THE BETTER FORGERY: Fidelity of Form, Responsive Verification, and When Skill Makes Deception More Harmful

Abstract

Abstract Most people assume that skill and value travel together: a better copy is a better thing. For forgeries the relation runs the other way. A clumsy forger is caught within a season, and the loss is one picture. Van Meegeren’s forgeries were bought by museums, entered the catalogues and shaped the scholarship on Vermeer. When they fell, a decade of attributions fell with them. This paper states, proves and bounds the law behind that reversal. The Competence Theorem (T1) holds that, at a fixed level of verification, the damage a forgery does rises with the fidelity of its form, because fidelity delays detection and reliance accumulates during the delay. The theorem holds only where the substance is missing. A copy that also reproduces the substance is not a forgery, and it does no damage of this kind, which explains why a good generic drug is harmless and a perfect fake drug is lethal. Verification, however, can respond. When verifiers work harder against better forgeries, the relation can reverse. The Concealment–Response Condition (T2) states exactly when it does: damage rises with fidelity if, and only if, the concealment that fidelity buys grows faster than the verification it provokes. The condition holds for any path along which reliance accumulates, whether linear, saturating or networked. Where the reliance placed on a forgery itself grows with its fidelity, the condition becomes easier to satisfy (T3). Where verification probes the substance rather than the form, it can defeat forgery without detecting it, by forcing the forger to supply the substance (T4), everywhere except in character, whose substance is authorship and cannot be forced. The law also has an upper boundary: in domains where excessive perfection is itself a signal, detection rises again at the top of the fidelity scale. The paper conjectures that verification does not respond at all below the frequency at which forgery becomes worth checking for (C1). It derives from the condition an ordering of domains (C2): · in currency, verification responds and the theorem can fail; · in art and in human character, verification barely responds, and the theorem holds at full strength; · fraud and the fluent agreement of AI systems lie between. Where verification relaxes as the form improves, damage rises fastest of all (T5). Where the verifier is itself a score, that is the normal condition, and the damage of a well-performed virtue is at its largest (C3). Because verification of character detects little and wrongs the innocent, a society’s protection against the accomplished hypocrite lies not in surveillance but in the formation of people whose good conduct is their own (C4). The paper concedes priority to the economics of counterfeit money, where quality and verification have been modelled together, and to the theory of costly signals, which asks when signals are credible where this paper asks how much damage an incredible one does before it is found out; and it states what is its own: the clause on substance, the general condition, and the predictions across domains. It registers two tests, on occupational fraud and on art forgery, neither yet run. Keywords: forgery; counterfeiting; verification; deception; fidelity; reliance; Competence Theorem; art forgery; occupational fraud; hypocrisy; AI sycophancy; social credit; Philosophy of Virtues. Highlights · At a fixed level of verification, the damage of a forgery rises with the fidelity of its form: a better forger does more harm than a worse one. · Verification that tests the substance rather than the form can defeat forgery without catching it, by forcing the forger to supply what was missing. · The law holds only where the substance is missing; a copy that reproduces the substance is not a forgery and does no such damage. · When verification responds, damage rises with fidelity if, and only if, the concealment fidelity buys outgrows the verification it provokes. · The theorem holds most strongly where verification cannot respond: in art, in human character, and wherever a score is the only check. · The fluent agreement of an AI system and the well-performed virtue of a scored citizen are the contemporary forms of the better forgery. · Against the accomplished hypocrite, surveillance is the wrong defence: it detects little and wrongs the innocent. The defence is character formed in freedom. Relevance Every society depends on forms it cannot check one by one: signatures, certificates, references, reputations, and increasingly the fluent answers of machines. The question of when a better imitation is more dangerous than a worse one is therefore not a question about art. It is a question about what a society can trust, and about how its institutions of verification must be built if trust is to remain safe. The answer this paper gives has a direct bearing on human virtue. The most dangerous hypocrite is not the clumsy one but the accomplished one. Character is the domain in which verification responds least to the quality of the performance. That is why the real thing matters more than the performance, and why no amount of polish can stand in for virtue itself. The answer also bears on two contemporary forms of the same structure: - AI systems. A system that agrees fluently is harder to catch than one that agrees clumsily, and its evaluations do not grow more demanding as it grows more fluent. - Scoring states. China’s social credit system rewards the well-performed form of a virtue, and its verifier is the score itself. Such a regime is built to maximise exactly the damage this paper describes.