Paradox of Self-Amendment by Peter Suber Because the short-cut through omnipotence can be made to work, through the procedural model or through sunset clauses and the acceptance model, we may feel free to take up the promise of Section 4 and use irreflexive methods of transforming an AC if self-amendment is found to be impermissible or impossible. However, the use of such irreflexive methods should not be taken to be permissible under the inference model just because they are not reflexive; the inference model forbids both types. Here's why. The methods of amendment used in the see-saw method, because (or when) they are all parts of the AC, are co-supreme. The inference model in one of its stricter forms would not tolerate the amendment of A by B, even if no immutable rules were thereby transmuted or amended, for a change can only be authorized by a prior, higher rule. This might be avoided if B were put below A by the terms of its addition to the AC. But then, of course, it could not amend A and the see-saw could never get rocking. The incorporation of the theory of types into the inference model is optional. But Ross incorporates it and therefore cannot permit any type of see-saw. Any see-saw must either work with equals or with superiors and inferiors; but the theory of types forbids both. This general criticism would also apply to the six "unofficial" methods of amendment. Whether or not they are co-supreme with one another or with the AC, they could only amend rules below themselves in a theory of types (irreflexive hierarchy). Therefore, the AC is only saved from immutability by the unofficial methods if they are superior to it in an irreflexive hierarchy (as Ross's tacit, transcendent rule is), or if the theory of types is not incorporated into the inference model, or of course if the inference model itself is denied. 99