有帮助的助手还是有成效的促进者? 调查人格如何影响语言模型的行为
Pedro Henrique Luz de Araujo1,2, Benjamin Roth1,3
1Faculty of Computer Science, University of Vienna, Vienna, Austria.
PloS one
|June 30, 2025
概括
将人格分配给大型语言模型 (LLM) 显著影响了他们在各种任务中的行为. 与标准提示相比,Persona赋值在LLM代中引入了更大的变化.
科学领域:
- 人工智能的人工智能
- 自然语言处理自然语言处理.
- 人与计算机的交互
背景情况:
- 大型语言模型 (LLM) 在各种应用中越来越多地使用.
- 控制LLM输出和行为对于可靠和安全的部署至关重要.
- 赋予人格是指导LLM生成的一种拟议方法.
研究的目的:
- 调查分配人格对七种不同的大型语言模型行为的影响.
- 分析个人如何影响客观和主观任务的LLM响应.
- 将人格驱动的世代与对照和无人格基线设置进行比较.
主要方法:
- 从12个类别 (例如,性别,职业) 分别分配了162个人到7个LLMs.
- 提示的LLM包含来自五个数据集的问题,涵盖客观和主观任务.
- 与对照组 ("有帮助的助手"的变体) 和一个空虚的角色设置进行了比较.
主要成果:
- 在所有测试的模型和数据集中,Personas显然增加了LLM代的可变性.
- 人格对模型行为的影响比控制设置更为明显.
- 人格驱动行为的某些方面在不同LLM中显示出一致性和概括性.
结论:
- 人格赋值是调节LLM行为和输出多样性的有效技术.
- 这些发现表明,人物可以可靠地影响LLM响应,提供定制的方法.
- 进一步的研究可以探索特定的LLM应用程序的个人设计细微差别.
相关概念视频
Language and Cognition
456
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.
456
Attribution Theory
13.4K
Behavior is a product of both the situation (e.g., cultural influences, social roles, and the presence of bystanders) and of the person (e.g., personality characteristics). Subfields of psychology tend to focus on one influence or behavior over others. Situationism is the view that our behavior and actions are determined by our immediate environment and surroundings. In contrast, dispositionism holds that our behavior is determined by internal factors (Heider, 1958).
13.4K
Language
430
Language is a unique communication system that uses words and systematic rules to organize and transmit information. Unlike other forms of communication, which may involve postures, movements, odors, or vocalizations, language relies on symbols and grammar. This makes human communication distinct from that of other species, who also communicate but do not use language in the same way humans do.
Corballis and Suddendorf (2007) and Tomasello and Rakoczy (2003) highlight the role of language in...
Corballis and Suddendorf (2007) and Tomasello and Rakoczy (2003) highlight the role of language in...
430
Social Cognitive Perspective on Personality
669
Social cognitive perspectives on personality emphasize the importance of conscious awareness, beliefs, expectations, and goals in shaping behavior. These perspectives incorporate behaviorist principles, such as learning through reinforcement and conditioning, but extend beyond them by highlighting human reasoning and planning. Unlike traditional behaviorist views, social cognitive theory focuses on how individuals reflect on their past experiences and plan for future outcomes by considering...
669
The Behavioral Perspective on Personality
445
Behaviorists view personality as primarily shaped by environmental reinforcements and consequences. According to this perspective, behavior is influenced by external stimuli, and individuals adjust their actions based on rewards and punishments. Over time, learning histories — accumulated patterns of reinforcement — play a significant role in shaping personality. Behaviors that lead to positive outcomes are reinforced, while those resulting in negative outcomes are diminished.
445
Introduction to Personality Psychology
18.0K
Personality encompasses a set of enduring traits and behavioral patterns that define how individuals think, feel, and interact, ultimately shaping their unique identities. The concept of personality has deep historical roots, deriving from the Latin term "persona," which means "mask." This term initially referred to the roles played by actors in ancient theater, signifying the different facets individuals display in various contexts.
Early Theories of Personality
The study of...
Early Theories of Personality
The study of...
18.0K


