Wikipedia talk:Snuggle

This talk page is being considered for deletion in accordance with Wikipedia's deletion policy.

Please discuss the matter at this page's entry on the Miscellany for deletion page.

You are welcome to edit this page, but please do not blank, merge, or move it, or remove this notice, while the discussion is in progress. For more information, see the Guide to deletion.

Start Snuggle

Other projects?

Is it possible to run Snuggle in, say, http://ca.wikipedia.org ?--Qgil (talk) 18:12, 17 June 2013 (UTC)[reply]

Good question! I'd love to. I'm currently gearing up to run an instance of Snuggle in http://pt.wikipedia.org. In order to get started, I need someone to act as a liaison to help me understand the community's needs and bug reports. Would you be willing to fill that role or help me find someone who will? --EpochFail^{(talk • work)} 19:10, 24 June 2013 (UTC)[reply]

I have asked at the Viquipèdia in Catalan. I have also volunteered as liasion if there is an interest. — Preceding unsigned comment added by QuimGil (talk • contribs) 18:57, 25 June 2013 (UTC)[reply]

Can we test snuggle also in german language version? Thank you, Conny (talk) 18:50, 25 June 2014 (UTC).[reply]

Hi Conny Totally. I added a bug to my tracker [1]. Is it OK if I bug you to help me fill out some language files? See [2] for an example of the English language file. If you can replace everything to the right of the colon ":" with equivalent German (or your best guess) and send me an updated file, that will help me get started. We'll need to iterate a couple of times, so don't sweat any mistakes. Once we see how the language bits show up in the UI, we can fix problems. --EpochFail (talk • contribs) 19:02, 25 June 2014 (UTC)[reply]

Sounds very interesting but...

... is it only for official mentors and coaches? Or can other people also use it, for example after application? The Banner talk 20:36, 6 July 2013 (UTC)[reply]

Its currently open for everyone to use, User:The Banner. Every action on Snuggle is visible to everyone else currently, so no worries about any bad faith people. Using Snuggle, you currently have the option of categorising newer editors, reporting them to AIV or to send them a teahouse message. You are also free to try direct Wiki contact with the editors. The categorising, in turn, helps Snuggle understand the difference between good and baith faith editors and will ease finding them. A simple sort by desirablity helps by making it much easier to find new editors of the same type

Hope this helps! Feel free to try it and leave further feedback or ask more questions! :)

Cheers,

TheOriginalSoni (talk) 20:55, 6 July 2013 (UTC)[reply]

I have looked a number of times to Snuggle, but it is not the field that interests me most. So I am moving out here. The Banner talk 20:00, 26 August 2013 (UTC)[reply]

Very interesting

First off, I am sorry to hear about your advisor. I understand the closeness of the student-advisor relationship and this must be a difficult time for you.

Secondly, I like the mockup that you posted. I believe that the tools other than rating should be more prominently displayed, and your proposal should do just that. Andrew³²⁷ 16:54, 11 July 2013 (UTC)[reply]

Thanks. I appreciate it. And thanks for the feedback and your work with Snuggle. --EpochFail^{(talk • work)} 18:52, 11 July 2013 (UTC)[reply]

Wiki-mentors, I made you a tool. IRC discussion Wed. July 17th @ 1600 UTC

Snuggle users and the Teahouse are co-hosting an IRC office hours session (Wed. July 17th @ 1600 UTC - #wikimedia-office ^connect) to discuss the state of new editor support in Wikipedia and introduce you to WP:Snuggle, a web-based tool designed to make finding good-faith newcomers who need help fast and easy. Give it a try by pointing your browser to http://snuggle.grouplens.org.

See the agenda for more info. --EpochFail (talk), Technical 13 (talk), TheOriginalSoni (talk) 17:06, 12 July 2013 (UTC)[reply]

Mentorship brainstorming

For folks interesting in brainstorming current issues with mentorship systems and what we might try to do differently, see also this idea. Come join the discussion! Siko (WMF) (talk) 17:11, 17 July 2013 (UTC)[reply]

Ensuring that mentors are suitably experienced

I wanted to follow up on a discussion that began during the Snuggle office hour session. One issue that I see a lot lately is that we have helpers and mentors in the various help spaces (Teahouse, Adoption program, #wikipedia-en-help, AfC review, etc.) who are not sufficiently experienced. I've seen users who are finishing up their own adoption programs jump right into trying to adopt new users, and I frequently see users with very little experience with the project come into #wikipedia-en-help to try to help new users. We don't really have any way of preventing this from happening, and we don't have any system of choosing who has enough experience to mentor.

I think it's very important that people who are helping our new users are experienced enough to answer most questions, and self-aware enough that they know when to say "I don't know" and pass the user on to someone who can help. How should we make sure this is the case? I do not like the idea of new userrights, and another dramatic RfA-like process is not something I seek to begin, but we need something: a set of criteria? A simple area where other experienced users can vouch for another user's experience? What are your thoughts? – GorillaWarfare (talk) 17:13, 17 July 2013 (UTC)[reply]

Some other areas onwiki use the "vouch for each other" style, I think. For instance, someone who wants to be an SPI clerk posts a note explaining who they are and why they want to help, and then usually a clerk will take them on as a trainee clerk if it's felt they're qualified. After a period of training, the checkusers and clerks will get together and determine if they should be promoted to full clerk. No formal vote or assignment of user rights, but neither does Joe Schmoe get to wander in and start doing the work without knowing how to do it - your peers have to take a look at you, approve you, and train you before you're set loose. This seems like a reasonable setup to use for mentors/adopters, as well: start out with some self-selected, known-to-be-qualified mentors, and have them (and anyone else in the community who wants to weigh in) be the gatekeepers and trainers or new mentors. If no current mentor is willing or able to take on someone who wants to be a trainee, then the trainee will have to wait (or, perhaps, apply to [other place], which could also appoint them? or something?). A fluffernutter is a sandwich! (talk) 17:31, 17 July 2013 (UTC)[reply]

One problem is the wide range of subjects covered by the parts of Wikipedia listed above. A SPA graphic designer might be able to answer questions about images better than an admin but wouldn't know anything about other policies. That is especially true in places like the TeaHouse, where there are enough eyes that truly bad answers are usually corrected. I support User:Fluffernutter's proposal for Adoption and AFC. Andrew³²⁷ 17:57, 17 July 2013 (UTC)[reply]

Agreed, but a big difference between an experienced mentor or helper and an unexperienced one is that the experienced ones tend to know when they don't know the answer, and also tend to know where to send people for more expert advice. – GorillaWarfare (talk) 18:19, 17 July 2013 (UTC)[reply]

Yes, indeed. One of the key aspects of a new mentorship program will be a clear path "Mentee to Mentor" which hopefully will insure that new mentors meet the minimum level of proficiency. slv (talk) 21:54, 17 July 2013 (UTC)[reply]

Fluffernutter's proposal sounds reasonable and useful to me. Such a setup would be immensely helpful in training potential mentors. Also pinging User:Theonesean to see if he'd be interested in something of this sort. TheOriginalSoni (talk) 07:10, 19 July 2013 (UTC)[reply]

Fluffernutter, that sounds like it could work. However, is it possible to break the mentors into sections based on what they need to know, and have a group of self-selected, known-to-be-qualified users be the ones to train and approve the new clerks, like the MedCom application process, but with training. And also, re: slv, each section (or the group of people at large) should develop a trainee "school" like so many mentors have done, that provides an outline of what they should no, and allows the mentors to test the proficiency of each recruit and tailor the training to them. the one sean 14:56, 19 July 2013 (UTC)[reply]

I'm not sure how I feel about the concept of training mentors, honestly. Mentorship, in my opinion, is a pretty informal process where an experienced Wikipedian takes on an unexperienced Wikipedian for mentorship, and explains questions and teaches about policies. I feel that if a new mentor needs training of this kind, perhaps they're not as experienced with the project as they should be. – GorillaWarfare (talk) 15:31, 19 July 2013 (UTC)[reply]

Maybe it shouldn't be a full hands-on training (i.e. not mentoring the mentors), but I think there should be an easy, centralized way to observe the progress and methods that mentors use, particularly newer ones. This is incredibly easy at The Teahouse, since there is only one area, the Q&A section, where all of us are watching and can observe other hosts' responses. Occasionally, when a response is incomplete, we can fill in the blanks or provide another perspective. We can also very easily see when a response is inappropriate (like in tone), or ensure that hosts are sending talkbacks to new users who don't understand or use their watchlist, and can talk to the host directly. We might consider making a centralized area where it is easy to access the user pages or other areas dedicated to mentoring-in-progress so that mentoring is a more visible and monitor-able event (among other mentors) rather than behind-the-scenes. I, Jethrobot ^{drop me a line} (note: not a bot!) 16:11, 19 July 2013 (UTC)[reply]

(For the record, I also support Fluffernutter's proposal.) I, Jethrobot ^{drop me a line} (note: not a bot!) 16:12, 19 July 2013 (UTC)[reply]

Yes, I have no issue with that kind of review. It's the idea of "mentoring the mentors" that I dislike. – GorillaWarfare (talk) 17:16, 19 July 2013 (UTC)[reply]

Hi TheOriginalsoni, thank you for bringing Theonesean into this conversation! Theonesean, I Jethrobot, these are very good points. A more centralized process will absolutely help with tracking progress of individual and projects, as well as enabling the sharing of best practices. I really like the idea of a “trainee school” a space where that can happen. To theonesean’s point, having some “specialty” (criteria set in preferences) is also useful for both mentors and mentees. We know newbies come in with different interests and contribution capacity – the one-fits-all format is not optimal. It takes time and commitment to reach a Superstart-Contributor status and we’re loosing lots of mentees that could do more micro-tasks-type work in the process. So maybe some tailoring will help with retention. Same goes for Mentors, there is variance in mentors preferences and capacity. The idea behind the Mentor-in-Training is multifold, to create a clear path from x-to-mentor while addressing concerns regarding mentors levels of experience/expertise, as well as the need for some common criteria that other experienced users can vouch for (validate). But most importantly, it’s also designed to increase the pool of well-prepared Mentors so we can respond to the demands of new generations of newbies. Maybe a better understanding of GorillaWarfare's 'dislike' will help us address concerns. :) Cheers! slv (talk) 22:00, 24 July 2013 (UTC)[reply]

I'm flattered!

And yeah, one-size-fits-all is doomed to failure. There was a really good TEDtalk [3] by Malcolm Gladwell about, among other things, there is no "perfect solution" to solving a problem. There are only perfect solutions. A specialisation, allowing mentors and mentees to focus on different areas of Wikiwork are perfect solutions. Specialisation, and allowing people to find what they like best, is the best way to do this. the one sean 04:28, 29 July 2013 (UTC)[reply]

Hi Theoneseas, thanks for sharing, love his books! I hadn't seen the talk, it illustrates well the opportunity for tailored/personalized user experience. :) slv (talk) 19:05, 29 July 2013 (UTC)[reply]

Solutions to the the snuggler-newcomer race condition

A potential problem was brought up in the last IRC session about Snuggle. If two Snuggle users are interacting with the same user at the same time, they might duplicate teahouse invites and other messages. There's a few potential solutions to this problem:

Warning for duplicate actions: Warn the snuggle user if they are trying to perform an action that was previously performed for a user.
Synchronous visibility: Make the concurrent activities of snuggle users visible to each other. This would manifest as an activity feed or other realtime visualization that would show snuggle users when they are stepping on each other's toes.
Local refresh: Refresh a the newcomers information when their row is selected in the user browser of Snuggle.

The easiest of these to implement is local refresh. Next is warning for duplicate actions. Synchronous visibility would require substantial restructuring to the code base and some experimental work with web sockets, but this restructuring would allow us to consider other types of synchronous activities such as a chat room and a live-updating "recent activity" log. Thoughts? --EpochFail^{(talk • work)} 19:05, 24 July 2013 (UTC)[reply]

To me the local refresh option seems the best of the three, and would be even better if supplemented with warning for duplicate actions but I think would be fine on its own and cut back on the majority of duplicate interactions. Jamesmcmahon0 (talk) 19:12, 24 July 2013 (UTC)[reply]

I completely agree with James. Andrew³²⁷ 19:33, 24 July 2013 (UTC)[reply]

I'd prefer synchronous visibility but if that's not feasible, local refresh supplemented with warning for duplicate ought to work. Chris Troutman (talk) 20:45, 24 July 2013 (UTC)[reply]

I'm with James on this one; if warning for dups can be implemented, that would be fantastic, but isn't strictly necessary to start. I, Jethrobot ^{drop me a line} (note: not a bot!) 17:12, 25 July 2013 (UTC)[reply]

Thanks for the responses, folks. I've added both the Local refresh and Duplicate action warning to the bug tracker. Once I've implemented the most recent mockups, solving this problem will be next on my plate. --EpochFail (talk • contribs) 21:32, 26 July 2013 (UTC)[reply]

RfC: Are the Category:Wikipedians and its subcategories appropriate for Wikipedia

There is an ongoing RfC going on at Category talk:Wikipedians#RfC: Is this category and current subcategories appropriate for Wikipedia that readers of this Wikipedian software page may be interested it. Technical 13 (talk) 12:19, 29 August 2013 (UTC)[reply]

First impressions

I just tried out Snuggle for a few minutes. Here are my first impressions. First the good:

The concept looks interesting - I like the idea with dealing with new users by user instead of by page or by action.
The interface is very good for getting an overview of new user's contributions. Good work on that one.

Things that could be improved:

I didn't really get a sense of why we were classifying users as good-faith and bad-faith. As in, I wasn't really sure what effect it would have on-wiki. This is probably the single thing that will most motivate people to use the software, so it would be worth really spelling it out.
Slightly related to the first thing - I was looking for a "welcome this user" link, but couldn't find one. The ability to add a one-click welcome would see a lot more welcomes being added, I would guess.
I was looking for links to the user's talk page next to the summary of their talk page posts, but there don't appear to be any. You could either have a link to the whole talk page right above the talk page summary, or links to the individual sections to the side of, or overlaid on, the edit summaries.
For users with many edits per day, the interactive edit graph can become very cramped, with an individual edit taking up just one or two pixels vertically. This makes it very hard to check individual edits - you might intend to click on a new edit, but accidentally click on one that you already checked, etc. Maybe you could implement some kind of scrolling or zoom feature to make this easier to navigate?

Other things I thought of:

What happens if a user blanks their talk page? Does Snuggle treat it as though they had never been left any messages?

I didn't spend very much time at all in doing this, so I may have missed some (most?) of Snuggle's features, but I hope this proves useful to the devs. Best — Mr. Stradivarius ^{♪ talk ♪} 01:08, 1 November 2013 (UTC)[reply]

Hi Mr. Stradivarius. Thanks for writing up your feedback. Your concerns/confusion around classification of newcomers is somewhat common. The classification UI is intended to be used only by other Snuggle users. It doesn't actually have an effect on the wiki. One common usage pattern that I Jethrobot showed me is as a means to clear an unclassified user from the list after interacting with them.

Regardless it seems that Snuggle "works" but is due for a redesign to make some of the interactions with the system more apparent. Would you be willing to spend some more time with it to help me and TheOriginalSoni with a redesign? --EpochFail (talk • contribs) 13:45, 1 November 2013 (UTC)[reply]

Woops. I nearly forgot to answer your other question. If a user blanks their talk page, it will look like they have no messages. The talk page space in Snuggle is a reflection of the current status of their talk page on wiki. I agree that this should be historical. I haven't had the time to spec that one out yet, but I could try a few sketches of what that might look like. I'll add something to the WP:Snuggle/Work log when I get a chance. Make sure to add that to your watchlist if you want to follow our progress. --EpochFail (talk • contribs) 13:48, 1 November 2013 (UTC)[reply]

Bug: Wrong reverter identified

I have just been using Snuggle and come across these edits. Snuggle said they were reverted by ClueBot NG, when they were actually reverted by AussieLegend.

I was going to report it as a false positive by CBNG, since the edits weren't vandalism. But now I have seen AussieLegend's explanation this is not necessary.

Yaris678 (talk) 12:57, 11 February 2014 (UTC)[reply]

Thanks for reporting Yaris678. I'm looking into it. --EpochFail (talk • contribs) 20:27, 12 February 2014 (UTC)[reply]

Note to self: supposed reverting edit [4] --EpochFail (talk • contribs) 21:46, 12 February 2014 (UTC)[reply]

Bug fixed. See [5] I still need to write an update script to fix the current data, but by pushing this fix out to production, new editors will be tracked correctly. --EpochFail (talk • contribs) 16:11, 13 February 2014 (UTC)[reply]

Thanks. Yaris678 (talk) 11:06, 14 February 2014 (UTC)[reply]

Bug: Wrong number of reverts

I decided to sort the newbies by number of reverts. The people at the top of the list all had several hundred reverts, despite having only ten or so edits.

Yaris678 (talk) 19:24, 12 February 2014 (UTC)[reply]

Thanks! Can you provide a screenshot of an offending user dossier to help me investigate? --EpochFail (talk • contribs) 20:24, 12 February 2014 (UTC)[reply]

Cool. Just added a screenshot to the thread. One newbie is expanded but the others also have a high number of reverts. Yaris678 (talk) 20:51, 12 February 2014 (UTC)[reply]

Thanks for the quick response. I filed the bug in my tracker [6] and will look at it as soon as I can. --EpochFail (talk • contribs) 21:24, 12 February 2014 (UTC)[reply]

Thanks.

BTW. I don't know if this is relevant to either bug, but the WMF changed the formatting of diffs in the API recently and it messed things up for STiki. Andrew will know the details. He managed to change STiki so that it dealt with the new format. I thought I should mention it, just in case it was the cause.

Yaris678 (talk) 22:00, 12 February 2014 (UTC)[reply]

Broken today?

So I tried to finally test the tool today, but the feed for users that I could review was empty. I got a brief flash of "Starting Snuggle..." or such, and that's all: nothing on the green, yellow or red tabs... --_{Piotr Konieczny aka Prokonsul Piotrus| reply here} 16:10, 22 August 2014 (UTC)[reply]

Piotrus, I'm not having trouble. Can you try again? If it still doesn't work, can you let me know what browser you are using? --EpochFail (talk • contribs) 16:28, 22 August 2014 (UTC)[reply]

Tried again today; ~~both Firefox and Chrome are stuck at "Loading...".~~ hmmm, it seems it takes a minute or so for that screen to load. Or could be a temporary server delay; I tried IE and Snuggle loaded - then I noticed it loaded in all three browsers (but F-fox wasn't loading it for about a minute...). Perhaps could add a progress bar or such, but low priority probably. --_{Piotr Konieczny aka Prokonsul Piotrus| reply here} 12:15, 24 August 2014 (UTC)[reply]

Thanks for giving it another try! Regretfully, this is a known issue that I haven't be able to get around. It seems that mongoDB misbehaves drops the index needed to sort users from memory. You should find that, after the system first loads, everything will be fast. I'll likely be switching to redis for the major version. --EpochFail (talk • contribs) 12:23, 24 August 2014 (UTC)[reply]

Is it live?

Hi, this was mentioned on the WM-l recently. How do I use it? Is there a manual? Tony (talk) 05:18, 27 August 2014 (UTC)[reply]

Hi Tony,

It is live. If you click on the big link at the top of the page it should work... although it may take a bit of time to load up after logging in.

The basic documentation is at WP:Snuggle. There is also a link to a video on that page that gives a tutorial. (The link used to go to a different page where the video was several clicks in, but I've just changed it)

Yaris678 (talk) 20:54, 27 August 2014 (UTC)[reply]

Actually, having said that, I can't get the summaries of users to expand. Is anyone else having that problem? I'm using Internet Explorer 11 on a Microsoft Surface with Windows AT. Yaris678 (talk) 20:59, 27 August 2014 (UTC)[reply]

Update: Snuggle is working on a different machine (Firefox and Windows 7). I suspect the problem was with IE11 or something to do with the Surface being touch-screen causing some kind of confusion. So Tony, it should work for you, but if you have the problem I had on my Surface, let us know. Yaris678 (talk) 10:05, 28 August 2014 (UTC)[reply]

I wonder if the /Notice template could be changed to something more obviously useful, like "Click on this link to start using Snuggle right now". The "version number" approach sounds kind of like "Here's stuff for developers that nobody else will care about". WhatamIdoing (talk) 17:00, 3 September 2014 (UTC)[reply]

Hey folks. I'm not sure what is going on with IE 11. Can someone confirm from a desktop to make sure it isn't something with the touch interface? (Side note: would be happy to support touch if I can get people to sign on as testers.) Per WhatamIdoing's comments, I have changed the notice to a big blue button. I think that's better, but still not as inviting as I would like. --EpochFail (talk • contribs) 17:32, 3 September 2014 (UTC)[reply]

Error message when inviting to teahouse

An Error message observed in Snuggle when attempting to invite someone to the teahouse.

I got the following error message when attempting to invite someone to the teahouse.

mediawiki: Mediawiki return an error while attempting to performing an action {u'fields': {u'header': u"You're invited to the Teahouse.", u'message': u, u'template': u':Wikipedia:Teahouse/Invitation'}, u'watch': True, u'user': {u'id': 22200887, u'name': u'Wenno123'}, u'action_name': u'invite to teahouse'}. mustposttoken:The 'token' parameter must be POSTed

Any thoughts on the cause? Yaris678 (talk) 10:55, 10 September 2014 (UTC)[reply]

Update: I was able to reprodice that error a few times but now I don't even get that far. I now get an error message in the preview window.

[ERROR] permissions: You must be logged in to preview a user action.

This despite the fact that it says in the corner that I am logged in. Yaris678 (talk) 11:41, 10 September 2014 (UTC)[reply]

Hi Yaris678, thanks for pointing this out. I found that bug 70448 -- which took down Wikilove -- might be the cause of these issues. I've pinged some of the devs involved to dig into it. I'll try to get a work-around in place in the meantime. I'll post updates here. --EpochFail (talk • contribs) 13:30, 10 September 2014 (UTC)[reply]

Done! It was related. Now fixed. --EpochFail (talk • contribs) 13:59, 10 September 2014 (UTC)[reply]

Cool. Thanks. It is working for me. Yaris678 (talk) 07:40, 11 September 2014 (UTC)[reply]

Snuggle --> Other wikis -- New project & Grant proposal

Hey folks. In order to be able to deploy Snuggle on other wikis, I need a source of revision scores (the scores used by ClueBot NG, WP:STiki, WP:Huggle, etc. to flag damaging revisions) on those wikis. For English Wikipedia, STiki has an API that provides these scores, but there's nothing comparable I could find for these other wikis. In trying to solve this problem for myself, I realized that this was a general issue. That a public revision scoring service could be helpful to more tools than Snuggle. So, I have been working with a few people (including STiki's developer West.andrew.g) to organize a project to set up such a public service.

If you have a moment, please take a look at the grant proposal (m:Grants:IEG/Revision scoring as a service) and endorse it if you see fit. Thanks! --EpochFail (talk • contribs) 14:37, 1 October 2014 (UTC)[reply]

Issues with Snuggle

It looks like the desirability scores are not being generated. I don't have time to look at this now, but I wanted to leave a note (and a watchlist ping) for everyone to know that I've noticed the problem and plan to put some hours in on it tonight. I'll ping again when I have more information (or a fix). --EpochFail (talk • contribs) 15:35, 8 October 2014 (UTC)[reply]

It looks like the server that synchronizes with STiki's API stopped. I gave it a kick and it fired right back up. It should take about 5 hours to catch back up. :) --EpochFail (talk • contribs) 04:44, 9 October 2014 (UTC)[reply]

Custom Cohort

The software is great for keeping track of the activity of a set of editors. Right now the candidates are automatically generated - would it be possible to instead supply a set of usernames? Then this could act as a sort of "mentoring dashboard". EdSaperia (talk) 10:03, 4 December 2014 (UTC)[reply]

Downtime: April 1-16 2015

Hey folks,

I'd like to start making note of my work to keep Snuggle online and useful, so expect to see these kinds of updates more often.

So, I noticed that Snuggle stopped synchronizing on March 31st when I went to the server to do some unrelated maintenance. I restarted the sync server and as of this morning, we are up-to-date again.

I'd like to have an alert set up for when synchronization gets far behind, but I don't have the time to invest in that at the moment (pull requests welcome!). If, when using Snuggle, you notice that the activity graphs don't show any activity in the last few days from editors, please feel free to ping me here or in #wikipedia-snuggle (I'm 'halfak'). --EpochFail (talk • contribs) 14:50, 16 April 2015 (UTC)[reply]

May 2015: Down?

My first time trying to use Snuggle, I see blank screens in all the categories (using Chrome and IE) and no activity since 4 May. Peter Chastain ^{[habla, por favor]} 08:40, 19 May 2015 (UTC)[reply]

Hey! Sorry for the delay. For some reason I didn't see your edit. I just noticed the system was down today and gave it a kick. I was just coming here to leave an update when I saw your message. It should catch up again in a few hours. Sorry for the trouble! I have some work scheduled to make sure that this type of downtime happens less often. See Phab:T102680. --EpochFail (talk • contribs) 19:50, 26 June 2015 (UTC)[reply]

video tutorial link is broken

Further information: http://tools.wmflabs.org/videotutorials/Snuggle/

Leads to a 'not here' message

Thanks. It looks like that tool was managed by the late User:Jackson_Peebles. It seems that the tool is now owned by User:Ijon. I'll ping him on IRC to see if I can get a status update. --EpochFail (talk • contribs) 14:27, 11 August 2015 (UTC)[reply]

Fixed now, thanks. Ijon (talk) 22:33, 11 August 2015 (UTC)[reply]

Thanks! valereee (talk) 14:46, 12 August 2015 (UTC)[reply]

not working

So I'm trying to use this for the first time, and when I click on 'start snuggle' I get the tabs, a flash that says 'Loading...' and then nothing. Same for all tabs. valereee (talk) 14:29, 11 August 2015 (UTC)[reply]

Thanks for reporting. It looks like the sync server went down. It should take about 10 hours to get back up-to-date.

There's some maintenance that I've been meaning to schedule that would make it so that the sync server automatically comes back up. I'm waiting until my work on m:R:Revscoring dies down to get that done. In the meantime, thanks for reporting issues. I'll try my best to be responsive. :) --EpochFail (talk • contribs) 14:34, 11 August 2015 (UTC)[reply]

Just checked in this morning and found that the problem was more substantial than I initially thought. It looks like there was a breaking change to the MediaWiki API that I'll need to adjust for. I'll post again once I have worked it out. --EpochFail (talk • contribs) 12:56, 12 August 2015 (UTC)[reply]

Figured out the problem. It looks like the username was hidden from some recent changes and that was causing my synchronizer to get caught up in processing that record. There really should be a contract where if you ask the API for a field, it provides it -- even with a null value if necessary. *sigh*

>>> [d for d in docs if d['type'] in ('new', 'edit') and 'user' not in d]
[{u'rcid': 754800455, u'pageid': 146167, u'timestamp': u'2015-08-05T22:13:35Z', u'userhidden': u'', u'revid': 674749914, u'old_revid': 674688103, u'type': u'edit'}, {u'rcid': 754800567, u'pageid': 146167, u'timestamp': u'2015-08-05T22:14:37Z', u'userhidden': u'', u'revid': 674750012, u'old_revid': 674749970, u'type': u'edit'}]

I'll set a default behavior and be done with it. --EpochFail (talk • contribs) 13:23, 12 August 2015 (UTC)[reply]

Server back online. It'll still take a few hours to sync up again. I'll check back, but feel free to ping. --EpochFail (talk • contribs) 13:29, 12 August 2015 (UTC)[reply]

Very cool, thanks for your responsiveness! valereee (talk) 14:21, 12 August 2015 (UTC)[reply]

Looks like we're all sync'd up. Let me know if you have any other issues. Snuggle is due for a major overhaul now that there are some new systems to support it where we had to do hacks in the past. (e.g. m:R:Revscoring) So please do file your feature requests by posting them here. --EpochFail (talk • contribs) 14:30, 13 August 2015 (UTC)[reply]

Will do, thanks! I'm still kind of figuring it out, at that point where I'm really just trying not to actually do anything WRONG with it. :) But I'll keep in mind that you're going to be doing some work! valereee (talk) 14:49, 13 August 2015 (UTC)[reply]

I still have problems with "Loading...". It appears to be not a server issue, since it depends on the settings. It always hangs when selecting "sorted by total edits". Also, in these cases, the display doesn't change when I switch back that setting; if I refresh, it logs me out, and I log back in, and I get the default settings again, and the server is obviously running. — Sebastian 18:33, 23 August 2015 (UTC)[reply]

I'm also getting no editors in the lists. Happy Squirrel (talk) 16:27, 30 January 2016 (UTC)[reply]

Looking into it now. Thanks for letting me know. --EpochFail (talk • contribs) 16:35, 31 January 2016 (UTC)[reply]

FYI, I'm having trouble even accessing the machine that Snuggle runs on. I'm looking into that now. See T125342. --EpochFail (talk • contribs) 17:31, 31 January 2016 (UTC)[reply]

Thanks to User:Yuvipanda, I was able to get back on the machine back online. It's synchronizing now. The system starts at the oldest edit of the recentchanges feed and marches forward. You should expect to be able to see complete data in a few hours. Thanks for your patience and sorry for the trouble. --EpochFail (talk • contribs) 19:38, 3 February 2016 (UTC)[reply]

desirability 9.223e+18

Why do all the editors I looked at (in the uncategorized tab) have the identical, absurdly high desirability of 9.223e+18? — Sebastian 05:20, 22 August 2015 (UTC)[reply]

hey Sebastian. That's an artifact of how the underlying Bayesian model works. As the number of observations (main namespace edits) increase, the values approach zero or infinity. It turns out that the number you see is a limit in python, so it essentially translates to "infinity". This doesn't have much practical meaning -- it certainly doesn't mean that we are 100% sure this editor is working in good-faith -- so I have been doing some work recently exploring how I can make it more intuitive. Regretfully, I haven't found any good options yet. But I'll be picking up user-level modeling in the m:ORES project soon and that will likely directly feed into Snuggle. --EpochFail (talk • contribs) 16:33, 31 January 2016 (UTC)[reply]

Thank you for your answer. My question actually had two components: One about the value itself, and the other why they're all the same. I get an idea for the former now, but I don't see the point of displaying any value at all, if they all compute to be the same. I don't know the formula you're using (I understand there can be good reasons for not revealing that here), but different users have a different number of edits, which could be included with some simple asymptotic function like 1-exp(-n/n₀). — Sebastian 17:40, 31 January 2016 (UTC)[reply]

Hey Sebastian, sorry to not be clear. The reason you see many values that are the same is because they are presented in sorted order (by highest desirability score first) and there are quite a few that reach the practically infinite threshold. You're right that we gain more confidence as more edits are saved and that means we can be more confident about editors who have made many edits. This acts against the intended purpose of Snuggle for targeting new editors who may not have already edited that much, but I haven't found a good alternative strategy. If you want to learn more about how the score is generated, check out this paper, "Modeling desirability" on page 5. If you have some ideas on how we might change things for the better, I'd be happy to explore updating the modeling/display strategy within Snuggle with you. :) --EpochFail (talk • contribs) 19:51, 31 January 2016 (UTC)[reply]

That seems to fit to what I remember. Sorry, I don't remember much since I only tried it out briefly and didn't use it anymore since my message of 18:33, 23 August 2015 above. I vaguely remember that I found the concept of just going by desirability a bit one-dimensional. There are moments when I just want to be nice and connect with nice people, and there are other moments when I feel up to the more challenging task of finding out whether a new user is nice or nasty. Who I want to connect with doesn't fit in a heuristic box of usefulness for Wikipedia. The more we draw on the variety of our users, the better for Wikipedia. So it would be nice if we could set up our own weighing function; maybe with a few predefined values for "nice people" and "potential troublemakers" and such.

But anyway, you now have one parameter for desirability and may not want to complicate matters, so let's look at the formula:

$desirability={\frac {p(scores|desirable)\cdot p(desirable)}{p(scores|undesirable)\cdot p(undesirable)}}$ .

In this, what does the condition '|desirable' refer to: (A) That a certain edit was found to be desirable? If so, what is the purpose of multiplying by p(desirable); shouldn't it just be 1 in that case? Or (B) The proportion of desirable edits for a given user? And what does 'p(scores)' mean? The syntax suggests it's the probability that there are scores, but there should be scores for everyone, shouldn't there? Or the probability for certain scores, such as above a certain threshold? And what is the population that p is averaged over: All editors, all edits, or all edits by a given editor? Obviously, it would help to see a diagram for the Bayesian model.

But regardless, the usual formula in such cases has the form ${\frac {p_{A}}{p_{A}+p_{B}}}$ , which behaves nicely by not exceeding 1. And since probabilities are already ratios, there may even be no need to compute a ratio of ratios: If $p_{A}+p_{B}=1$ , you don't need a denominator. — Sebastian 01:11, 1 February 2016 (UTC)[reply]

"In this, what does the condition '|desirable' refer to:" It stands for the the probability of seeing this observation (the score) given the distribution of desirable scores. This translates to "The probability of seeing this score if the newcomer is editing in good-faith". See conditional probability for more discussion.

p(scores)

would refer to the probability of seeing these scores given the distribution of all scores -- independent of whether we are looking at the distibution of good-faith or bad-faith scores. These probability distributions are built by extracting scores from a set of manually labeled good- and bad-faith editors. See m:Research:Newcomer desirability modeling for more about where this data comes from and how I fit beta distributions to model it.

I think that you'll find that while a probability and ratio encode the same information, they exhibit very different properties of scale. E.g 1/10 ~ 0.1 ~ 1:9. In this case, we are building a probability ratio of good-faith vs. bad-faith. You're right though that we could limit this to a range between zero and 1. That would look like this:

desirable\_probability={\frac {p(scores|desirable)\cdot p(desirable)}{p(scores|desirable)\cdot p(desirable)+p(scores|undesirable)\cdot p(undesirable)}}

This may be easier to represent, so I'd be interested in exploring that. We might run into some precision issues with a floating point value that we don't hit with the ratio, but I'm not sure if that'd be a real problem or not. --EpochFail (talk • contribs) 02:57, 2 February 2016 (UTC)[reply]

Thank you, this helps, especially the link to m:Research:Newcomer desirability modeling. But I still don't have a clear definition of the various parameters, especially I don't know what p(desirable) is concretely. It would help if you could point me to the code on bitbucket, and if you could clarify how the image File:Newcomer_desirability_scores_(theoretical).png fits to the formula. Do I understand it correctly that x denotes the score of a given user, the left curve is p(scores|desirable), and the right one p(scores|undesirable)?

Can the problem be worded as follows?: Given a distribution of scores p(scores|some_user) for a given user, how similar to p(scores|desirable) and p(scores|undesirable) is it? A standard way to compute this would be by orthonormalization (replacing p(scores|[un]desirable) with two functions u and v that span a basis), and then computing the scalar product of p(scores|some_user) with these two (i.e. computing a := ∑ p(scores|some_user)·u and b := ∑ p(scores|some_user)·v), which will give you the similarity for each. Then a/(a+b) gives you the desirability on a scale from 0 to 1. — Sebastian 08:35, 3 February 2016 (UTC)[reply]

p(desirable)

== 0.80 (the probability that an editor is editing in good-faith given our observations). Here's where the code lives: https://bitbucket.org/grouplens/snuggle I ended up fitting the betas to the observed data using R's fitdistr method and then applying the parameters to pythons scipy.stats.beta -- specifically this code. I think you have the rough sense for what the graph is capturing. There are minor issues with your notation that mean different things, but I think we're on the same page.

This strategy for modeling is the basics of bayesian reasoning. It seems that your math matches my intent from the

desirable\_probability

equation in my last message, so I think we're in agreement there too. --EpochFail (talk • contribs) 21:31, 3 February 2016 (UTC)[reply]

ARWikipedia

Hi :) Does Snuggle work on Arabic Wikipedia?--Reem Al-Kashif (talk) 16:13, 12 October 2015 (UTC)[reply]

hey Reem Al-Kashif. Sorry for the late reply. Somehow I didn't get a watchlist ping. Regretfully we don't have Snuggle running for arabic yet. We're working in that direction though. We need to get a damage-detection model working before the good-faith detection model can work too. See m:ORES for our work on massive, cross-language damage-detection. If you want to get updates while we make progress, add m:R:Revision scoring as a service to your watchlist. --EpochFail (talk • contribs) 16:30, 31 January 2016 (UTC)[reply]

Hi EpochFail. Don't worry. Thank you for your reply. I can contribute with translation into Arabic whenever needed :)--Reem Al-Kashif (talk) 17:09, 31 January 2016 (UTC)[reply]

Reem Al-Kashif, Great! I'll reach out to you soon -- when we get the arwiki damage-detection model deployed in m:ORES since we'll need help to make sure that it is working as intended. Expect that ping some time in the next couple of weeks. --EpochFail (talk • contribs) 17:13, 31 January 2016 (UTC)[reply]

EpochFail fantastic! :)--Reem Al-Kashif (talk) 18:56, 31 January 2016 (UTC)[reply]

@EpochFail:, how's the status now? I'm asking, because it seems this tool might be useful on Polish wikipedia as well. //Halibu tt 11:05, 2 November 2017 (UTC)[reply]

Halibutt, regretfully, I've not got someone to pick up Snuggle engineering work *yet*, but I've been working with Rubberpaw to get some funding to experiment with new deployments of Snuggle. Maybe we could target Polish Wikipedia with those experiments. --EpochFail (talk • contribs) 02:32, 3 November 2017 (UTC)[reply]

@EpochFail:, how much funding are we talking about? Anything I/we can help with? //Halibu tt 08:49, 3 November 2017 (UTC)[reply]

@Halibutt:, I've put in a grant proposal large enough to hire a part-time developer and a full-time research manager for a year to test Snuggle (and one other project) with Wikimedia projects in several languages. Polish Wikipedia is high on the list. I hope to hear back on late December. Since the nonprofit I'm starting (which will support this research) is new, we tend to need grants that will allow us to bring on new staff. --Rubberpaw (talk • contribs) 20:36, 3 November 2017 (UTC)[reply]

Migration coming!

Hey folks, I'm going to be migrating Snuggle to a more recent version of linux (Ubuntu 12.04 --> Debian 8.6). To finish the process off, this will require a little bit of downtime. I'll be doing this migration tomorrow, Nov. 8th at 20:30 UTC. If all goes as planned, you should expect about 15 minutes of downtime. --EpochFail (talk • contribs) 15:56, 7 November 2016 (UTC)[reply]

Migration is complete. Let me know if you run into any trouble. --EpochFail (talk • contribs) 21:41, 8 November 2016 (UTC)[reply]

Active

@EpochFail: Is this still active? I'm getting errors when I try to use it.

I also think it is not working. --Gian (talk) 06:45, 18 May 2018 (UTC)[reply]

@EpochFail: Any news? Is this project no longer operating? Should it be marked as historic? I receive the following error having signed in: database: An database error occurred while attempting to getting a set of users with query {u'category': None, u'direction': u'ascending', u'min_edits': 1, u'skip': 0, u'namespace': u'all', u'limit': 10, u'sorted_by': u'desirability.ratio' Many thanks, Nick Moyes (talk) 11:16, 4 March 2019 (UTC)[reply]

Sorry for the delay! I was on vacation. I'm looking into it now. --EpochFail (talk • contribs) 17:43, 7 March 2019 (UTC)[reply]

OK looks like we're back online. Sorry for the trouble! It might take a couple of hours to catch back up. --EpochFail (talk • contribs) 18:00, 7 March 2019 (UTC)[reply]

Difficulty using Snuggle

This is my ~third attempt to run Snuggle, and it mostly worked this time. It apparently needs wmflabs.org scripts, and snuggle-en.wmflabs.org cookies. I could easily log in with my account. However, no edits at all were displayed in any of the tables. There is the heading, "Username and stats | Interactive edit graph | Talk page activity", but below it, just blank grey. Not sure why this is: whether the server is down or if there is some incompatibility. HLHJ (talk) 23:58, 9 November 2019 (UTC)[reply]

HLHJ, Snuggle is no longer being maintained so its likely that it doesn't work anymore. ~~ CAPTAIN MEDUSA^talk 00:28, 10 November 2019 (UTC)[reply]

Thank you, CAPTAIN MEDUSA. I'm writing about editing tools; do you know of a place I could cite for it not being maintained anymore? HLHJ (talk) 02:53, 10 November 2019 (UTC)[reply]

I should also mention that some of the content in the interface seems not to be delivered over a secured connection, which might be a privacy problem. HLHJ (talk) 17:09, 10 November 2019 (UTC)[reply]

Calling EpochFail -- is it true that Snuggle is deprecated? If so, we need to indicate it as such somewhere. (Thanks for all you do.)---- φ OnePt618^Talk φ 06:16, 3 February 2020 (UTC)[reply]

Hey folks! I don't have much time to spend on Snuggle these days. I've been seriously thinking of decommissioning it. But I'd much rather find another maintainer. Any suggestions? --EpochFail (talk • contribs) 21:55, 3 February 2020 (UTC)[reply]

Are there any updates? I have not had a chance to try the tool out but it seems like a great idea. Thank you for working on it.--Jetam2 (talk) 20:06, 26 May 2020 (UTC)[reply]

Seeking new maintainer

@EpochFail, CAPTAIN MEDUSA, and OnePt618: I think it would be good to ask for new maintainers at village pump/technical, and/or to ask it through sign post. —usernamekiran (talk) 14:39, 3 November 2020 (UTC)[reply]

Usernamekiran, We might find some maintainers at the village pump. ~~ CAPTAIN MEDUSA^talk 13:03, 4 November 2020 (UTC)[reply]

I have started a discussion at Wikipedia:Village pump (technical)#Maintainers needed for Snuggle. ~~ CAPTAIN MEDUSA^talk 13:08, 4 November 2020 (UTC)[reply]

Thanks for taking this on, folks. I'm really happy to help hand over the details to someone new. :) --EpochFail (talk • contribs) 18:11, 24 November 2020 (UTC)[reply]

502 Bad Gateway

I'd like to try Snuggle, but currently it's 502'ing... pandakekok9 (talk) 09:35, 21 December 2020 (UTC)[reply]

Is it just me?

Everytime I tap on any Snuggle link it brings me to a page that says "502: Bad Gateway", is this a problem on my end? Is something wrong with the program? glowing regards, paperandscissors (contact) 06:07, 17 March 2021 (UTC)[reply]

@Paperandscissors No, it is not just you - the project is dead for over months now, the developer does not seem to have time and/or motivation to bring it online again. CommanderWaterford (talk) 09:32, 2 April 2021 (UTC)[reply]

Snuggle officially offline

Hey folks! I've taken Snuggle offline officially. The code is still available, and I have a database backup should someone want to bring it online again. Just let me know and I'll help you get started. The system is built in python and runs on flask/wsgi. --EpochFail (talk • contribs) 21:32, 14 January 2022 (UTC)[reply]