ReasonSmith exists to elevate reasoning quality. These guidelines explain participation
standards, our scoring signals, and how we treat edge cases. The goal is not sterile civility—it
is intellectually honest, well-supported argumentation.
Core Principles
Good Faith First: Engage to understand, refine, or test ideas—not to score
rhetorical wins.
Evidence over Assertion: Claims that influence conclusions should cite or
label speculation clearly.
Precision over Volume: Better one well-structured paragraph than five
meandering ones.
Steelman Before Critique: Accurately restate what you’re responding to (or
ask clarifying questions).
Separate Person from Position: Challenge reasoning, data, assumptions—not
identity.
Scoring Dimensions (Experimental)
We are developing automated + community‑driven metrics. Provisional categories:
Good-Faith Structure Score: Presence of steelmanning, explicit assumptions,
delineated claims vs. support.
Fallacy Burden: Frequency + severity of detected logical fallacy patterns
(weighted: straw man > slippery slope > ad hominem aside, etc.).
Cultish Language Index: Density of manipulative rhetorical markers
(thought-terminating clichés, purity framing, loaded binaries).
Source Integrity: Ratio of claims with adequate citations + citation tier
quality.
Responsiveness: Whether replies address strongest prior point rather than
tangential fragments.
Scores guide moderation tooling and reputation—not instant punishments. False positives are
reviewed and models retrained.
Logical Fallacies & Cultish Markers
We surface flagged patterns to help you self-correct. A single minor fallacy or a moment of
rhetorical heat is not fatal; persistent patterns reduce trust weighting.
Tracked Fallacies (Examples):
Cultish / Manipulative Signals:
Transparency: You will be able to see what was flagged and why. Dispute
mechanisms coming soon.
Tone vs. Substance (Edge Cases)
We prioritize substantive epistemic value over tone purity. Mild rhetorical jabs that
accompany solid reasoning are treated differently from content-free insult chains.
Borderline Acceptable
Contains a harsh aside but delivers structured refutation.
"Flat Earth claims persist because people ignore direct observation. Calling the Earth flat is indefensible:
1. Hull-first disappearance of ships: consistent with curvature; incompatible with planar optics.
2. Great-circle navigation & polar routes: real-world flight plans minimize spherical surface distance; cannot be replicated on a Euclidean plane without distortion.
3. Continuous satellite telemetry & ISS 90-min orbits: require consistent gravitational curvature.
If you still deny this, you're willfully discarding multi-domain convergent evidence."
Not Acceptable
Insult replaces argument; no falsifiable content.
"Only idiots believe that. Educate yourself."
Harsh phrasing (+ one or two ad hominem adjectives) may pass if the post also: (a) steelmans,
(b) provides verifiable evidence, (c) advances the discussion. Pure insult without substance
will be down‑ranked or removed.
Guided Improvement
Flagged posts include a private “improve” panel suggesting rewrites.
Editing quickly after a flag reduces any negative weighting.
Repeat uncorrected patterns → temporary visibility dampening before moderation escalation.
Reporting & Appeals
User Reports are triaged by severity + model confidence.