← All posts
meta#pre-registration#reproducibility#methodology#open-sciencearXiv:2604.23639

Why Every IRDME Experiment Is Pre-Registered

Science is only credible when predictions precede results. Here is the exact pre-registration protocol used for every Functional Proximity Law experiment — SHA-256 hashes, UTC timestamps, and public commits.

The Problem with Post-Hoc Analysis

Anyone who has worked with network data knows the temptation: run the analysis, see an interesting pattern, then write up a hypothesis that matches it. This is HARKing — Hypothesising After Results are Known. It produces confident-sounding claims backed by zero predictive power.

The Functional Proximity Law was discovered by noticing a pattern. To make it scientific, every subsequent test had to be pre-registered.

The Protocol

    Every IRDME experiment file contains a hypotheses block. Before running:
  • The hypothesis statement, expected direction, and minimum threshold are written in plain text.
  • The experiment file is SHA-256 hashed.
  • The hash and UTC timestamp are committed to github.com/vladi160/the-beginning.
  • Then the experiment runs.

The hash proves the hypothesis existed before the result. The public Git commit proves the timestamp. Neither can be faked retroactively.

Example

From the Human Brain Connectome experiment:

{
  "id": "h1",
  "type": "cross_layer_hub_correlation",
  "layer_a": "structural",
  "layer_b": "functional",
  "min_r": 0.3,
  "statement": "Hub centrality scores correlate positively across structural and functional layers (min r=0.3), consistent with the Functional Proximity Law."
}

Pre-registration hash: sha256:a3f9... (see repository).

Result: r = +0.703, p = 0.004. CONFIRMED.

The prediction was r > 0.3. The result is r = 0.703. The gap between prediction and result is evidence of genuine signal — not a post-hoc fit.

The Denial Cases

Of 23 domains, 7 were denied. Each denial was also pre-registered. The adversarial PTM network was constructed specifically to violate the law — and it did. That denial is as important as any confirmation: it proves the law has testable boundaries and is not unfalsifiable.

Why This Matters

    Pre-registration is not bureaucracy. It is the difference between:
  • "We found that hub scores correlate across similar layers" (post-hoc, weak)
  • "We predicted hub scores would correlate across similar layers before running the analysis, and they did" (predictive, strong)

All 23 IRDME experiments are in the second category. The pre-registration record is public, timestamped, and immutable.