The Kaiying Shared-Reality Bargaining Framework: A Minimal AI-Assisted Protocol for Locating a Reasonable Bargaining Space

Many disputes become difficult before the parties reach the real bargaining stage. The parties may disagree about facts, may interpret fairness in self-serving ways, or may sincerely regard their own preferred outcome as reasonable while viewing the other side as extreme. The Kaiying Shared-Reality Bargaining Framework (KSRB) is a deliberately simple protocol for using general-purpose AI as an external reference before human bargaining begins. The parties first construct a factual description that both accept. They then submit the same factual packet and the same prompt to independently chosen general-purpose AI models. The models are not told what either party wants, what either side’s reservation value is, or which party selected them. Each model returns a defensible bargaining range. In the simplest implementation for low-value disputes between non-professional parties, each side uses one ordinary, independently chosen, non-customized general-purpose AI model. In repeat-player or high-stakes settings, model choice should instead be jointly constrained. Each party then selects a meaningful human bargaining range inside the range supported by its model. If the submitted ranges overlap, the parties bargain within the overlap and AI exits. If they do not overlap, each party reveals the full range supported by its model. If those full ranges overlap, that overlap becomes the reference bargaining corridor. Only if the model-supported ranges themselves do not overlap does the process return to the shared facts, prompt, or model choice for diagnosis. Parties may optionally use several models. In that case, obviously isolated model outputs should not define the bargaining space; the party instead uses the central region supported by the majority of its chosen models. The framework is not an arbitrator, autonomous negotiator, or enforcement mechanism. Its purpose is narrower: agree on reality first, use independent AI judgments to locate a reasonable bargaining space, then let humans bargain.