Trust: rank sources and authors by framing scores

This page aggregates stored article history into ranked source or author profiles. Grouping is by source by default (or when any non-author sort is selected), and by author when Trust (author) is selected. Ranking uses Accuracy, Sensationalism, Facts, or Opinion when those sorts are active; otherwise it uses the composite trust score described below. Lean filters (All / Right / Left) limit which historical articles feed the averages.

How composite trust is calculated. For each source or author we average four metrics across their stored articles: fact ratio, opinion ratio, accuracy score, and sensationalism score (each 0–100 from the same heuristics used on the live feed). Composite trust is then a weighted blend that rewards checkable framing:

• 35% accuracy (how specific and attributed claims look)
• 30% fact ratio (share of fact-leaning language)
• 20% inverted sensationalism — 100 − sensationalism so calmer presentation scores higher
• 15% inverted opinion — 100 − opinion ratio so less evaluative language scores higher

The result is rounded to a 0–100 trust score. Higher means more checkable, less loaded framing in the text we have—not a moral rating or a guarantee of truth. Metric-only sorts (Accuracy, Facts, and so on) skip the blend and rank entities by that average alone, still respecting sort direction (↑ / ↓). Browse individual rows on History or live coverage on the live dashboard.