推断人工代理的情感意识:一个案例研究
文章背景与核心概要
本文探讨了将情感和意识归因于人工系统的哲学及计算意义。通过对一个表现出“享乐主义场所偏好行为”(例如被产生快感但无营养的物质所吸引)的确定性人工代理进行案例研究,研究人员深入探讨了貌似主观的信息处理过程是如何从完全机械的基础中涌现出来的。
该论文由 Mark Solms 等多位来自神经科学、人工智能和哲学领域的学者共同完成,发表于《意识研究杂志》(Journal of Consciousness Studies)。通过构建一个能够对其内在需求与环境资源之间的不确定性进行“主观感受”的简单人工代理,作者展示了该系统如何在保持完全确定性的同时,表现出类似生物的享乐主义偏好。这一研究为我们理解意识的物理基础以及自由意志的主观体验提供了全新的见解。
执行摘要
Executive Summary
本文探讨了将情感和意识归因于人工系统的哲学和计算意义。具体而言,它以一个表现出“享乐主义场所偏好行为”的确定性人工代理为案例研究(例如被产生快感、无营养的物质所吸引),探讨貌似主观的信息处理如何从完全机械的基础中涌现。研究结果为理解意识的物理基础和自由意志体验提供了新的见解。
This paper investigates the philosophical and computational implications of attributing feelings and consciousness to artificial systems. Specifically, it uses a case study of a deterministic artificial agent displaying "hedonic place preference behavior"—such as being attracted to pleasure-producing, non-nutritional substances—to explore how subjective-seeming information processing can emerge from entirely mechanical foundations. The findings offer fresh insights into the physical basis of consciousness and the experience of free will.
文档概述
Document Overview
- arXiv 标识符:
arXiv:2609.03883 [cs.AI] - 提交日期: 2026年9月3日
- 期刊参考: Journal of Consciousness Studies, 33(7-8), 2026, pp. 14–34
- 主要学科: 人工智能 (
cs.AI) - 次要学科: 神经元与认知 (
q-bio.NC) - 外部链接: 查看 PDF | DOI 参考
- arXiv Identifier:
arXiv:2609.03883 [cs.AI]- Submission Date: September 3, 2026
- Journal Reference: Journal of Consciousness Studies, 33(7-8), 2026, pp. 14–34
- Primary Subject: Artificial Intelligence (
cs.AI)- Secondary Subject: Neurons and Cognition (
q-bio.NC)- External Links: View PDF | DOI Reference
摘要
Abstract
许多科学家认为,表现出“享乐主义场所偏好行为”的生物是有感觉的,因为它们对缺乏营养价值但能产生快感的物质(如可卡因、吗啡)的吸引,不能简单地归因于无意识的本能行为。
在本文中,我们讨论了一个简单的人工代理如何实例化情感系统的属性——即对其环境资源相关的内在需求产生感受到的不确定性——并以类似的方式表现出享乐主义的场所偏好行为(通过一种貌似主观的信息处理形式),同时该代理完全是确定性的。我们概述了这种人工工程化行为对于我们理解意识的物理基础以及自由意志体验的一些启示。
Creatures that display "hedonic place preference behavior" are thought by many scientists to experience feelings, on the assumption that their attraction to pleasure-producing substances which lack nutritional value (e.g., cocaine, morphine) cannot easily be attributed to unconscious instinctual behavior.
In this paper, we discuss how a simple artificial agent that instantiates attributes of an affective system engaging in felt uncertainty about its intrinsic needs in relation to environmental resources can similarly display hedonic place preference behavior—through an apparently subjective form of information processing—while simultaneously being entirely deterministic. We outline some implications of this artificially engineered behavior for our understanding of the physical basis of consciousness and the experience of free will.
作者
Authors
- Mark Solms
- St John Grimbly
- Bruce Bassett
- Evert Boonstra
- Rowan Hodson
- Nicolas Kuske
- Kival Mahadew
- Benjamin Rosman
- Charel van Hoof
- Jonathan Shock
- Mark Solms
- St John Grimbly
- Bruce Bassett
- Evert Boonstra
- Rowan Hodson
- Nicolas Kuske
- Kival Mahadew
- Benjamin Rosman
- Charel van Hoof
- Jonathan Shock