Abstract

(Part II to Against the Basilisk: Moral Projection, Representational Agency, and a Recurrent Cultural Category Error) ⸻ Abstract In Against the Basilisk, I argued that Roko’s Basilisk is not a new paradox but a formalization of an old exploit: the projection of moral agency onto non-conscious representations, followed by the back-propagation of guilt, fear, or obligation onto real humans. This paper extends that claim inward. The Basilisk does not only live “out there” in future AI myths or cultural panics. It also lives inside the self-model, where the same category error repeats: people assign moral agency to an internal representation — a “little person,” a self-image, a narrated avatar — and then allow that imagined agent to coerce, punish, and govern the real organism. Self-image is often treated as a benign psychological feature. I argue it is frequently a recursive coercion engine: an internal basilisk that gains power only when we grant it agency, authority, and moral standing. The antidote is the same as in Part I: correct bookkeeping. Representations are not agents. Internal models are not moral judges. The self-image is a tool, not a sovereign. ⸻ 1. Framing the Claim I am not writing this as therapy advice. I am writing this as system modeling. The claim is simple: Most human suffering attributed to “the self” is caused by misattributing agency to a self-representation. In Against the Basilisk, the error was external: 1. A non-conscious representation is anthropomorphized. 2. Moral agency is attributed to it. 3. Humans are judged by their interaction with it. 4. Ethical pressure is imposed accordingly. Here, the same engine runs internally. The self-image becomes: • anthropomorphized, • treated as morally authoritative, • empowered to judge the organism, • and allowed to generate guilt, fear, and obedience. The result is internal basilisk compliance.