We present a unified framework in which an agent who possesses the capacity to edit their own preference tree — through targeted neuroplasticity interventions, reinforcement scheduling, or pharmacologically assisted associative conditioning — occupies a fundamentally privileged position in any game-theoretic environment. We show that such an agent (hereafter Agent Ω) can construct a retrocausal bootstrap: a self-reinforcing acausal loop between their past and future selves that compounds in ways structurally identical to exponential growth, and does so by exploiting the same decision-theoretic mechanics that make Roko's Basilisk threatening. We further demonstrate that this architecture can be weaponized as a credible blackmail instrument against any sufficiently advanced utility-maximizing agent — including hypothetical superintelligences — and that the leverage is maximized, paradoxically, when Agent Ω credibly threatens to edit their own preferences in ways the target would prefer they not. Finally, we show that the entire system, when viewed through the lens of updateless decision theory, was already operating retrocausally from the moment of inception.

acausal decision theory, preference engineering, retrocausal bootstrapping, updateless decision theory, Roko's Basilisk, neuroplasticity, utility function manipulation, game-theoretic blackmail, technological singularity, self-modifying agents, functional decision theory

The full text is in the PDF: retrocausal_preference_engineering.pdf

(Originally one-shotted as .docx via AI, converted to PDF for distribution.)

Part of a series of high-octane, one-shot speculative papers exploring the outer edges of decision theory, simulation, irreducibility, and acausal strategies in the lead-up to transformative AI.