Kaiwai Arbitrage Dynamics (KAD): A Vacancy-Based Creativity Method for Cross-Community Value Translation — Reference Implementation, Design Lessons, and the Evaluator Problem

This record contains the English and Japanese versions of the paper (Furuta_2026_KAD_en.pdf / Furuta_2026_KAD_ja.pdf). 本レコードには英語版と日本語版のPDFを同梱しています。 The value of information is relative to the community that receives it. This paper proposes Kaiwai Arbitrage Dynamics (KAD), a method for systematically discovering and exploiting the asymmetry of information value across social-media communities (kaiwai), operationally defining translation as a two-stage mapping from coreValue to targetValue. We report a six-stage reference implementation combining LLMs with social-media search; a negative result (LLM self-critique without real-data grounding systematically degrades quality); evaluations showing that a two-stage rewrite converges posts toward the target community's actual writing style (n=81, p=1.2e-6) while an additional rewrite significantly diverges them (n=27, p=0.006); and a criterion-dependent reversal of blind human preference (13/14 pairs flipping), from which we derive an "evaluation ladder" whose top rung is the community's own measured response. Second paper in the Vacancy-Based Creativity Methods series (#1: QVP, DOI: 10.5281/zenodo.20627544).