Talk:Community health initiative/Interaction Timeline

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search


Archive: 1

Irrelevant results due to long-ago edits[edit]

I ran a comparison of my staff and volunteer accounts for January and February 2018, and the results were basically useless. One column was mostly filled with w:en:WT:MED, and the other column was mostly filled with my sandbox. Since someone pinged my staff account to WT:MED once or twice (but not during the stated time period), and since I've occasionally used my volunteer account to test edit conflicts in my sandbox (but, again, not during the stated time period), the resulting timeline was filled with claims about "a day between interactions", even though there were no interactions on these pages.

The Village Pumps (and similar large discussions/noticeboards) are a particular problem for this model. It is very common to have two people posting to a board such as w:en:WP:ANI within moments of each other but without having any actual interaction. Whatamidoing (WMF) (talk) 15:30, 12 March 2018 (UTC)

@Whatamidoing (WMF): Yes, this is one of the areas for improvement we're already starting to see emerge. Our definition of "pages where both users have edited" is currently 'all time' (e.g. if I made one edit to the Beyonce article in 2008 and you made one edit today, her article would be displayed on the timeline.) And yes, common pages like the VP are often more noisy than helpful. This definition is proving to be too broad and we're starting to think of ways to fix this. Ideas welcome here or on our Phabricator task phab:T189237.
"the resulting timeline was filled with claims about "a day between interactions", even though there were no interactions on these pages." Could you please share a link when this happened?
Thank you! — Trevor Bolliger, WMF Product Manager 🗨 22:24, 12 March 2018 (UTC)
The no-interaction time stamp is probably due to the 'all time' setting for interactions.
I think you might get some performance improvements by reducing 'all time' to the specified time window, at least for highly active editors. Whatamidoing (WMF) (talk) 02:06, 13 March 2018 (UTC)

Autocomplete[edit]

Hi. Thanls a lot for the tool. Since you do not provide two different fields for users, the tool allows input of more. But the third user is autocompleted by a weird little set of values (AutoIP, AutoImage, and so on, 9 values), which throw an error 400. Could you check this, please? IKhitron (talk) 17:48, 15 March 2018 (UTC)

@IKhitron: Yes, this is a problem we fixed before, but it decided to break again! It's filed at phab:T188289. Thank you for reporting it! — Trevor Bolliger, WMF Product Manager 🗨 18:20, 15 March 2018 (UTC)
(-: Thanks. IKhitron (talk) 18:22, 15 March 2018 (UTC)

Timeline V1.1 changes[edit]

Hello all! We thought we reached a stopping point with the Timeline but found there were a few changes we want to make after some initial feedback. These changes will go live shortly, and we'll post an update when it's complete.

  • phab:T189849 — Change 'N time between interactions' to 'N time between edits'
  • phab:T189847 — Make the start date use a default start date of 30 days before current calendar date
  • phab:T189237 — Change definition of "pages where both users have edited" to respect the start and end dates

We believe this will make the Timeline easier to use by removing clutter from activity not pertinent to the selected date range. Feedback welcome here or on Phabricator. Thank you! — Trevor Bolliger, WMF Product Manager 🗨 23:42, 26 March 2018 (UTC)

IP Users?[edit]

When trying to intersect with an IP user, the tool fails out with Error 400. — xaosflux Talk 20:05, 5 April 2018 (UTC)

Created phab:T191587. — xaosflux Talk 01:09, 6 April 2018 (UTC)
Oh yikes! Thanks for finding and filing. We'll get on it soon. — Trevor Bolliger, WMF Product Manager 🗨 16:44, 9 April 2018 (UTC)

Real case feedback[edit]

Hello, The Anti-Harassment Tools team has completed V1.1 and the tool is ready for use. The Interaction Timeline shows a chronologic history for two users on pages where they have both made edits.

The purpose of the tool is to better understand the sequence of edits between two users in order to make a decision about the best way to resolve a user conduct dispute. Here are some test cases that show the results and also some known limitations of the tool. We would like to hear your experience using the tool in real cases. You can leave public feedback on talk page or contact us by email if the case needs discretion or you would prefer to comment privately. SPoore (WMF) (talk) , Trust and Safety Specialist, Community health initiative (talk) 17:40, 9 May 2018 (UTC)

Clicking to expand/collapse the inline diffs works really well, but there's an awkward detail. When you expand the inline diff it adds a colored bar with the username and date. That colored region has the same look&feel as the colored area above it - like a single expanded colored region. Clicking the upper colored area collapses the inline diff - good. Clicking the lower part of the colored area (trying to collapse the diff) behaves as a link to the diff page. Then after mistakenly loading the diff page, the back button doesn't really work right because it's an "infinitely scrollable page". The link to the diff page should be presented differently.
I think you can/should eliminate the darker-gray horizontal dividing bars between days. The gray date-boxes on the left already do a good job of fulfilling the same purpose. The bar is redundant whitespace, and it can arbitrarily split edits that are very close in time. Alsee (talk) 12:29, 10 May 2018 (UTC)
@Alsee: I agree that the link for the revision is "unfortunately discoverable." I've created phab:T194439 to make it better, thank you for sharing your perspective! I'm of two minds for the darker-grey day separator boxes. I was skeptical to add them but they do make the timeline feel a little more manageable and less like a never-ending scroll feed. We're discussing with the WMF design team to do a repaint to make the fonts, colors, and icons match the Wikimedia style guide. When we do the repaint we'll probably remove those grey boxes. — Trevor Bolliger, WMF Product Manager 🗨 22:39, 10 May 2018 (UTC)
At this point I think the tool works well, at least for me. Have you gotten any feedback from administrators who are supposed to be using it? It was not used during my recent appearance at the English Wikipedia's Administrator's Notice board where I was accused of 'following' another editor when instead the interaction tool demonstrates the exact opposite. I'd like to know if I am interpreting the tool correctly. So here is a scenario: Editor A makes an edit to the article on Cats. Editor B is simultaneously editing the article on Dogs and has made no recent edits to the article on Cats. Editor B then stops editing the article on Dogs and reverts the edit of Editor A thirty seconds after Editor A has made their edit to the article on Cats. Two different things could be happening here. Perhaps Editor B has the articles on Cats on their watchlist and got a notification. Or, perhaps Editor B has the contributions of Editor A in a browser window so they can 'follow' the edits of Editor A. A one-time 30-second reversion is not proof of harassment. If Editor B continues to appear within minutes of Editor A's edits to a variety of articles and reverting edits, might that mean that 'harassment' is occurring? Is there any other reason why an editor would respond so quickly to the edits of another to revert? Is there another interpretation of the interaction tool? Best Regards, Barbara (WVS) (talk) 20:55, 10 May 2018 (UTC)
@Barbara (WVS): Yes, this is a perfect use case for the Interaction Timeline. It only shows edits on pages where both users have edited, so for your example only edits to the page Cats would display — not any edits to Dogs, because Editor A has never edited Dogs. This is an example we made up to illustrate how it could be used.
We are currently working on spreading the news about this tool. We are going slow because we want to incorporate any feedback about the tool, like Alsee above. My colleague User:SPoore (WMF) is working on getting this tool in the hands of people who could most benefit from it. — Trevor Bolliger, WMF Product Manager 🗨 22:39, 10 May 2018 (UTC)

Real world use of the tool[edit]

I've been keepng an eye on disussions posted to ANI on the English Wikipedia but not have seen the tool being used. I will pay closer attention.

I have another scenario:

  • Editor A makes an edit to the Cat article.
  • Editor B makes an edit to the Dog article a few minutes before editor A makes their edit to the Cat article.
  • Editor A has never edited the Dog article.
  • Editor B appears on the Cat article to revert an edit that Editor A has made a few moments beforehand.

Will the interaction tool leave out the information that editor B was editing the dog article minutes before editor B went to the cat article to revert the edits of editor A?

How about user talk pages?

  • Editor A leaves a comment on the talk page of Editor C.
  • Editor B leaves a comment responding 'quickly' to the comment of editor A on Editor C's talk page.
  • Will this appear on the editor interaction analyser?

Thank you and Best Regards, Barbara (WVS) (talk) 13:47, 29 May 2018 (UTC)

Hello @Barbara (WVS): and thank you for your questions. The Timeline only shows edits to pages where both users have edited within the selected time frame. The edits could be to different sections of the page or have been reverted, but both users must have published edits stored in page history. The tools treats User talk pages like any other page.
For your examples, only the edits to 'Cat' will display on the Timeline. For the user talk page, if you query the Timeline for interactions between User A and B, the Timeline will show User_talk:C. — Trevor Bolliger, WMF Product Manager 🗨 21:39, 29 May 2018 (UTC)

How can the Interaction Timeline be useful in reporting to noticeboards?[edit]

Hi everybody,

I’m Trevor, a Wikimedia Foundation product manager with an update about the Interaction Timeline. We built this tool to make it easier to understand how two people interact and converse across multiple pages on a wiki. The Interaction Timeline compliments a few related existing tools such as the Interaction Analyzer and Intertwined Contribs because it shows one consolidated chronologic list of edits by two users, only on pages where they have both made edits within the provided time range. This tool has the attention of the Wikimedia Foundation’s Anti-Harassment Tools for new feature development and maintenance. We have dozens of ideas for new features but need your input on how to make this tool even more useful as you look into incidents of user misconduct.

Today we’re interested in how we can make the Interaction Timeline generate a summary of results with links and data. Our goals are to assist users to make well informed decisions in incidents of user misconduct and to keep on-wiki discussions civil and focused on evidence. We think the Timeline fits into the important workflows for evaluating user misconduct:

  1. You become aware of an incident of user misconduct (in the wild, or it has been reported on a noticeboard)
  2. You investigate the situation with your tools of your choice
  3. You report back what you’ve found to help the decision move forward.

We’d like to build functionality that can bridge the gap between 2 and 3. The Timeline already generates some statistics about the users and their interactions (the number of pages where they both edited, the number of edits they both made, the amount of time between edits, etc.) and we think generating some wikitext with this information could be beneficial. This wikitext could then be copied+pasted into an on-wiki discussion. For example:

Between 2017-01-01 and today on test.wikipedia.org {{User|Test-apples}} and {{User|Test-bananas}} both made 37 edits to the same 13 pages. The shortest amount of time between edits occured on the article [[Candy]] on 2018-03-23 when User:Test-apples made [[Special:Diff/350114|this edit]] and a few seconds later User:Test-bananas made [[Special:Diff/350115|this edit]]. You can see a chronological list of all their edits on [https://tools.wmflabs.org/interaction-timeline/?wiki=testwiki&user=Test-apples&user=Test-bananas&startDate=1483228800 the Interaction Timeline] or a table view on [https://tools.wmflabs.org/sigma/editorinteract.py?users=Test-apples&users=Test-bananas&users=&startdate=20170101&enddate=&ns=&server=testwiki the Interaction Analyser]. — ~~~~

Which would render as:

Between 2017-01-01 and today on test.wikipedia.org Test-apples (talk · contribs) and Test-bananas (talk · contribs) both made 37 edits to the same 13 pages. The shortest amount of time between edits occured on the article Candy on 2018-03-23 when User:Test-apples made this edit and a few seconds later User:Test-bananas made this edit. You can see a chronological list of all their edits on the Interaction Timeline or a table view on the Interaction Analyser.


Would this type of shortcut prove beneficial in your harassment or wikihounding investigations? If so, what type of information would you like to see included in the wikitext? Thank you! — Trevor Bolliger, WMF Product Manager 🗨 21:56, 14 June 2018 (UTC)

Has this tool been used yet?[edit]

I've been searching the ANI discussions on the English wikipedia and have not been able to find one instance when this tool has been used. I noticed that its availability has been made known to many on en:wikipedia. I would like to see an example of its use. My impression at ANI is that an editor who is skilled in cherry picking has more 'success' than any tool can counter. I want to be wrong and I want to see this tool widely used. We are losing a lot of editors. Best Regards, Barbara (WVS) (talk) 21:32, 23 June 2018 (UTC)

Hello, Barbara (WVS), and thank you for your question. We can see that people are accessing the tool, but we too were expecting to see more links to it included on ANI. We are currently working to determine if this is an awareness issue or if the tool is not effective. The Interaction Timeline currently has a minimal feature set and we would like to get input about the next enhancements we should build. We designed and built this tool because we saw that cherry-picking diffs is time intensive to verify or challenge, and I still believe it has the potential to accomplish this. — Trevor Bolliger, WMF Product Manager 🗨 21:47, 25 June 2018 (UTC)