跳转至

面向人类状态跃迁的生理学世界模型

文章背景与核心概要

现代连续多模态传感技术使得人们能够在日常生活中对人类生理机能进行前所未有的观测，从而打破了仅限于临床偶尔访问的局限。然而，现有的健康人工智能系统通常侧重于识别当前状态、评估风险或分析孤立的生物标志物，却未能对生理状态如何随现实世界的事件、行为、语境和干预措施发生动态变化进行建模。

为了填补这一空白，本文提出了生理学世界模型（Physiological World Model, PWM）——一个旨在学习全人（whole-person）生理状态跃迁的事件条件框架。其核心贡献包括：引入了“人类状态跃迁 Token”（HumanState Transition Token），这是一种结构化且带有质量评分的单元，将事件前生理状态、事件/行动、语境、干预措施、事后生理轨迹以及观察到的结果紧密相连；划定了四个能力等级，从基础状态代表性到有界干预规划；设计了四种严格的数据采集与验证协议；并提出了六大基准任务，用于评估状态表示、多时间尺度预测、个性化响应预测、替代干预模拟、有界规划以及在分布偏移下的可靠性。该框架为个性化健康管理、行为干预设计和临床医生监督的决策支持提供了切实可行的途径。

📌 执行摘要 (Executive Summary)

现代连续多模态传感技术使我们能够在日常生活中对人类生理进行前所未有的观测，而不再局限于偶尔的临床访问。然而，现有的健康人工智能系统通常专注于识别当前状态、评估风险或分析孤立的生物标志物，未能对生理状态如何响应现实世界的事件、行为、语境和干预措施进行动态变化建模。

为了解决这一空白，本文引入了生理学世界模型（PWM）——一个旨在学习全人生理跃迁的事件条件框架。核心贡献包括： * 人类状态跃迁 Token： 一个结构化、具质量评分的单元，将事件前生理状态、事件/行动、语境、干预措施、事后生理轨迹以及观察到的结果连接起来。 * 四个能力等级： 从基本的生理状态表示延伸至有界的干预规划。 * 严格的验证协议： 四种不同的数据采集与验证框架。 * 六项基准任务： 评估表示能力、多时间尺度预测、个体化响应预测、替代干预模拟、有界规划，以及在分布偏移下的可靠性。

Modern continuous multimodal sensing enables unprecedented observation of human physiology throughout daily life, moving beyond isolated clinical visits. However, existing health AI systems typically focus on recognizing current states, assessing risks, or analyzing isolated biomarkers, failing to model how physiological states dynamically shift in response to real-world events, behaviors, contexts, and interventions.

To address this gap, this paper introduces the Physiological World Model (PWM)—an event-conditioned framework designed to learn whole-person physiological transitions. Key contributions include: * The HumanState Transition Token: A structured, quality-scored unit linking pre-event physiological states, events/actions, contexts, interventions, post-event trajectories, and observed outcomes. * Four Capability Levels: Ranging from basic state representation to bounded intervention planning. * Rigorous Validation Protocols: Four distinct data acquisition and validation frameworks. * Six Benchmark Tasks: Evaluating representation, multi-timescale forecasting, individualized response prediction, alternative intervention simulation, bounded planning, and reliability under distribution shift.

📖 摘要 (Abstract)

连续多模态传感现在允许我们在日常生活中观察人类生理，而不是仅在偶尔的临床访问中进行观察。然而，大多数健康人工智能系统旨在识别当前状态、估计风险或分析个别生物标志物。它们没有直接模拟生理状态如何响应现实世界事件、行为、语境和干预而发生变化。在这里，我们提出了生理学世界模型（PWM），这是一个用于在全人层面学习这些变化的事件条件框架。我们引入了人类状态跃迁 Token（HumanState Transition Token），这是一种结构化的、质量评分的单元，它连接了事件前的生理状态与事件或行动、相关语境和干预信息、事件后的生理轨迹、观察到的结果以及数据质量。我们描述了从状态表示到有界干预规划的四个能力等级，以及四个数据采集和验证协议。我们还提出了六个基准任务，涵盖人类状态表示、跨多个时间尺度的预测、个体化响应预测、替代干预的模拟、有界规划以及在分布偏移下的可靠性。总之，该框架为个性化健康管理、行为干预设计和临床医生监督的决策支持提供了实用的路径，同时明确将预测与因果推断分开，并使不确定性、安全性、治理和使用限制显性化。

Continuous multimodal sensing now allows human physiology to be observed throughout daily life rather than only during occasional clinical visits. However, most health artificial intelligence systems are designed to recognize current states, estimate risks or analyse individual biomarkers. They do not directly model how physiological states change in response to real-world events, behaviours, contexts and interventions. Here we propose the Physiological World Model (PWM), an event-conditioned framework for learning these changes at the level of the whole person. We introduce the HumanState Transition Token, a structured, quality-scored unit that connects the physiological state before an event with the event or action, relevant context and intervention information, the physiological trajectory after the event, observed outcomes and data quality. We describe four capability levels, from state representation to bounded intervention planning, together with four data acquisition and validation protocols. We also propose six benchmark tasks covering HumanState representation, forecasting across multiple timescales, individualized response prediction, simulation of alternative interventions, bounded planning and reliability under distribution shift. Together, this framework provides a practical path towards personalized health management, behavioural intervention design and clinician-supervised decision support, while clearly separating prediction from causal inference and making uncertainty, safety, governance and limits of use explicit.