Skip to content

R0051/2026-03-31/Q002/SRC05/E01

Research R0051 — Fact-Checking Gap
Run 2026-03-31
Query Q002
Source SRC05
Evidence SRC05-E01
Type Statistical

Of 16 annotated indicators, only 2 content indicators (clickbait title, logical fallacies) significantly predicted credibility.

URL: https://researchgate.net/publication/324639995_A_Structured_Response_to_Misinformation_Defining_and_Annotating_Credibility_Indicators_in_News_Articles

Extract

Zhang et al. (2018) analyzed 40 widely-shared public health and climate change articles using 16 credibility indicators annotated by 6 trained annotators. The 16 indicators included content indicators (title representativeness, clickbait title, quotes from outside experts, citation of organizations and studies, calibration of confidence, logical fallacies, tone, inference) and context indicators (originality, fact-checked, representative citations, reputation of citations, number of ads, spammy ads, number of social calls, placement of ads and social calls).

After model convergence, only 2 content variables remained significant predictors: clickbait title and logical fallacies (slippery slope). For context-based signals, 6 variables remained: fact-checked (reported false), fact-checked (reported mixed results), number of social calls, number of mailing list calls, and placement of ads and social calls.

Notably, "calibration of confidence" was among the 16 tested indicators but did not survive model convergence — suggesting it was not a reliable predictor of credibility in this sample.

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Contradicts Limited empirical validation — confidence calibration did not predict credibility
H2 Supports Real empirical work was done but results were limited
H3 Contradicts Genuine research with results was produced

Context

The finding that "calibration of confidence" was tested but did not survive as a significant predictor is relevant — it suggests that even where the Coalition attempted to formalize confidence assessment, the empirical results did not support its utility in their framework.