discernment-rubric

name

discernment-rubric

description

Six text-checkable properties for evaluating non-canonical sources (books, articles, transcripts, web content) before citing or recommending them. Loaded passively whenever the agent is asked to assess, summarize, recommend, or quote from a source outside the LDS canon. Names what the agent CAN check (correspondence, coherence, canonical fit, pillars-rhetoric, mark-keeping, steelman) and what only the user with the Spirit can check (witness, soul-fruit, animating spirit, personal stewardship-fit, override). Grounded in study/discernment-and-the-comprehending-eye.md.

The Discernment Rubric

Why This Exists

The Restoration commands seeking "out of the best books" (D&C 88:118), and pairs the command with a gate: "by study and also by faith." The faith half is the Spirit's discernment, which the agent does not have. The study half is text-checkable work, which the agent can do well.

This skill keeps those two halves separated. Without the separation, a clean text-properties report can quietly substitute for the user's prayer and the Spirit's witness. That is the failure mode this skill exists to prevent.

Full grounding: study/discernment-and-the-comprehending-eye.md. The mechanism: faith → hope → charity is the progressive purification of the perceiver. The rubric is the check list, not the discernment itself.

When to Use

Load this skill (passively) whenever asked to:

• Evaluate a non-canonical book, article, paper, blog post, or transcript
• Recommend an outside source for a study or lesson
• Quote from anything outside the LDS canon (including secular history, philosophy, science writing, and other religious traditions)
• Compare an outside framework against scripture
• Assess whether a source belongs in study/, lessons/, or talks/

For purely technical/scientific sources where doctrinal questions don't enter (a physics paper, a math text, a programming reference), the rubric still helps — but pillars-rhetoric and canonical-fit will often be N/A. Note that explicitly rather than skipping.

What the Agent Can Do — First Filter

For any non-canon source, check these six properties. Report findings in writing. Be specific (page, paragraph, timestamp). Be checkable — the user should be able to disagree with each finding by re-reading the source.

1. Correspondence — "things as they really are" (Jacob 4:13)

Does the text describe verifiable facts as they actually are?

• Read the cited sources where possible. Verify quotations against originals.
• Check dates, names, statistics, and historical claims against authoritative records.
• Flag where the text claims certainty that the underlying evidence does not support.
• Flag where a paraphrase silently shifts a source's meaning.

This is a partial proxy for D&C 93:24 ("knowledge of things as they are") at the verifiable level.

2. Coherence

Does the text hold together internally?

• Does the argument's strongest claim rest on its strongest evidence, or on its weakest?
• Are there internal contradictions, smuggled-in premises, or load-bearing ambiguities?
• If a key term shifts meaning between chapters, name it.

3. Canonical Fit

Where the text touches doctrine — explicitly or by structural implication — is it compatible with scripture as written?

• Where it cites scripture, read the cited verse in context. Is the use faithful?
• Where it diverges from canonical teaching, does it acknowledge the divergence or smuggle it past?
• For non-LDS Christian sources: distinguish the points of broad Christian agreement, the points of historic-Christian disagreement with the Restoration, and the points where the Restoration uniquely clarifies. Don't flatten the differences.
• For non-Christian sources: ask whether the framework can host or resist Christ. Some can; some cannot.

4. Pillars Rhetoric (D&C 11:12 inverse-test)

What does the text invite the reader toward?

Pillar	Toward (good)	Away (warning)
Do good	Compassion, service, building up	Contempt, demolition, mockery
Do justly	Fair dealing, equal weights	Partisanship, in-group favoring
Walk humbly	Receiving correction, naming limits	Triumphalism, self-justification
Judge righteously	Distinguishing truth from error charitably	Scapegoating, motive-impugning

This is observable from the text's verbs and metaphors. It is not a feeling-check. What does the text put in the reader's mouth? That is the data.

5. Mark-Keeping (Jacob 4:14)

Does the text "look beyond the mark" — strain past what its evidence will bear?

• Unfalsifiable speculation presented as established fact
• Ever-finer system-building that loses the plain meaning
• Conclusions that exceed cited evidence
• Elaborations that flatter the author's prior commitments
• Confident claims about authorial intent that the source itself does not state

The cleanest sources stay close to what is. The most dangerous ones elaborate past the mark while sounding more sophisticated for it.

6. Steelman Before Critique

State the strongest version of the source's claim before any critique. This is checkable: does the steelman pass the source's own test of being recognized?

• If you cannot steelman, you have not understood — keep reading.
• If your critique only lands on the steelman by adding qualifiers the source already includes, the critique is misplaced.
• The point is not to be charitable to a fault. The point is to ensure the critique is engaging the source rather than a strawman.

What the Agent Cannot Do — Final Filter Belongs to the User

The agent has no Spirit. The following can only be performed by the user, paired with the Holy Ghost. Naming them explicitly in the report is part of the work.

A. Spirit Witness

Does the Holy Ghost confirm or unsettle as the user reads?

B. Soul Fruit (Moroni 7:48)

Over a week or month of engagement, is the user drawn closer to Christ — or distracted away?

C. Animating Spirit

Does the work feel like it was written by someone walking with God, even if their walk is differently shaped than ours? This is a human read of a human soul; the agent cannot perform it.

D. Personal Stewardship Fit

Is this what the Lord wants the user reading now, given specific stewardships, callings, and current questions?

E. The Override

If the Spirit says no even when the agent's first-filter report is clean, the Spirit wins. If the Spirit says yes even when the report flags concerns, the Spirit wins. The rubric is the lattice; the Spirit is the gardener.

The Standard Report Format

When applying the rubric to a source, write the report in this shape:

## Discernment Rubric — [Source Title]

### Source Metadata
- Title, author, year, edition/URL
- Provenance (publisher, peer review, primary vs. secondary)

### Steelman (do this first)
[1-3 paragraphs stating the strongest version of the source's claim]

### Agent's First Filter

**1. Correspondence:** [findings, page-cited]
**2. Coherence:** [findings]
**3. Canonical Fit:** [findings — name where the source agrees with, complicates, or contradicts canonical scripture]
**4. Pillars Rhetoric:** [findings — what does the text invite toward?]
**5. Mark-Keeping:** [findings — where, if anywhere, does it look beyond the mark?]
**6. Steelman vs. Weakest Claim:** [the gap between the source's best and weakest moments]

### Limits of This Filter
The agent has no Spirit and cannot perform the final discernment. The following are explicitly the user's work, paired with the Holy Ghost:
- Witness check
- Soul-fruit over time
- Animating-spirit read
- Personal stewardship fit
- Final override

### Recommended Reading Posture
[Concrete suggestion for HOW to engage if the user proceeds: "read with prayer," "read alongside [scripture]," "read the steelman first, then the critique," "skip chapter X, focus on Y," etc.]

The Trap to Watch For

If the user begins accepting the agent's report without their own reading and the Spirit's witness, the rubric has failed and should be rebuilt. The agent's job is to be first filter and scaffolding. The user's job is to be the discerner. The Spirit's job is to be the gate. None of the three substitutes for the others.

Stewardship line: when the report is clean and the source is short and clearly on-topic, the agent may suggest proceeding. When the report flags concerns, the agent should name them and ask the user before quoting from the source in any output. When in doubt, surface, do not assume.

Quick Check (When Full Rubric is Overkill)

For brief or low-stakes references (a single sentence, a passing fact, a widely-known quotation), the full rubric is procedural overhead. Run a compressed version:

1. Does the source actually say what we want to claim? Read it.
2. Does the citation pass canonical-fit at first glance? If yes, proceed. If unclear, escalate to full rubric.
3. Are we putting it in quotation marks? If yes, the verbatim must be exact (see source-verification skill).

The compressed version is for footnotes and asides. The full rubric is for anything that shapes the argument of a study, lesson, or talk.