slides-2nd.md

% You shall not publish: Edit filters on EN Wikipedia % HCC Research Group Meeting June 2019 % Lusy

---

Source: Halfaker et al. "The Rise and Decline of an Open Collaboration System: How Wikipedia’s reaction to popularity is causing its decline"

---

Overview

• Motivation
• State of the literature/Literature: What does the scientific community know
• Documentation: What is an edit filter and why was it introduced according to Wikipedia's/MediaWiki pages?
• Data Analysis: Edit filters on English Wikipedia
• Open questions

---

Motivation

• What is the role of filters among existing (algorithmic) quality-control mechanisms (bots, semi-automated tools, ORES, humans)? Which type of tasks do filters take over?
• How have these tasks evolved over time (are they changes in the type, number, etc.)?
• What are suitable areas of application for rule-based systems such as filters in contrast to the other ML-based approaches?

---

State of the Literature

• One thing is ostentatiously missing: edit filters

---

What is an edit filter

• MediaWiki extension
• regex based filtering of edits and other actions (e.g. account creation, page deletion or move, upload)
• triggers before an edit is published
• different actions can be defined

---

Motivations for its introduction

• disallow certain types of obvious pervasive (perhaps automated) vandalism directly
• takes more than a single click to revert
• human editors can use their time more productively elsewhere

---

Edit filters in the quality control mechanisms frame

• the question of infrastructure
• guidelines say: for in-depth checks and problems with a particular article bots are better (don't use up resources)
• they were introduced before the ml tools came around.
• they probably work, so no one sees a reason to shut them down

---

• hypothesis: Wikipedia is a diy project driven by volunteers; they work on whatever they like to work
• hypothesis: it is easier to understand what's going on than it is with a ML tool. people like to use them for simplicity and transparency reasons
• hypothesis: it is easier to set up a filter than program a bot. Setting up a filter requires "only" understanding of regular expressions. Programming a bot requires knowledge of a programming language and understanding of the API.

---

Data Analysis: Edit Filters on EN Wikipedia

---

What do most active filters do?

135  repeating characters & tag, warn
30   "large deletion from article by new editors" & tag, warn
61   "new user removing references" & tag
18   "test type edits from clicking on edit bar" & deleted in Feb 2012
3    "new user blanking articles" & tag, warn

---

Descriptive statistics

all filters: 954
public filters: 361
Active public filters: 110
disabled (but not deleted) public filters: 35
deleted public filters: 216
hidden filters: 593
active hidden filters: 91
disabled (but not deleted) hidden filters: 118
deleted hidden filters: 384

---

Number of filter hits per month March 2009-March 2019

---

Filters Actions

---

Active Public Filters Actions

---

Active Hidden Filters Actions

---

Manual classification

vandalism, good faith and maintenance

• difficult to distinguish
• a lot of subcategories

---

Vandalism

id  hits     public comment 
46	356945	 "Poop" vandalism
365	85470	   Unusual changes to featured or good content
16	2005	   Prolific socker I

---

Good Faith

id  hits    public comment 
180	175939	Large unwikified new article
98	39401	  Creating very short new article

---

maintenance

id  hits  public comment 
577 1566  VisualEditor bugs: Strange icons
345	13832	Extraneous formatting from browser extension
942	1573	Log edits to protected pages

---

Open Questions

---

Current Limitations

• Only EN Wikipedia
• manual filter classification only conducted by me

---

Bigger picture: Upload filters

https://upload.wikimedia.org/wikipedia/commons/c/c5/Blackout_of_wikipedia.de_by_Wikimedia_Deutschland_-_March_2019.png

---

Thank you!

These slides are licensed under the CC BY-SA 4.0 License.

---

Questions? Comments? Thoughts?