had already been discussed several times on our village pump (1, 2, 3, 4, 5), things remained the same. To illustrate, this is the list of new users who were interested in being mentored, where they were never chosen by mentors. Allied to this, there was complaints that the system was underused by mentoree, because users were often chosen by mentors and simply disappeared after that. I was a little concerned about a "stampede of mentors" after the implementation of the new mentoring system, by making this process automatic, but instead, our mentoring queue are gone and the feedback from the community has been great so far.

At the moment, the features are available for 80% of new accounts. As Catrope explained in a previous message, we keep 20% of new accounts without the tools so that we can check if the tools have any bad impact compared to our control group (the 20%).

Basque Wikipedia is the only case with a 100% activation. We did it because they are a very small wiki and people create accounts on this wiki only because they attend editing workshops. We agreed to deploy the features to 100% of newcomers so that newcomers don't have to explore their preferences during workshop time.

At the moment, since our tools are really changing (they are prototypes), we need to keep a control group for a wiki as big as Portuguese.

As I had told Catrope, I have not understood that was a control group behind those missing 20%. I have assumed it was a security limit or something like that, so I started this ticket by asking "would it be possible"...

We haven't explored yet when we would be able to switch from 80% to 100%, but your input is definitely welcomed there so that we can think about it and take a decision at some point (this decision requires time and various inputs though). Is it something your community discussed about? If so, can you share a link to the conversation? Thanks! :)

I believe that only I am aware that the portal is activated for only 80% of new users. As it was more technical and I wasn't sure why, I ended up not sharing it. Since this is an experiment, I think we can work on it together and share these results with the community after a period (a few months, idk...) and better discuss the plausibility of increasing it to 100%.
So far, there has been no broad discussion related to growth experiments, or at least not formally. If you need inputs, I would recommend contacting Jo directly, as he is the most active user dealing with new users and the most active in answering questions in the HelpDesk.

@Albertoleoncio -- thanks for working with us on this feature and discussing it. @Trizek-WMF may have other thoughts and responses for you, but I have a question: you said "the feedback from the community has been great so far" about the mentoring feature. What are they saying? We are interested to know what is working well so that we can think about how to augment those good parts.

@Albertoleoncio, thank you for the details.

I very familiar with your mentoring issue! :) As you may know, I'm a volunteer on French Wikipedia. A few years ago, we made the same observation as yours. So we decided to have a more attractive message that starts with "hello, I'm [Mentor], and I can help you" and a direct link to the mentor's talk page. It led to more questions to mentors and the fall of the old, less efficient mentoring program -- which was built the same way as yours. So I'm really happy to know that assigning a mentor to each newcomer solves the old issue you had!

Concerning the 100% activation, the team has to discuss about it in details. If this topic of the 80%/100% raises on the community, would you please inform us about it?

If you need inputs, I would recommend contacting Jo directly, as he is the most active user dealing with new users and the most active in answering questions in the HelpDesk.

Thank you for the tip. I can't find if this user understands English, any idea?

What are they saying?

Actually, not a lot. I have heard comments from this topic in our village pump and informally, from our Telegram group. I think we can open a new topic, like a RfC, to see if our mentors have more opinions after those ~3 weeks. If you have some questions in mind, I can translate it and publish there.

@Albertoleoncio, thank you for the details.

If this topic of the 80%/100% raises on the community, would you please inform us about it?

Sure!

If you need inputs, I would recommend contacting Jo directly, as he is the most active user dealing with new users and the most active in answering questions in the HelpDesk.

Thank you for the tip. I can't find if this user understands English, any idea?

I think he does. I can step in, if necessary.

@Albertoleoncio -- thanks for working with us on this feature and discussing it. @Trizek-WMF may have other thoughts and responses for you, but I have a question: you said "the feedback from the community has been great so far" about the mentoring feature. What are they saying? We are interested to know what is working well so that we can think about how to augment those good parts.

Initially it looked great (technically it looks great), but after some weeks of use I've found that almost 100% of the questions received via the new system do not seem to have any kind of development after answered, and of those a considerable amount are either pure crap or stuff which are obviously a dead end (ppl trying to create their own biographies there, ideology motivated editions, etc). I am generally extremely tolerant with newbies, but this is approaching the point where I'll remove the name from the tutor list, since I fell it's basically a waste of time.

@Darwinius, I'm sorry about this bad experience. What kind of messages do you receive? Can you share some links so we can see some examples?

We had the same kind of feedback coming from French Wikipedia. I'm a volunteer mentor there and, like you, we receive questions about creating autobiographies and promotional articles. As a consequence, some mentors have complained about the low quality of questions. It was important to check if it was a fact or a feeling.

We took all questions over 3 days (57 questions) and checked both if:

  1. They were actual questions
  2. The context of the message was providing enough information so that the message could be treated (even if it was not a question).

We were two evaluators of the messages, me and one of the users who was raising up this issue.
The outcome is the following:

  1. 9 messages over 57 are not questions (15,7%)
  2. 9 messages over 57 were not giving information to be replied (15,7%)

9 and 9 doesn't mean we have 9 bad questions. Some messages are questions but cannot be answered ("How are you?"), and, on the other hand, some messages, without being a question, can be answered ("Hi, my goal was to have the article being visible on Google.").

These results were fine for the French community, since the percentages were reasonable, if evenly distributed among mentors.

If you like, we can run the same study for Portuguese Wikipedia. Based on the results, we can check which kind of action we could take to increase the quality of questions asked to mentors. What do you think?

Yes, the questions definitively should be analyzed, because despite the system being cool technically, there's some kind of flaw that is rendering it quite below an optimal state.

I've keep all the questions I received in my talk page, and can do a quick analysis of them:

https://pt.wikipedia.org/wiki/Usu%C3%A1rio_Discuss%C3%A3o:DarwIn#Pergunta_de_Willy_Fraze_(12h34min_de_5_de_outubro_de_2020) (down from there)

Total of 20 questions coming from the new system

  1. Question OK, no feedback (but maybe he found the solution himself)
  2. Garbage
  3. Question OK, no feedback, only edition by user
  4. Question OK, no feedback, only edition by user
  5. Just "create biography", no feedback, only edition by user
  6. Question OK, no feedback
  7. Question OK, no feedback, only edition by user
  8. Question OK, had feedback (9), but stopped editions there.
  9. Feedback from 8 (it is kind of weird that they have to create a different topic for that)
  10. Question OK, no feedback, but apparently solved the problem and he edited afterwards
  11. Weird (garbage) question, no feedback, only edition by user
  12. Question OK, no feedback, only edition by user
  13. Wants to create own biography, no feedback, only edition by user
  14. Question OK, no feedback
  15. Garbage
  16. Just "create page", no feedback
  17. Question OK by user (16), no feedback, together with (16) only editions by user
  18. Question OK, no feedback, but apparently solved the problem and he edited afterwards
  19. Question OK, no feedback, but apparently solved the problem and he edited afterwards
  20. Question OK, no feedback, only edition by user

Only 1 in 20 provided some kind of feedback, but didn't edit afterwards
Only 3 in 20 appeared to be useful, but no way to be sure since no feedback.

I generally dedicate my time as sysop to monitor our filters, where many newbies are caught. There I provide help and assistance that I feel and know are very useful for newbies (such as granting confirmed status to escape the filters, helping them to save their articles in an educative way, teaching about mistakes, etc), and I'm doing this for more than 4 years, so I consider myself to be reasonably seasoned on newbie assistance. The feeling I have now is that shifting that time from the filters to this new tutor system is a bad option for me, so I'm considering abandoning it. But I'm probably a special case (very few sysops actively monitor the filters), and others may find this new system useful.

A possible problem I've detected is that, since we have not banned IPs from editing user talk pages (yet), persistent IP vandalism is causing a number of them to be protected against non confirmed users. I suspect this knocks down the new tutor system, as it blocks newbies from communicating with their tutor.

Thank you for sharing these examples.

Let me filtering these questions using same questions as on fr.wp, since we are assessing on the quality of questions:

#UserIs it an actual questionHas context to write a reply
1Will Bungoyesno
2Willy Frazeyesyes
3waji gangnono
4XirleyGruyesyes
5Marcelle Marques de Andradeyesyes
6Gustavo Silva Xaviernoyes
7Layanne Maurícioyesyes
8intellectual Gochiyesyes
9Turackyesyes
10Turacknoyes
11Vitydyesyes
12Son of wisdomyesyes
13Socimbrayesyes
14GSMBAHIAyesyes
15Manuel Guedes Martinsyesyes
16Eu s3ó pr0nono
17Ddgf8611noyes
18Ddgf8611yesyes
19KIQAFyesyes
20César Argumedosyesyes
21Elltreezyyesyes
Total os "no"s5/213/21
Total os "no"s, %23,8%14,2%

The quality score is quite similar to French Wikipedia (15,7%), where the sample was bigger. And I only see two questions that have a "no" on each column, which is good: the majority of messages yo receive can have replies.

A lot of questions that are not directly about editing are around knowing more about Wikipedia. Wikipedia's workflows are basically unknown for global audience. I don't know about your language, but the French press only has a few articles about Wikipedia. French global audience would know more about Facebook or Twitter just by watching TV or reading newspapers. And we, as communities are overall not good at documenting how Wikipedia works. On all wikis I visit, I see that documentation or first steps are spanning from "confusing" at best to nothing in the worse case scenario". Hence this idea of having a mentor to guide you when you start.

You also have people who just say hi, because you welcomed them; just some basic (but pleasant) social interactions. Is Wikipedia a social experience? I won't reply to this tricky question, but I know that it is more pleasant to salute others. ;-)

Only 1 in 20 provided some kind of feedback, but didn't edit afterwards
Only 3 in 20 appeared to be useful, but no way to be sure since no feedback.

People not responding immediately is very frequent. They might come back after a while. For instance, I just got a thank as a volunteer for a reply to a question I posted 3 months ago. It is sometimes a long run, and newcomers (and most users) aren't wiki-nerds like we can sometimes be. :) Most newcomers have a dedicated time for Wikipedia. They might come back later if they liked the previous experience (knowing that someone can help you helps). On the other hand, sometimes problems are solved by users themselves without waiting for a reply, which is also fine. They will know that you are here to help them, when they will be back.

Also, people creating a new section to reply are common, just because our wikis aren't really designed in a way people are used to. Usually, to post a reply, there is a field that clearly says "type your reply here". Not on Wikipedia. Maybe newcomers don't know how to reply? The new reply tool, available on your wiki as a Beta feature, would help a lot to solve this issue!

The feeling I have now is that shifting that time from the filters to this new tutor system is a bad option for me, so I'm considering abandoning it. But I'm probably a special case (very few sysops actively monitor the filters), and others may find this new system useful.

I totally understand!

From one mentor to another mentor, it is sometimes a bit discouraging to reply and see no action. And then, after a while, you have some great collaborations that start. It may take time, but, in my opinion, it worth it on the long run. Easing newcomers' life is a key point there. I personally use StructuredDiscussions, that has some limitations but allows newcomers to have their own conversation instead of struggling at understanding a classical Wikipedia talk page.

Helping newcomers who are blocked while editing is also a way to help. I think the most important thing is to help them, whatever the way. :)

A possible problem I've detected is that, since we have not banned IPs from editing user talk pages (yet), persistent IP vandalism is causing a number of them to be protected against non confirmed users. I suspect this knocks down the new tutor system, as it blocks newbies from communicating with their tutor.

Thank you for this information. The "ask your mentor" module is just a form to help newcomers posting, like the "add topic/adicionar tópico" form you have on a talk page. If a newcomer can't post on their mentors' page, it is a problem. However, I randomly checked about 10 mentors' pages and none of them were protected. Do you know how many of them are protected?

Thank you for this information. The "ask your mentor" module is just a form to help newcomers posting, like the "add topic/adicionar tópico" form you have on a talk page. If a newcomer can't post on their mentors' page, it is a problem. However, I randomly checked about 10 mentors' pages and none of them were protected. Do you know how many of them are protected?

Mine was protected (and it was supposed to be still protected, but I removed the protection myself), since I was receiving death threats from an IP there. Other sysops which are active against vandalism are also subject to attacks and vandalism, so it is relatively common to have user talk pages protected, sometimes permanently or for a long time. I don't know if any of the current tutors is in this situation, but this is something frequent and recurrent, at least until we revoke the ability for IPs to edit talk pages.

My sympathies to you since I already had the same unpleasant experience.

We have some ideas around mentors' protected pages:

My sympathies to you since I already had the same unpleasant experience.

One gets used to that after a while 😛

We have some ideas around mentors' protected pages:

I believe this last one would solve the problem.

From one mentor to another mentor, it is sometimes a bit discouraging to reply and see no action. And then, after a while, you have some great collaborations that start. It may take time, but, in my opinion, it worth it on the long run. Easing newcomers' life is a key point there. I personally use StructuredDiscussions, that has some limitations but allows newcomers to have their own conversation instead of struggling at understanding a classical Wikipedia talk page.

What do you mean with StructuredDiscussions "allows newcomers to have their own conversation"? If I well understood, it only affects who activates it on the beta functionalities, and nobody else, so it would have no effect on newbies. Is this incorrect?

StructuredDiscussions (SD) are a different way to have a talk page. Each new message is a new topic, a proper page. This way, instead of having to find back their old message on a talk page, users have a direct access (and can watch) a dedicated subpage. On wikis where StructuredDiscussions are available (like on French Wikipedia), the talk page owner can enable them on their Beta preferences. Then anyone visiting the talk page has to use it. See my volunteer talk page on French Wikipedia you may see there some conversations I have with newcomers. Please note that SD will be replaced by the Reply tool.

The Reply tool is also available on some wikis. It is a different tool. This one only affects the user who enables it on their Beta preferences, like you describe. There are some plans to have it deployed to every new accounts though. If you think it would help your wiki, you should contact the team in charge of this feature.

The Reply tool is also available on some wikis. It is a different tool. This one only affects the user who enables it on their Beta preferences, like you describe. There are some plans to have it deployed to every new accounts though. If you think it would help your wiki, you should contact the team in charge of this feature.

Apparently what we have available as beta at wiki.pt is the reply tool, not SD. I've activated it now, and will give feedback as you suggested.

The reply tool will only work for you, not for people who are reply on on your talk page.

What was initially requested has already been done, so...