Research methodology
How this platform separates observation, interpretation and judgment.
Version 1.0, as of Sep 22, 2026. No expert review: this document was written for this platform, and nobody has gone through it independently.
Statement types
Every substantive block is marked with the kind of statement it makes. That prevents the most common error in DeFi reporting: turning a metric straight into a risk verdict.
- Fact
- A checkable observation or definition. Must be verifiable; a source is required as soon as it goes beyond a definition.
- Data
- A measured value with a named origin. Requires a source.
- Interpretation
- A reading of the observation. Can be wrong without the observation being wrong.
- Assumption
- An assumption that must hold for the statement to carry.
- Risk
- A named risk, not its probability of occurring.
- Uncertainty
- An explicitly open question. Not a shortcoming, a finding.
- Assessment
- A judgment a person is accountable for. Never an audit, never a guarantee.
Source hierarchy
Not every source carries the same weight. The tier follows from the source type and is not assigned freely per entry.
- Tier 1Primary data (on-chain)
- Tier 2Official protocol documentation
- Tier 3Regulator / legislator
- Tier 3Central bank
- Tier 4Peer-reviewed research
- Tier 5Working paper
- Tier 5Professional research
- Tier 6Secondary media
- Tier 7Community sources
Why retrieval and spaced review
The Academy asks retrieval questions after every lesson and places completed lessons on a review ladder of 1, 3, 7, 14, 30 and 60 days. That is not a design whim: spacing and retrieval practice show robust learning advantages over rereading in the research literature, with feedback playing an important role. What that evidences is the effectiveness of the method — not the factual accuracy of any particular lesson here, and the settings studied are mostly schools and classrooms. Both limits are stated on the records themselves.
Sources: EVD-2026-0002, EVD-2026-0003
Five methods this Academy applies, and what the research says about each:
1. Retrieval practice
Every lesson ends with retrieval questions whose feedback names what an answer missed. Practice testing is one of the two techniques with the highest utility rating in Dunlosky et al.'s review.
Sources: EVD-2026-0003, EVD-2026-0021
2. Distributed practice
Completed lessons sit on a review ladder of 1, 3, 7, 14, 30 and 60 days that follows actual retrieval performance. Distributed practice is the second technique with a high utility rating.
Sources: EVD-2026-0002, EVD-2026-0021
3. Worked example, then variation
Every formula sits next to a worked example, and where the figures are movable, the interactive model starts on exactly those figures. Studying examples instead of solving problems measurably reduces the load on novices — the worked-example effect.
Sources: EVD-2026-0024
4. Text and diagram side by side
Models show the curve the input sits on, and the readout stands next to the control that moves it. Words plus a matching graphic, signaling the structure and leaving out decorative detail are the principles with documented effect; the effects are small to moderate and shrink the closer studies get to real classrooms.
Sources: EVD-2026-0022, EVD-2026-0023
5. Estimate first, then check
A question the reader should commit to comes before the controls. Self-explanation and elaborative interrogation are given a moderate utility rating in the review — documented, but on narrower evidence than retrieval and spacing.
Sources: EVD-2026-0021, EVD-2026-0023
All four sources study learners in schools, universities or laboratories with academic material. None establishes that this Academy implements the methods correctly or that its content is factually right — both questions are noted on the records themselves.
Analyst notebook
The structure for recording your own analysis. Eleven fields, in this order:
- Research question
- Hypothesis
- Data sources
- Observations
- Calculations
- Interpretation
- Counterargument
- Risk
- Uncertainty
- Conclusion
- Evidence
This platform stores no user text. It has neither user accounts nor a database, and its privacy policy says exactly that. The notebook is therefore a structure for you to keep yourself — server-side storage would be a product decision with data-protection consequences, not merely an implementation question.