PE 输出从 scalar 升级为 distribution

Blind spot (现状盲点)

当前 [`packages/vz-cognition/`](../../packages/vz-cognition/) 的 prediction-error owner 输出是否包含**价值分布**信息（如 quantile / categorical distribution / 分布矩），还是只输出标量 surprise？需要核查。如果只是 scalar，那么我们错失了**"分布塌缩 = 病理状态"**这一根本健康信号——R7 self-track 的"心理健康"将只能依赖行为代理指标（如对话频率、响应长度），而非有原则的内禀信号。

Adoptable suggestions (可落地动作)

1. 1.在 [`docs/specs/prediction-error-loop.md`](../specs/prediction-error-loop.md) 加入"PE distributional readout"小节：定义最简实现——PE owner 的 snapshot 输出契约从 `(mean_surprise, ...)` 扩展为 `(mean, distribution_summary)`，其中 `distribution_summary` 至少包含 quantile（如 IQR）+ entropy + asymmetry（skew）。PROPOSED

   Not a runnable A/B candidate — evaluated by the path above, not ablation.

2. 2.在 [`docs/specs/lifeform-vitals.md`](../specs/lifeform-vitals.md) 把"价值分布方差/熵/asymmetry 漂移"列为新增 vitals slot，作为 lifeform health 的一级指标。PROPOSED

   Not a runnable A/B candidate — evaluated by the path above, not ablation.

3. 3.评估证据先行：在 evaluation 加一族"分布编码完整性"指标（不进入控制主链），先用 SHADOW 模式跑一段时间观察是否能稳定区分"健康对话 vs 关系破裂前兆"。PROPOSED

   Not a runnable A/B candidate — evaluated by the path above, not ablation.

Traceability

No plugins / runs linked yet. Scaffold a suggestion to start.

Expected benefit (预期收益)

- 让 R7 self-track 的"心理健康"有**有原则的定义**，而不是依赖行为代理指标。 - 让 `lifeform_vitals` 触发的 needs / regime 切换不再依赖均值漂移，而是基于"分布漂移"——更早、更准确地检测到关系破裂前兆。 - 与 R-PE "PE 是一级信号" 完全相容，不破坏任何契约。

Cited paper (引用论文)

**Botvinick M, Kurth-Nelson Z, Muller T, Dabney W. *Depression as a disorder of distributional coding*. arXiv:2507.16598, 2025.** - 文档位置：[`research/papers/dm/depression-as-disorder-of-distributional-coding-2507.16598.pdf`](../../research/papers/dm/depression-as-disorder-of-distributional-coding-2507.16598.pdf) - 作者权重：Botvinick（DeepMind Director of Neuroscience Research）+ Kurth-Nelson（DeepMind）+ Dabney（DeepMind，distributional dopamine 的发明者）。 - 摘要原文（精炼）： > Major depressive disorder persistently stands as a major public health problem. While some progress has been made toward effective treatments, the neural mechanisms that give rise to the disorder remain poorly understood. In this Perspective, we put forward a new theory of the pathophysiology of depression. ... We spotlight three previously separate bodies of research, showing how they can be fit together into a previously overlooked larger picture. The first piece of the puzzle is provided by pathophysiology research implicating dopamine in depression. The second piece, coming from computational psychiatry, links depression with a special form of reinforcement learning. The third and final piece involves recent work at the intersection of artificial intelligence and basic neuroscience research, indicating that the brain may represent value using a distributional code. - 关键观点：把 **VTA 多巴胺功能异常 + 计算精神病学 + distributional coding** 三块拼图缝合成一个 depression 病理学的统一理论。**核心可借鉴点**：value 应该用 distribution（不是 scalar）编码；当价值分布塌缩，就出现 depression-like 状态。这给 R7 的 lifeform vitals 提供了一个有神经科学合法性的"内禀健康指标"模板。 ---