Research Literacy

Weighing evidence and sources, and reading a paper past its abstract

The two-axis filter this site's own /protocol research dossiers run on in real code — evidence tier (how good the underlying data is) and source tier (who's telling you, and what they stand to gain) — tracked as two separate labels, because merging them into one trust score is the single most common failure in how people evaluate what they read. Works through all six evidence tags and all six source tiers against real, named cases (a whale-longevity paper's CIRBP finding, the taurine reversal, POM Wonderful's FTC case, Coca-Cola's Global Energy Balance Network), teaches where an actual paper hides its own numbers using the power-posing study and its own lead author's public retraction as the worked example, names four checkable red flags (publication bias, conflict of interest, an unreplicated small finding, correlation mistaken for causation) against real cases for each, and closes by running the discipline on one real claim end to end — the marshmallow-test delay-of-gratification finding, from its thin original sample through the replication that shrank it — against what AQA's own EPQ mark scheme actually rewards. Seven modules, seven lessons, evidence- and source-tiered throughout.

52 min of reading