Reliability: HIGH — Official W3C Community Group page.
Relevance: DIRECT — Primary authoritative source on group status.
Bias assessment:
Institutional: W3C (neutral standards body).
Self-reporting: Group describes its own mission and status.
Key contribution: States mission to "help shift the Web toward more
trustworthy content without increasing censorship or social division."
Strategy focuses on data sharing using JSON-LD and specifying "credibility
indicators."
Reliability: HIGH — Official specification document.
Relevance: DIRECT — The original credibility signals specification.
Key contribution: Defines credibility signals as "small units of
information used in making a credibility assessment." Organized hierarchically
by subject type. Signals are observational features, not evaluative
judgments — they describe what CAN BE observed, not how to WEIGHT
observations.
Reliability: HIGH — Official specification; represents group consensus.
Relevance: DIRECT — The curated subset of signals approved as "promising"
by the group.
Key contribution: CRITICAL — This is the most mature output of the group.
Only signals reviewed and approved as "promising" are included. Examples
include journalism awards, site longevity, and corrections policies as source
reliability indicators. These are BINARY observational signals (present/absent
or measurable), not GRADED quality assessments.
Key contribution: The CCIV was SUPERSEDED by the Credibility Signals
specification. The document itself notes it is archival and directs users to
use Credibility Signals instead.
Reliability: MODERATE — Blog post from a participating organization.
Relevance: HIGH — Provides context on the initiative's goals and design
philosophy.
Bias assessment:
Institutional: Meedan is a participating organization (positive bias toward
the initiative).
Key contribution: Articulates the design philosophy: signals are
DESCRIPTIVE indicators that help users make their own credibility decisions,
not PRESCRIPTIVE quality judgments.
Relevance: DIRECT — Current landing page showing latest outputs.
Key contribution: Lists specifications, meeting agendas, and reports. The
most recent dated specification is the Reviewed Credibility Signals
(2020-02-24).
Key contribution: Describes goal to "understand the veracity, quality and
credibility of online information" — a research community, not a standards
body.
Relevance: HIGH — Best evidence of current activity level.
Key contribution: CRITICAL — Quarterly check-in meetings in 2024
(February 7, May 1, August 21, September/TPAC). The August 2024 meeting
featured guests discussing Japan's Originator Profile initiative. This
pattern (quarterly meetings featuring external presentations) is consistent
with a group in networking/maintenance mode rather than active specification
development.
Relevance: HIGH — Best evidence of specification development activity.
Key contribution: CRITICAL — Search found no commits more recent than
October 2019. This indicates specification development has effectively
stopped, even though the community group still meets.
Relevance: HIGH — Later version of the signals specification.
Key contribution: Signals are organized in "related groups in hierarchical
sections" with signals at the lower level and groupings at the higher level.
The hierarchy is ORGANIZATIONAL (grouping signals by topic) not EVALUATIVE
(ranking evidence quality). Subject types include the claim, article, source,
and media.
Relevance: HIGH — Overview of the group's approach and its relationship
to other initiatives.
Key contribution: Describes credibility analysis as "gathering, organizing,
and analyzing evidence to help people make credibility decisions." The output
is a credibility assessment report with a credibility score. However, the
specification focuses on WHAT signals to gather, not HOW to weight or combine
them into quality judgments.
Reliability: MODERATE — Preprint, not yet peer-reviewed.
Relevance: HIGH — Recent survey referencing credibility signals work.
Key contribution: Shows continued academic interest in credibility signals
as features for automated systems, suggesting the vocabulary has some
downstream adoption in NLP research even if the specification itself is not
actively developed.