User:Hearvox/WikiSignals: Difference between revisions – Wikipedia

 

Line 83: Line 83:

* [https://www.mediawiki.org/wiki/Wikimedia_Enterprise/Credibility_Signals Wikimedia_Enterprise/Credibility_Signals]

* [https://www.mediawiki.org/wiki/Wikimedia_Enterprise/Credibility_Signals Wikimedia_Enterprise/Credibility_Signals]

=== WikiSignals user script ===

=== WikiSignals ===

We’ll be demoing our too mockup at WCNA 2025:

* Script: [[User:Hearvox/WikiSignals/wikisignals.js]]

* : [[User:Hearvox/WikiSignals/wikisignals.js]]

* WCNA 2025: [https://wikiconference.org/wiki/Submissions:2025/WikiSignals Program Submission]

* WCNA 2025: [https://wikiconference.org/wiki/Submissions:2025/WikiSignals Program Submission]

Our tests for fetching and compiling domain data are at GitLab:

* API-calls script: [https://gitlab.wikimedia.org/toolforge-repos/veri-fyi/-/blob/WikiSignals/ComputeCredibilityIndicatorsAsync.ipynb?plain=0 ComputeCredibilityIndicatorsAsync.ipynb]

* Domain-data output: [https://gitlab.wikimedia.org/toolforge-repos/veri-fyi/-/blob/WikiSignals/CredibilityIndicators_PerennialSources_MBFC_20250919.json JSON] | [https://gitlab.wikimedia.org/toolforge-repos/veri-fyi/-/blob/WikiSignals/CredibilityIndicators_PerennialSources_MBFC_20250919.csv CSV]

{{#section-h:User:Hearvox/WikiSignals/Reliability-Indicators|WS-Nav}}

{{#section-h:User:Hearvox/WikiSignals/Reliability-Indicators|WS-Nav}}

WikiSignals is a proposed credibility-related Wikipedia Editing Tool.

Determining the reliability of a source could be easier if data to help with that determination was more accessible. WikiSignals compiles a range of available data so Wikipedians can better evaluate source credibility.

The project grew out of WikiCredCon 2025. It’s a collaboration between Veri.FYI, Factiverse, SimPPL, AfroCrowd, Iffy.news, and the Internet Archive Reference Explorer: organizations pioneering efforts to improve access to source-credibility information.

Wikipedia’s core mission is to be a source of credible information. This requires maintaining the credibility of not only internal articles but also external references. Wikipedia has a Perennial Sources list that ranks sources by their reliability. However, this list is only available for a handful of languages. The Wikipedia English is the the most developed and even that has ranks for only a few hundred websites (out of countless millions).

WikiSignals aims to close this gap. Participants in the WikiCred initiative are developing a Wikipedia editing tool that automates editor tasks related to credibility. The tool provides reliability-related signals for sources being considered for citations, references, and external-link lists. It returns website domain-level data^[1] as well, including global traffic rank, registration date, and social media usage statistics — information that can be used to judge the quality of the reference.

The WikiSignals tool can expand to assist with other editing tasks. We plan to interview Wikipedia editors to find and address problems that WikiSignals might address, particularly with respect to the user workflow of editing. Our broader goal is to become part of an effort to build a comprehensive reference dataset of credibility indicators for news outlets, science journals, and possibly other source types.

• Check external reference-source credibility
• List reliability-related signals for reference.
• List social-media signals for reference’s domain.
• Save URLs in Wayback Machine archive.
• Return saved archive URLs in wikitext format.

1. Interview active Wikipedia editors (July-August)
2. Design and present mockup edit-form User Scripts (June-August)
3. Aggregate/evaluate credibility indicators for website domains, URLs and social media (July-September)
4. Incorporate credibility indicators into User Script (July-September)
5. Add feedback form for reporting errors and requesting features (August-September)
6. Optimize UI/UX (June-October)

• Factiverse— Verify your information and mitigate risk. AI-powered fact-checking solutions tailored to media and finance that accelerate research, access credible sources, and protect organizational reputations.
• Iffy.news— Tools for mis/disinfo research.
• Internet Archive Reference Explorer— Enter a Wikipedia URL and click “Load References” to display the credibility of citation resources. Part of IA’s Turn All References Blue team.
• SimPPL— Rebuilding trust on the social internet. A research collective that designs open access tools and conducts research to cultivate transparency and authenticity online.
• Veri.FYI— Assess websites for misinformation risk based on publicly available data. Veri.FYI takes a URL or domain as an input and outputs indicators that can be used to assess the source’s reliability.

People: Chris Rusnick (Localmotion127), Gaute Kokkvoll (Gautek), Swapneel Mehta (SwapneelM), Sherry Antoine, Barrett Golding (Hearvox), Chris Lombari (iamojo).

Domain-level reliability signals gathered by this project (mostly machine-harvestable):

• Domain registration date
• Global site rank: Similarweb ranking (P10768) or Tranco
• Date website first archived (e.g., Wayback Machine)
• Press association: member of (P463)
• Inclusion in vetted index (e.g., Google News and Local News Initiative)
• Correct URL: official website (P856) (most Infobox website URLs have the old, now-incorrect http:// protocol)
• Local address (in standardized format): headquarters location (P159)
• Ownership (P127)
• Funding (P8324)
• Library of Congress authority ID (P244)
• Library of Congress Control Number (LCCN) (P244)
• Chronicling America newspaper ID (P4898) (same as LCCN)
• Online Computer Library Center (OCLC) (P243)
• International Standard Serial Number (ISSN) (P236)
• Freedom Forum newspaper front page ID (P6136)
• Media Bias/Fact Check ID (P9852)) with assessment outcome (P9259) — reliable, unreliable, or mixed
• Wikipedia citations (domain via Citations Database)

For more on indicators, see W3C CredWeb CG: Reviewed Credibility Signals, NewsQ: Signals, The Trust Project: Trust Indicators, AAM: Trust Initiative, and HackerNoon: Emoji Credibility Indicators.

URL-level reliability signals:

• Wikipedia citations
• Wayback Machine captures: oldest, newest, total
• HTTP Status code (e.g., 200)

Social-site data for news/science media compiled by this project:

See Wikidata Identifiers for The New York Times and JAMA.

We’ll be demoing our too mockup at WCNA 2025:

Our tests for fetching and compiling domain data are at GitLab: