Community health initiative/Blocking tools and improvements/Block notices

Block notices are the messages that appear to blocked users that inform them of their block status and parameters. In 2018 the WMF's Mobile Web and Anti-Harassment Tools teams made some changes to the mobile block notice and measured how often they appear to users across different devices.

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

🗣   We invite you to join the discussion!
🛠   Track in Phabricator at T189724.

Block notices can be displayed in various situations. Commonly they are displayed when a blocked user tries to edit, but they may also be displayed when attempting to log in or send an email, or on other special pages. This analysis concentrates on block notices shown when attempting to edit.

Summary edit

  • Block notices appear very often on the largest Wikimedia projects, sometimes outnumbering actual edits. (6.2 million blocked edits occured on English Wikipedia over a 30-day period.)
  • The desktop wikitext editor sees the vast majority of impressions by a wide margin. (98% on English Wikipedia, 89.5% on Spanish, and 98.7% on Russian.)
  • The VisualEditor and mobile block notices may occur less frequently, but still display to thousands of people every month. There are hurdles to optimizing their displays, given the legacy of desktop-first design.

Display edit

Block notices may differ according to the type of block, the page the user is attempting to access, the wiki's local customization, or the user's device.

Here are the block notices shown for a block made by a sysop against another user, when the blocked user attempts to edit:

Components edit

When a user is blocked, the system can display the following information. Not all needs to be displayed:

  • The current session's username or IP
  • The target username, IP address, or autoblock ID number
  • The admin who configured the block
  • The timestamp of when the block was set
  • The timestamp (and/or calculated time remaining) of when the block expires
  • The block reason
  • If the block is partial or sitewide

The desktop messages include information on how to report an incorrect block. The mobile block notices do not, because the workflow is not optimized for mobile.

Desktop notices edit

On desktop, one of the following messages will be displayed to blocked users attempting to edit, log in, email, perform certain actions or access certain special pages, depending on their block:

Block type English translation Documentation
Sitewide block made by a sysop against a user, IP or range blockedtext blockedtext
Automatic block against an IP address that has been used by a blocked user autoblockedtext autoblockedtext
Partial block made by a sysop against a user, IP or range blockedtext-partial blockedtext-partial
Block against an IP address created by the MediaWiki software, e.g. due to blacklisting systemblockedtext systemblockedtext
Block created by the MediaWiki software when a user or IP address has more than one block blockedtext-composite blockedtext-composite

Like all messages, any of these may be locally overridden on particular wikis. Currently, the 'blockedtext' message is overridden on several Wikipedias, while the others are less commonly overridden.

Mobile notices edit

On mobile, a shorter message is displayed, with key features about the block:

  • Reason for the block
  • User who made the block (only if the block was made by a user)
  • Link to the block's entry on Special:BlockList (only if the block is listed there)

Data edit

 
Charts of how frequently block notices display on 20 Wikimedia wikis, December 2018 to January 2019.

In December 2018 we enabled tracking on block notices to roughly answer the question "how many times do these notices display?" An interactive dashboard of graphs is available at this Grafana link and a screenshot of data from December 19, 2018 to January 18, 2019 is embedded on this page. Here are some preliminary findings:

  • Far and away, the most common format which people see the block notices is on the desktop wikitext editor. (They display as green in the charts.)
    • English Wikipedia — 98.47% of all block notices that appeared were the desktop wikitext editor, resulting in 6.1 million impressions.
    • Spanish Wikipedia — 89.5%
    • Russian Wikipedia — 98.74%
    • French Wikipedia — 94.06%
    • Italian Wikipedia — 62.21%
  • The second-most common display of the block notice is the VisualEditor. The percentage of these is much lower than expected. A next step for investigation would be to calculate the ratio of wikitext edits to wikitext block notices, and visual edits to visual blocks notices.
  • The occurrence of API block notice displays vary greatly from wiki to wiki, but appears steady across all 30 days. These are likely unsophisticated spam bots that are blocked but still attempt to edit.
  • The ratio of edits to blocked edits varies greatly on a wiki by wiki basis. (See table below.) On some wikis there are more attempted edits that are blocked (RU, EN) but on others the edits greatly outweigh the block notice impressions (ES, IT).
    • Another aspect that should be explored is page protection as a spam, vandalism, and misconduct mitigation tool.
English Wikipedia EN % Total Spanish Wikipedia ES % Total Russian Wikipedia RU % Total Italian Wikipedia IT % Total
Blocks Weekly 9,937 0.88% 282 0.48% 40,019 2.04% 273 0.49%
IP (current) 86,870 7.66% 2,387 4.02% 1,856,750 94.81% 1,034 1.84%
IP Range (current) 8,502 0.75% 24 0.04% 47,856 2.44% 30 0.05%
User (current) 1,038,076 91.56% 56,886 95.90% 54,213 2.77% 55,183 98.08%
Total (current) 1,133,807 59,320 1,958,366 56,263
Monthly Block notices Wikitext 6,100,000 98.47% 40,000 89.54% 1,000,000 98.74% 10,700 62.21%
Visual 13,000 0.21% 772 1.73% 4,400 0.43% 1,000 5.81%
Mobile web 52,000 0.84% 2,400 5.37% 4,800 0.47% 2,000 11.63%
API 30,000 0.48% 1,500 3.36% 3,600 0.36% 3,500 20.35%
Total 6,195,000 - 44,672 - 1,012,800 - 17,200 -
Monthly Edits IP 792,886 19.66% 177,745 33.18% 86,226 19.30% 114,539 29.19%
User 3,241,069 80.34% 358,007 66.82% 360,554 80.70% 277,861 70.81%
Total 4,033,955 - 535,752 - 446,780 - 392,400 -
Monthly Editors IP 298,260 68.08% 75,217 79.45% 34,917 73.99% 35,859 79.91%
User 139,818 31.92% 19,457 20.55% 12,274 26.01% 9,013 20.09%
Total 438,078 - 94,674 - 47,191 - 44,872 -
Ratios Notices : All blocks 5.46 0.75 0.52 0.31
All blocks : Notices 0.18 1.33 1.93 3.27
Notices : Edits 1.54 0.08 2.27 0.04
Blocks set : Edits 0.00 0.00 0.09 0.00
Notices : IP Edits 7.81 0.25 11.75 0.15
Blocks set : IP Edits 0.01 0.00 0.46 0.00
Notices : Editors 14.14 0.47 21.46 0.38
Blocks set : Editors 0.02 0.00 0.85 0.01
Notices : IP Editors 20.77 0.59 29.01 0.48
Blocks set : IP Editors 0.03 0.00 1.15 0.01
Mobile : Desktop notice 0.01 0.06 0.00 0.17

Data sources:

Blocks set April 30 - May 7, 2018 https://docs.google.com/spreadsheets/d/1_4GZ2WUurxaehlNeab5mF7VDgOgHd-PXjFjsLD5tY2Q/edit#gid=1703473757
Block notices displayed Dec 19 - Jan 18, 2019 https://grafana.wikimedia.org/d/000000618/blocknotices?orgId=1&from=now-7d&to=now
Edits + Editors Nov 2018 https://stats.wikimedia.org/v2/#/it.wikipedia.org/contributing/edits/normal|bar|1-Month|~total
Blocks active Dec 20 https://quarry.wmflabs.org/query/32169


Other notes:

  • Ideally all of this data would be from the same period of time (e.g. January 1 to January 31, 2019.) This did not happen for this initial report, for time constraints. I'll aim to recalculate everything in early February.
  • Data could be affected by adblock and privacy blockers. The actual numbers will be higher.
  • API edits include mobile app edits, bot edits, and a longtail of other sources
  • The visualeditor 'returned' data is calculated via API and the 'shown' is calculated via tracking pixel. In short, 'shown' means a human saw it.
  • There are usual seasonality changes around late December for the holidays — edits usually decrease. We'll leave the tracking on until March 1 at a minimum to capture two full calendar months of data. (phab:T214214)
  • There appears to be something interfering the data from properly gathering on German Wikipedia. We do not have time to investigate, it will just have to be an unsolved mystery.

Considerations for the future edit

Intended audiences & design edit

Blocks are set to quarantine damage from occurring to the wikis, they are not set as punishments. There are two intended audiences who may see these block notices, and one unintended audience:

  1. People who are temporarily blocked who can return after cooling down
  2. People who are not welcome to ever return
  3. Accidental bystanders who happen to be on the same IP address or range.

In the case for #1 and #3, the block notice should likely be informative on when the block expires and/or how to request the block be lifted. For #2, the block should be clear but not motivate the people to evade the block. With the current block software it is impossible to differentiate between these, but maybe in the future we want to offer different styles of block notices.

Templated reasons edit

The block notice field supports wikitext. Many wikis have used this functionality to build desktop-appropriate block messages, such as the English Wikipedia screenshot above. This message does not appropriately display in the VisualEditor or on mobile (web or apps.) A new approach to block reasons should be considered to 1) allow for better display across experiences and 2) allow for more structured logging, to better help determine the efficiency of blocks.

Some Phabricator tasks for this problem include: phab:T190946 and phab:T189717.

See also edit