Wikipedia:Community health initiative on English Wikipedia/Interaction Timeline

This page documents a feature the Wikimedia Foundation's Anti-Harassment Tools team may build. Development of this feature is complete.

🗣 We invite you to join the discussion!

Tracked in Phabricator
Task T166807

The Interaction Timeline shows a chronologic history for two users on pages where they have both made edits.

The goal of the feature is to help administrators understand the sequence of events between two users so they can confidently make a decision about how to best resolve a user conduct dispute. Non-administrators may also use the tool to determine whether to file a complaint about another user, and to compile evidence for that complaint.

Use the Interaction Timeline

The Interaction Timeline can be found at https://tools.wmflabs.org/interaction-timeline/

How it works

If you provide two usernames and a wiki, the Interaction Timeline will display a chronologic list of edits made by the two specified users on pages where both users have edited. For example:

User:Apples edits on Washington, Texas, and Kansas
User:Bananas edits California, Texas, and Kansas
The Timeline would only show their edits on Texas and Kansas.

The start and end date options allow you to narrow what information is displayed on the Timeline and calculates pages where both users have edited. For example:

User:Apples edits the Beyonce article on January 1 and edits the Adele article on January 2.
User:Bananas edits the Adele article on January 2 and the Beyonce article on January 3.
If the Timeline is queried to show edits from January 1 to January 2, it will show edits by both users to Beyonce.
If the Timeline is queried to show edits from January 1 to January 3, it will show edits by both users to both Adele and Beyonce.

The edits are displayed on a vertical timeline, with the edits made by one user on the left and the edits made by the other on the right. Clicking on the edit card opens a diff in the Timeline for quick review. The diff can be opened in a new tab by middle click. All dates and times on the Timeline are displayed in UTC.

When activity shifts from one user to the other, the Timeline calculates and displays the amount of time between interactions in small red text. We believe this is helpful in understanding if two users are rapidly editing over each other, potentially stalking each other or edit warring.

Known limitations

The Timeline works best when using a narrow date range.
The Timeline does not display participation on Structured Discussions threads.
The Timeline is not optimized for RTL display support.

In the future we would like to add new functionality to alleviate these problems. This may be a way to hide all edits from a certain page, all edits in a certain namespace, or a tool that helps you locate the closely-discussing interactions in a wide date range. Your feedback on this talk page or via email will help us build the best possible version of this tool!

Examples to test

User:Test-apples and User:Test-bananas on test.wikipedia.org
- Link to timeline: https://tools.wmflabs.org/interaction-timeline/?wiki=testwiki&user=Test-apples&user=Test-bananas
- Scenario: Wikihounding. Bananas makes a bad edit, and Apples then follows them around to other pages they've edited.
User:Test-carrots and User:Test-durian on test.wikipedia.org
- Link to query: https://tools.wmflabs.org/interaction-timeline/?wiki=testwiki&user=Test-carrots&user=Test-durian
- Scenario: "He said, she said" dispute, likely boomerang. Durian criticises Carrots until Carrots snaps back. Durian then cleans up their trail and reports Carrots to ANI.
User:Derby pie and User:Sweets lover
- Link to query User:Derby pie and User:Sweets lover on test wiki
- Scenario: Edit warring across articles and reporting each other on ANI with accusations of wikihounding and a ban discussion.
Tinker toys and Baby rattle
- Link to query User:Tinker toys and Baby rattle
- Scenario: Cross article edit warring and user talk page with user warning template.

Leave us feedback

We'd love to hear if this tool is adequately meeting your expectations and needs! We encourage and welcome feedback and ideas of how we can make the Timeline even better at our talk page.

Project information

Updates

December 20

User interviews (mentioned in the November 8 update, below) are underway for the Interaction Timeline. We hope to complete the interviews in January and will work with our designer to see if there any changes we can make so the Timeline is more useful. We won't start development until April 2019 at the earliest.

If you have thoughts about the Timeline, please let us know on the talk page! We still believe the tool has potential for locating and identifying evidence for user disputes.

November 8

After several months of existence, the Interaction Timeline is not seeing the adoption we were hopeful for. This may be due to a handful of reasons so we will be performing some interviews with Wikipedia administrators and other users who investigate user disputes. We have some theories on what would make the tool more useful and look forward to seeing what we learn over the coming weeks.

After our interviews we will either conduct more interviews, leave the Timeline as-is, or work with our designer to address the opportunities raised in the interviews.

Problems we believe the Timeline can solve

As we've observed how user disputes are reported, we've identified some common problems:

Wherever a case is raised, users will often only provide cherry-picked diffs and links to bolster their side of the argument. Knowing the entire story is important to make an informed decision, so users must seek more evidence if they would like to know the full story.
Digging through Special:Contributions, page histories, diffs, Edit Interaction Analyser, and other tools requires switching between multiple tabs. It can be very time consuming and easy to get lost.
Comparing timestamps relative to each other requires a lot of mental processing. Some disputes occur over multiple pages, and determining an accurate sequence of events can be time consuming and complicated.

Project goals

In the end, we want users to be able to make decisions about conduct disputes:

in a time efficient manner (which will ideally result in more disputes being resolved and fewer reports going ignored.)
that result in a fair, appropriate response (the software should show an honest story and enough information for the admin to understand the entire picture)
that leave the admin and community feeling confident about the decision (A lot of these cases are fuzzy: this tool will not be a success if its users are not sure how best to employ it.)

Use Cases

Wikihounding

User:Apple and User:Banana are two contributors. Apple believes Banana is wikihounding them. Banana claims they are editing as per their usual behaviors.
The admin will seek to understand:
- Is Apple new to Banana's domain, or is Banana following Apple? — sequence
- Is Banana following Apple deliberately to frustrate them? — time between edits, context of edit
- What is the nature of both Apple's and Banana's edits? — contents of each edit
Data to show:
- The number of intersecting pages both users have edited (incl. talk pages)
- Who edited each page first (most likely recently, not of all time ever.)
- Amount of time between edits
- Info about the edit itself
  - The edit summaries
  - Link/preview of the diff
  - Was it a revert or rollback?
  - bytes added or removed?
Data we could show, if helpful:
- Edits by other users in between Apple and Banana edits
  - Pro: showing other edits may better reflect the entire story behind a page — maybe these two are just ships passing in the night, as opposed to deliberate hounding.
- Edits Apple and Banana make elsewhere, outside of their direct interaction.
  - Pro: shows if one is fixated on the other (e.g. if Banana only edits what Apple is editing, then it looks like a fixation. If their interaction with Apple is only a fraction of their total contributions, it could be coincidental.)

"He said, she said" harassment

User:Carrot accuses User:Durian of harassment on their talk page, providing a specific diff link. Durian claims this was in response to continual harassment from Carrot.
The admin will seek to understand:
- What was the impetus for this dispute and who instigated it? — contents of each edit
- To what degree is each party guilty of harassment? Or was someone entirely goaded into an outburst? — sequence
Data to show:
- The sequence of interactions the two users had:
  - Edits made to the same articles, talk pages, user talk pages, and elsewhere.
- Amount of time between edits
- Info about the edits themselves
  - The edit summaries
  - Link/preview of the diff
  - Was the edit a revert or rollback?
  - Where there bytes added or removed?
  - Did any of the edits trigger any ORES or AbuseFilters?

Conduct vs. content dispute

User:Eggplant claims that User:Fig is harassing them. Fig states they are just cleaning up Eggplant's messy edits and their claim is baseless.
The admin will seek to understand:
- What did Fig say or do to Eggplant that could be construed as harassment? — content of each edit
- Did Eggplant's edits deserve to be cleaned up? Does Fig's argument have any merit? — content of each edit
- To what extent is Fig following Eggplant: a few pages or many? — pagenames
- How closely is Fig following Eggplan? — time between edits
Data to show:
- The number of pages both users have edited (incl. talk pages)
- Who edited each page first
- Amount of time between edits
- Info about the edit itself
  - The edit summaries
  - Link/preview of the diff
  - Revert or rollback?
  - bytes added or removed?

Preliminary sockpuppet investigation

User:Grape and user User:Horseradish have suspiciously similar styles, and are adding virtually identical promotional content. It is not yet clear whether a Checkuser investigation is required, or evidence is required to justify a Checkuser request.
The admin will seek to understand:
- Do Grape and Horseradish coordinate to harass someone?
- Do Grape and Horseradish edit the same set of unrelated articles? Are they both targeting all articles related to a particular company?
- Are Grape and Horseradish pushing an identical biased agenda on those articles?
- Do Grape and Horseradish magically follow each other around, or are they unrelated users finding these articles being discussed on a wikiproject or notice board?
- Do Grape and Horseradish team up to revert edits of other users, concealing coordinated edit-warring and exceeding revert limits?
- Do Grape and Horseradish show up together in unrelated debates, stacking identical votes and using the same arguments?
- Do Grape and Horseradish mysteriously show up to defend each other on other people's user talk? Do they separately show up arguing on the same other-people's user talk?
- Do Grape and Horseradish both edit an extremely obscure page, such as an archive?
Data to show:
- The number of pages both users have edited (incl. talk pages)
- Who edited each page first
- Amount of time between edits
- Info about the edit itself
  - The edit summaries
  - Link/preview of the diff
  - Revert or rollback
  - bytes added or removed?

Existing features

Editor Interaction Analyser
- My research notes
- Allows up to 20 users, and start/end dates, namespace
- Shows an overview of the pages where both/all users have edited, min time between edits, and the edit count for each user on all pages, highlighting who edited each page first in blue
- 'timeline' links to the per-article contributions for just the provided users, including edit summaries. It also shows the amount of time between interactions (...3 hours... or ...25 seconds...)
WikiBlame — Allows users to search a page's revision for the first inclusion/removal of a specific string
Intertwined Contributions — Combines all contributions of 2+ users, not intersecting contributions.
Intersect Contributions — Shows the pages where both users have edited, not the specific edits.

Log formats

Many MediaWiki tools use a log format which could be used in this feature:

RecentChanges
- (diff | hist) . . PAGENAME; 22:55 . . (+2,926)‎ . . ‎USERNAME (talk | contribs)‎ EDITSUMMARY (Tag: TAGS)
Article history
- (cur | prev) 00:07, 15 March 2017‎ USERNAME (talk | contribs)‎ . . (19,665 bytes) (+866)‎ . . EDITSUMMARY (undo | thank) (Tag: TAGS)
Special:Contributions
- 22:55, 15 March 2017 (diff | hist) . . (+2,926)‎ . . PAGENAME ‎ EDITSUMMARY (current) (Tag: TAGS)
Edit Interaction Analyzer
- 20:40, 15 March 2017 (diff | hist) USERNAME (talk | contribs) . . (12,968 bytes) (+275) . . EDITSUMMARY

These all include:

A link to the diff
A permalink to the revision
Timestamp
Byte change
Username
Edit summary
Link to article history (except history)

Wireframes

First draft

For our first iteration of wireframes, we've anonymized a real example of two users who accused each other of harassment on English Wikipedia's ANI. We identified all the pages where they had both edited and plotted them in a vertical timeline. We extrapolated the section they edited from the edit summary.

Each edit block would have a link or option to view the link, either in-line on the tool itself or in a new tab.

These are wireframes — the exact dimensions, colors, and text will be finalized later. The feedback we need at this stage is on the general layout, arrangement, and organization.

Variation 2
Variation 3, has three significant differences:
1. The ability to add another user to the timeline from the top. We imagine this is also where other filtering parameters would be placed (such as date range, namespaces to exclude, etc.)
2. Both users' edits are on a shared timeline in the center, rather than two parallel timelines.
3. The amount of time between interactions is calculated and displayed in red text.

Second draft

This wireframe is based on Variation 3 above. We made minor changes, most notably showing how in-line diffs could function.

Third draft (current draft)

We've reduced the white space in all of the wireframes. Two of the wireframes are variations of the same one; one without a gray background and less divisions between dates and the second with the gray background and boxes to show the difference between dates. We've also included some options for in-line diffs.