在大型语言模型上进行大型道德机器实验
Muhammad Shahrul Zaim Bin Ahmad1,2, Kazuhiro Takemoto1,3
1Department of Bioscience and Bioinformatics, Kyushu Institute of Technology, Iizuka, Japan.
PloS one
|May 21, 2025
概括
这项研究分析了52个大型语言模型 (LLM),用于自动驾驶汽车的道德决策. 较大的模型,特别是超过100亿参数的模型,与人类道德更好地保持一致.
科学领域:
- 人工智能的人工智能
- 技术中的伦理学
- 自主系统 自主系统
背景情况:
- 大型语言模型 (LLM) 正在快速发展,引发了关于其道德决策的质疑.
- 将LLM集成到自动驾驶系统中需要评估它们与人类道德价值观的一致性.
- 之前的研究研究了有限数量的LLM,需要进行更广泛的分析.
研究的目的:
- 评估52个不同的LLM在自动驾驶场景中的道德判断.
- 评估LLM伦理决策与人类偏好的协调程度.
- 调查模型大小,更新和架构对道德对齐的影响.
主要方法:
- 联合分析框架用于评估伦理困境中的LLM反应.
- 对专有 (GPT,Claude,Gemini) 和开源 (Llama,Gemma) LLM 的评估.
- 分析包括模型大小 (参数),版本更新和架构在内的因素.
主要成果:
- 专有和更大的开源LLM (>10B参数) 与人类的道德判断更接近.
- 在开源模型中,观察到模型大小和与人类判断的偏差之间存在负相关性.
- 模型更新并没有持续增强伦理一致性;一些LLM过度强调特定的伦理原则.
结论:
- 增加LLM大小可能会自然地改善类似人类的道德判断.
- 在自动驾驶中实施LLM需要平衡伦理与计算效率的协调.
- 文化背景对于人工智能在自主系统中的道德决策至关重要.
相关概念视频
02:16Stereotype Content Model
13.9K
The Stereotype Content Model (SCM) was first proposed by Susan Fiske and her colleagues (Fiske, Cuddy, Glick & Xu, 2002; see also Fiske, 2012 and Fiske, 2017). The SCM specifies that when someone encounters a new group, they will stereotype them based on two metrics: warmth—or that group’s perceived intent, and how likely they are to provide help or inflict harm—and competence—or their ability to carry out that objective. Depending on the warmth-competence...
13.9K
01:27Language and Cognition
303
Language serves as a bridge between ideas and communication, influencing how individuals perceive and interact with the world. Psychologists have long debated whether language shapes thought or vice versa. This discussion gained grip with Edward Sapir and Benjamin Lee Whorf in the 1940s, who proposed that language determines thought, a concept known as linguistic determinism. They suggested that the vocabulary and structure of a language influence how its speakers think and perceive reality.
303
02:20Milgram's Obedience to Authority
6.3K
Obedience to authority is classically demonstrated in a more famous series of social psychology experiments performed by Stanley Milgram. He was a social psychology professor at Yale who was influenced by the trial of Adolf Eichmann, a Nazi war criminal. Eichmann’s defense for the atrocities he committed was that he was “just following orders.”
6.3K