Methodology
This page explains how SelfControlAtlas produces content: the four-layer content system, the de-identification rules, the evidence standards, and the quality checks applied before publication. It is the reference for anyone who wants to understand how much to trust a given page.
Last updated: 2026-10-23 · Last reviewed: 2026-10-23
1. The four-layer content system
Content is organized into four layers, each with a distinct epistemic status. The layers are kept separate so that a research finding is never confused with a community observation, and a single person's story is never presented as a universal fact.
Layer 1 — Experiences
De-identified first-person accounts of recovery. An experience is one person's report, not a fact about everyone. It is structured (trigger, strategy, outcome, timeline) so it can be compared with others, but it is never generalized beyond itself.
Layer 2 — Patterns
Recurring observations aggregated across many experiences. A pattern says "this appears frequently across accounts," not "this is universally true." Patterns are supported by counts or clear recurrence, never by a single anecdote.
Layer 3 — Problems
Common questions (e.g., "What is a flatline?", "Should I track my streak?") answered by synthesizing patterns and evidence. Answers state what is known, what is uncertain, and where to get professional help.
Layer 4 — Evidence
Research summaries tied to published sources and graded by the evidence tiers below. Evidence is kept separate from community data so that a research finding is never confused with a community observation.
2. Why four layers?
The four-layer structure addresses two common errors in this space:
- Treating anecdotes as evidence. An experience is one person's report. Without a separate "evidence" layer, it is easy to mistake a vivid story for a proven effect.
- Treating findings as guarantees. A research finding is an average or an association, not a promise about any individual. Without a separate "patterns" layer, it is easy to overstate what a study means for a given person.
By keeping the layers separate, the site can be both honest about its sources and useful to readers.
3. De-identification rules
Before any community-derived content is published, editors apply these rules:
- Remove usernames, real names, and aliases.
- Remove contact details (email, phone, social handles).
- Generalize or remove specific locations.
- Remove dates that could identify a person or event.
- Remove any detail that, in combination, could re-identify an individual.
The objective is that no reader can identify the person behind an experience, and no personal information about community members is published.
4. Evidence standards and tiers
Research summaries are graded using evidence tiers, from strongest to weakest:
- Official classifications — e.g., WHO/ICD-11 definitions of compulsive sexual behavior disorder.
- Systematic reviews and meta-analyses — the highest-quality syntheses of primary research.
- Primary studies — individual peer-reviewed studies, with their limitations noted.
- Expert opinion and case reports — labeled as such and given less weight.
- Community consensus — useful for describing shared experience, but never presented as scientific evidence.
Editors state the tier a claim rests on, and say so explicitly when evidence is mixed or absent.
5. Separation of evidence and experience
A core rule: evidence and experience are never merged into a single undifferentiated claim. A research summary cites sources; a pattern cites recurrence across accounts. The site does not dress a community anecdote up as a study, and does not dress a study up as a universal truth.
This separation is the single most important methodological commitment, because it prevents the two most common errors in this space: treating anecdotes as evidence, and treating findings as guarantees.
6. Quality checks before publication
Before a page is published or updated, editors check:
- The content layer is stated on the page.
- De-identification rules have been applied (for community-derived content).
- Sources are cited where a factual or research claim is made.
- Uncertainty is labeled where evidence is weak.
- Crisis and professional-help signposts are present where a page touches on distress or self-harm.
- The page is consistent with the Disclaimer and Terms of Use.
7. How problems are answered
Problem pages follow a consistent structure:
- What is being asked — restated plainly.
- What the evidence says — with the tier stated.
- What the community reports — clearly separated from evidence.
- What is uncertain — stated honestly.
- When to seek professional help — signposted.
This structure prevents a problem page from becoming disguised medical advice.
8. How the knowledge graph is built
The knowledge graph links entities across the site so readers can trace connections:
- A trigger page connects to the strategies that address it.
- A problem page connects to the evidence that bears on it.
- An experience connects to the patterns it illustrates.
Links are directional and labeled, so readers move from "what is this?" to "what does the evidence say?" to "what have others experienced?" without losing track of which is which.
9. Corrections and versioning
The current version of any page is authoritative. When an error is corrected:
- The correction is made on the page.
- Material corrections may be noted.
- Readers can report errors via the Contact page.
10. Sourcing practice
When the site cites research, it prefers primary and secondary sources that readers can locate:
- Official classifications and clinical references (e.g., WHO/ICD-11).
- Peer-reviewed systematic reviews and meta-analyses.
- Individual peer-reviewed studies, with limitations noted.
Where a source is behind a paywall or not publicly linkable, the site still names the source so readers can find it. The site does not cite sources it cannot verify, and does not fabricate citations.
11. Honesty about uncertainty
Uncertainty is a normal part of research on problematic pornography use. The site labels it rather than hiding it. Where evidence is mixed, weak, or absent, the page says so in plain terms, rather than presenting a confident-sounding but unsupported conclusion.
12. Limitations
- Research summaries reflect sources available at the time of writing and may be superseded.
- Community patterns are observational, not experimental.
- No content on this site constitutes medical, psychological, or therapeutic advice.
See the Disclaimer for the full statement of limitations.
Last updated: 2026-10-23.