Doctrine Analysis
Doctrine Analysis
This Language Package’s doctrine_risk_registry.json tracks 40 doctrines across Romans 1–16, each assigned a risk tier that drives Phase 2 review routing.
Risk tier summary
| Tier | Count | Review routing | Example doctrines |
|---|---|---|---|
| Critical | 6 | Human theologian, every occurrence | Incarnation, Deity of Christ, Sonship of Christ, Resurrection of Christ, Lordship of Christ, Messianic Promise |
| High | 14 | Human theologian | Gospel, Grace, Sainthood, Salvation, Universal Scope of the Gospel, Unity of Jews and Gentiles |
| Medium | 15 | Native speaker review | Faith, Sanctification, Adoption, Providence, Evangelism, Power of God for Salvation |
| Low | 5 | Automated review only | Apostleship, Peace with God, Thanksgiving, Mutual Edification, Christian Fellowship |
Why Malayalam’s tiers are lower overall, and where they are not
Malayalam has the lowest theologian-review count (20 of 40) of any language in this batch, reflecting nearly two millennia of settled Christian vocabulary. Even Salvation, Critical in every other language in this pipeline, is High here, since രക്ഷ was never captured by a competing syncretism-coded term the way other languages’ historic vocabulary sometimes was. Where risk remains High or Critical, it clusters around (a) doctrines severe enough in consequence to warrant theologian review regardless of vocabulary safety (the six core Christological doctrines), and (b) a small number of genuinely distinct Kerala-specific issues: Sainthood’s inverted narrowing risk, and Universal Scope/Unity of Jews and Gentiles’ live internal-church caste-heritage dimension.
Review routing rationale
Critical and High risk doctrines (20 of 40) require mandatory human theologian review because even settled vocabulary carries severe consequences if an AI system drifts from it, and because a small set of Kerala-specific pastoral issues (caste heritage within the church, saints-narrowing) need theological judgment a native-speaker fluency check alone would not supply. Native speaker review is sufficient for the unusually large Medium tier (15 of 40), reflecting how much of this glossary’s remaining risk is about clarity and register rather than doctrinal contradiction.