Common Writing Principles

Common Writing Principles

These principles apply to all content types across the wiki. Domain-specific style guides (risk, response, models, ATM) build on these foundations. When scoring pages, the objectivity rating dimension measures adherence to these principles.

1. Epistemic Honesty

The wiki makes many quantitative claims that are uncertain. Present them honestly.

Ranges Over Point Estimates

Always use ranges for uncertain quantities. A single number implies false precision.

Label Uncertainty Explicitly

Use hedging language proportional to actual uncertainty:

• High confidence (sourced data): "According to IPS, 36% of deceased pledgers met the 50% threshold"
• Moderate confidence (reasonable inference): "This suggests roughly $25-70B in capital is at stake"
• Low confidence (rough estimate): "A portfolio might cost on the order of $2-8M"
• Speculation: "It's plausible that backfire risk could make this net negative"

Words and Phrases

Structural Patterns

• Pair optimistic calculations with critical analysis. If you show a naive calculation (e.g., "even a 1pp shift is worth $250M"), immediately explain why the naive math is probably wrong (selection bias, hidden costs, backfire risk, etc.).
• Include "Key Uncertainties" sections on analysis pages. Name the 3-5 things you're most unsure about and explain how they'd change conclusions.
• Show deflators explicitly. When presenting cost-effectiveness estimates, show both the naive calculation and the deflated version with a table explaining each adjustment.
• No false precision. Use "≈$1M" not "$1,234,567" when the number is a rough estimate. Round to the level of actual knowledge.

2. Language Neutrality

The wiki should be readable by someone with no prior exposure to the effective altruism community. Avoid insider language that creates in-group/out-group framing.

The Core Rule

Describe things by what they are, not which community they belong to.

When "EA" Is Fine

• As a proper noun referring to the movement: "effective altruism" (spelled out on first use)
• In entity names: "EA Forum", "Centre for Effective Altruism"
• In direct quotes
• On pages specifically about the EA movement (e.g., the field-building section)

Avoid Defining by Negation

"Non-EA founders" tells the reader nothing about who these people are. "Founders without documented connections to the effective giving community" is longer but actually informative. Or just say "the other 5 founders" if context makes it clear.

3. Analytical Over Prescriptive

The wiki analyzes; it doesn't advocate. Present tradeoffs and let the reader decide.

Voice

Consider the Subject's Perspective

When writing about interventions that target real people or organizations:

• Ask: how would the subject feel reading this? If a page about "pressuring founders to donate" would read as hostile surveillance, reframe it.
• Distinguish collaborative from adversarial interventions. Tax planning help is collaborative; public shaming is adversarial. Name this distinction explicitly.
• Acknowledge legitimate reasons for disagreement. If someone chose a non-binding pledge deliberately, don't assume it was an oversight.

Red-Team Your Own Conclusions

Analysis pages should include honest counter-arguments:

• What if the intervention backfires?
• What if selection bias makes the expected value near zero?
• What legitimate reasons might someone have for disagreeing?
• Are you writing from the conclusion backwards, or following the evidence?

Anti-Patterns

The "False Certainty" Trap

Symptom: Table headers say "True Cost" and "Realistic EV". Estimates presented as single numbers without ranges. No acknowledgment of uncertainty.

Fix: Add "Est." prefixes, use ranges, include a "Why These Numbers Might Be Wrong" section.

The "Insider Language" Trap

Symptom: Page uses "EA" as an adjective throughout. Assumes reader knows what GiveWell, GWWC, and LTFF are. Defines outsiders as "non-EA."

Fix: Spell out abbreviations, name specific organizations, describe things by their properties not their community affiliation.

The "Prescriptive Analysis" Trap

Symptom: Page framed as analysis but actually advocates for a specific position. Uses "we should" and "the right approach."

Fix: Reframe as "this analysis suggests" and present counterarguments with equal rigor.

The "Self-Importance" Trap

Symptom: Page describes itself as the "canonical source," "definitive guide," "most comprehensive analysis," "investment-grade," or "rigorous" analysis. Uses superlatives about its own coverage ("most thoroughly documented," "first analysis to..."). Frontmatter descriptions or note blocks claim the page is uniquely authoritative.

Why it matters: Readers interpret these claims as the wiki asserting global authority—as if no better source exists anywhere. Even if intended to describe scope within the wiki, it reads as self-promotional marketing copy. It undermines the epistemic humility the wiki aims for.

Fix: Describe what the page covers, not how good it is. Let quality speak for itself.

Note: Superlatives about external sources are fine when attributed ("FLI calls their index 'the most comprehensive'"). The issue is claiming superlatives about the wiki's own content.

The "One-Sided Framing" Trap

Symptom: Lists benefits of an intervention without honest concerns. Doesn't red-team its own conclusions. Treats the intervention as obviously good.

Fix: Add "Honest concern" notes to each intervention. Include a section on why the whole premise might be wrong.

Relationship to Scoring

These principles are measured by the objectivity rating dimension (0-10):

• 1-2: Heavy insider jargon, false certainty, prescriptive advocacy throughout
• 3-4: Some insider language; estimates without uncertainty ranges; one-sided framing
• 5-6: Mostly neutral language; some uncertainty acknowledgment; mostly analytical
• 7+: Fully accessible to outsiders; all estimates properly hedged with ranges and caveats; analytical throughout; honest counter-arguments included

See the Rating System for how objectivity fits into derived quality scores.