Flow-by-Flow: Content-Judgment Bypass for Governing AI Output in High-Loss Domains

Flow-by-Flow：高风险领域 AI 输出治理的内容判断绕过机制

Abstract: Prior work showed that human-in-the-loop oversight becomes structurally untenable in high-loss domains when AI output velocity $V$ exceeds human cognitive capacity $C_{max}$. The operative constraint, however, is not $V$ alone but $V \times L$, where $L$ denotes per-item cognitive load. $L$ consists of triage, judgment, and response, which respond asymmetrically to AI capability improvement.

摘要： 先前的研究表明，当 AI 输出速度 $V$ 超过人类认知能力 $C_{max}$ 时，高风险领域中的“人在回路”（human-in-the-loop）监管在结构上将变得不可持续。然而，实际的制约因素并非仅是 $V$，而是 $V \times L$，其中 $L$ 代表单项任务的认知负荷。$L$ 由分类（triage）、判断（judgment）和响应（response）组成，它们对 AI 能力提升的反应是不对称的。

Triage cost does not decline as models become more capable, because semantic indeterminacy is inherent in general-purpose design. Response cost is invariant to accuracy improvements. Only judgment cost faces downward pressure, and this pressure often operates by inducing omission rather than genuine reduction. Capability improvement therefore restructures $L$ rather than reducing it.

随着模型能力的增强，分类成本并不会下降，因为语义不确定性是通用设计的固有属性。响应成本对于准确性的提升是不变的。只有判断成本面临下行压力，且这种压力往往是通过诱导遗漏而非真正的减少来发挥作用的。因此，能力的提升重构了 $L$，而非减少了它。

Governance mechanisms based on evaluating whether AI output is correct either delegate that evaluation to AI and inherit hallucination risk, or delegate it to humans and face the $V \times L$ ceiling. We propose Flow-by-Flow, a governance paradigm that controls supervisory load without evaluating content. A cognitive cost score based on formal, countable features imposes nonlinear costs on high-volume production, while an institutional capacity cap keeps processing volume within $C_{max}$.

基于评估 AI 输出是否正确的治理机制，要么将评估委托给 AI 并继承幻觉风险，要么委托给人类并面临 $V \times L$ 的上限。我们提出了 Flow-by-Flow，这是一种无需评估内容即可控制监管负荷的治理范式。基于形式化、可计数特征的认知成本评分对大批量生产施加非线性成本，同时通过制度性的容量上限将处理量保持在 $C_{max}$ 以内。

We derive four design invariants for any content-judgment-bypass exceedance pathway: no content judgment, no scalable consumption of examiner capacity, identity-bound per-application friction, and no batch clearance. One reference implementation is discussed to show that these invariants are jointly satisfiable, while its practical difficulties are explicitly acknowledged. An illustrative Monte Carlo analysis across 1,000 parameter draws suggests that composite multi-metric flow control outperforms supervision reinforcement alone in 90.8% of trials.

我们为任何“内容判断绕过”的超额路径推导出了四个设计不变量：无内容判断、无审查者容量的可扩展消耗、与身份绑定的单次申请摩擦力，以及无批量清算。文中讨论了一个参考实现，以证明这些不变量可以共同满足，同时也明确承认了其实际操作中的困难。一项跨越 1,000 个参数抽取的蒙特卡洛模拟分析表明，在 90.8% 的试验中，复合多指标流量控制的表现优于单纯的监管强化。