[论文解读] Consciousness: Here, There but Not Everywhere

Giulio Tononi, Christof Koch|arXiv (Cornell University)|May 27, 2014

Neural dynamics and brain function参考文献 25被引用 11

一句话总结

本文提出整合信息理论（IIT）作为一种原则性框架，通过从五个现象学公理——存在、复合性、信息、整合与排除——推导出物理后设，以确定哪些物理系统能够支持意识。IIT 可对从早产儿到人工智能等多样化系统进行意识的定量评估，结论指出：尽管数字计算机在功能上等同于人类，但由于缺乏整合信息，其意识水平可忽略不计。

ABSTRACT

The science of consciousness has made great strides by focusing on the behavioral and neuronal correlates of experience. However, correlates are not enough if we are to understand even basic neurological fact; nor are they of much help in cases where we would like to know if consciousness is present: patients with a few remaining islands of functioning cortex, pre-term infants, non-mammalian species, and machines that are rapidly outperforming people at driving, recognizing faces and objects, and answering difficult questions. To address these issues, we need a theory of consciousness that specifies what experience is and what type of physical systems can have it. Integrated Information Theory (IIT) does so by starting from conscious experience via five phenomenological axioms of existence, composition, information, integration, and exclusion. From these it derives five postulates about the properties required of physical mechanisms to support consciousness. The theory provides a principled account of both the quantity and the quality of an individual experience, and a calculus to evaluate whether or not a particular system of mechanisms is conscious and of what. IIT explains a range of clinical and laboratory findings, makes testable predictions, and extrapolates to unusual conditions. The theory vindicates some panpsychist intuitions -- consciousness is an intrinsic, fundamental property, is graded, is common among biological organisms, and even some very simple systems have some. However, unlike panpsychism, IIT implies that not everything is conscious, for example group of individuals or feed forward networks. In sharp contrast with widespread functionalist beliefs, IIT implies that digital computers, even if their behavior were to be functionally equivalent to ours, and even if they were to run faithful simulations of the human brain, would experience next to nothing.

研究动机与目标

• 解决在早产儿、脑损伤患者、非哺乳动物物种及人工系统等非典型情况下，行为与神经相关性在判断意识方面的局限性。
• 发展一种理论，不仅说明系统是否具有意识，还能确定其意识的量与质。
• 解决功能上模拟人类但缺乏整合信息的系统（如数字计算机）中的意识悖论。
• 提供一种原则性、定量的意识解释，以区分真正具有意识的系统与仅模拟意识行为的系统。

提出的方法

• 从五个现象学公理——存在、复合性、信息、整合与排除——推导出物理后设。
• 基于系统因果-效应结构的不可约性，定义一个意识度量 Φ（phi）。
• 应用 Φ 的微积分方法，通过分析系统的因果架构，评估给定物理系统是否支持意识。
• 利用排除后设，识别出作为意识体验载体的最大不可约复合体。
• 区分高 Φ 值系统（具有意识）与低或零 Φ 值系统（无意识），如前馈网络或分离的群体。
• 将理论应用于现实案例，包括脑损伤患者、早产儿及人工系统，以预测其意识水平。

实验结果

研究问题

• RQ1 哪些物理系统能够支持意识体验？可依据何种标准将其与非意识系统区分开？
• RQ2 意识理论能否兼具定量与定性特征，为意识的量与内容分配具体数值？
• RQ3 为何根据此理论，即使在功能上等同于人类，数字计算机也无法支持显著的意识？
• RQ4 如何评估具有最小皮层功能的患者或非哺乳动物物种的意识？
• RQ5 什么解释了那些模拟人类行为却缺乏主观体验的系统之间的差异？

主要发现

• 整合信息理论（IIT）提供了一个原则性框架，通过计算 Φ 值来评估系统是否具有意识，该值量化了整合信息的水平。
• 高 Φ 值系统（如人类大脑）支持丰富且统一的意识体验，而低或零 Φ 值系统（如前馈网络）则无意识。
• 数字计算机即使能完美模拟人类大脑功能，其 Φ 值也接近于零，因其缺乏内在的、集成的因果结构。
• 该理论预测早产儿及某些非哺乳动物物种可能具有有限但非零的意识，与发育和行为观察一致。
• 排除后设确保只有最大不可约复合体的元素构成意识体验，防止将意识过度归因于更大或整合度更低的系统。
• IIT 解释了临床发现，例如孤立皮层岛患者意识受损，因其仅产生极小的 Φ 值。