A portable method for testing what we believe about reality.
The Recursive Corrigibility Framework is a lifetime of challenging my own ideas, written down as a procedure that people and computational systems can both follow. It is published here to be used, taught, and stress-tested.
Wonder without surrender. Critical inquiry without protected conclusions.
Reasoning that can be checked, taught, and corrected.
Most arguments about consciousness, life, God and mind stall because nobody agrees on what would count against them. This site documents a method for fixing that, and shows the method at work.
Decades of self-correction
The framework grew out of a lifelong habit of treating every belief, my own first, as a claim that has to keep earning its place. Good-faith criticism is how it has evolved.
Built for any kind of reasoner
Each step is explicit enough to teach to a student or enforce on a language model. The aim is a method that works the same for a person, a machine, or an intelligence not yet built.
No protected conclusions
Nothing here is exempt from testing, including the framework. If an argument on this site fails, the useful response is to show where. That is how the work improves.
One loop, applied to every claim.
RCF does not tell you in advance what reality must be. It asks how much of a proposed answer has actually been earned.
- Claim
- Decompose
- Rivals
- Test
- Update
- Propagate
- Repeat ↻
What exactly is being claimed?
Separate observation, inference, causal explanation, mathematics, interpretation and ontology before letting evidence for one settle another.
What is the strongest serious rival?
A useful alternative keeps the most successful evidence while differing on the commitment actually in dispute.
What should come out differently?
Find the observations, interventions, proofs or counterfactuals that change the relative support of the competing explanations.
What actually failed?
Trace a failure to the assumptions that produced it, instead of abandoning the whole worldview or protecting it automatically.
The framework, audited by its own rules.
A method that demands evidence from everyone else has to meet the same standard. This is how RCF scores its own central claim.
RCF should have to earn its own complexity.
- Claim
- Following RCF makes reasoning more accurate and better calibrated.
- Type
- Empirical claim about outcomes
- Strongest rival
- A short checklist, or no framework at all, does just as well.
- Test
- The same held-out questions under four conditions: no framework, the full framework, a compressed version, and a mirror-image version biased the other way. Score calibration, errors caught and agreement with ground truth.
- Status
- PROVISIONAL until that comparison is run and published.
- Would change it
- Any step that does not change outcomes gets cut. If the full framework loses to the simpler versions, the simpler version wins.
For people, and for machines.
Learn the method
The framework is documented in full, from its methodology to the ontology it currently supports.
Run the method
The Interlocutor on this site answers from these articles and runs every reply through the framework. Each answer labels its claims by type, source and confidence, names the strongest rival, and is checked by code before you see it.
How the ideas connect.
Arguments tested in public, against serious people.
Much of the writing here engages working scientists and philosophers directly, with sources quoted in full so readers can check both sides.
Latest essays
- Responding to Joel Dietz (MIT)’s “77.6++%” Defense of Michael Levin’s Biological Causal Platonism and His Demonstrable Misrepresentations of My Actual Argument
Joel Dietz’s 77.6% audit of Michael Levin leaves nonphysical ingress at 0% earned support. A source-checked response to misattributions, seven claims, and executable tests. - Consciousness Naturalized: A Falsifiable Substrate-Agnostic Consciousness Theory
A falsifiable, substrate-agnostic research programme for consciousness, comparing organizational and thermodynamic hypotheses with rival theories through formal tests, empirical predictions, and explicit revision conditions. - Process Philosopher Matthew David Segall: Evidence and Scholarly Criticism IndexAn annotated research index of Matthew David Segall’s process philosophy, critical responses, peer-reviewed evidence, universal prehension, divine valuation, and the questions that remain open.
- Douglas Brash’s Chladni Plate Challenge to Michael Levin’s Biological PlatonismWhat Michael Levin’s planarian findings establish, what they do not, and why a flexible Platonic explanation can absorb the evidence without discriminating itself from causal alternatives.
- What Is Truth? What I Actually Affirm, and What My Disagreement with Matthew David Segall Is Really AboutWhat does truth require when an opponent publicly substitutes a position you did not hold? An evidence-led examination of Segall, representation, ontology, and the burden of argument.
- No Evidence, No Examples, No Named Errors, No Citations: How Michael Levin’s “Gut Feeling” Selective Dismissal of Criticism Shields His “Physicalism Is False” Claim From the Objection It Most Needs to AnswerMichael Levin’s dismissal of AI-assisted criticism offers no named logical errors or evidence. This analysis identifies the specific claims, fallacies, and unanswered objections.
- Michael Levin’s Xenobots: Are They Real? Do They Prove That “Physicalism Is False”?What xenobot experiments actually establish, what independent studies corroborate, and why novel biological behavior does not by itself refute physicalism.
- Matthew Segall on Whitehead: What Does Segall’s God Interpretation Actually Add?Matthew Segall’s distinctions about Whitehead’s God matter. Do they answer the specific objections about suffering, divine power, preservation, and evidential responsibility?
Bring the strongest objection you have.
The goal is not a system that shares my conclusions. The goal is a system, and an intellectual culture, that can be shown where its conclusions fail and can actually change when reality pushes back.
I am Nathan Sweet, an independent transdisciplinary researcher and writer developing methods for more evidence-responsive, corrigible reasoning across human and computational systems.