自然语言处理中的价值观渗透问题——以ChatGPT为例

Research on the Issues of Infiltration of Values in Natural Language Processing:Taking ChatGPT as an Example

[研究目的]以ChatGPT为例,探讨自然语言处理中的价值观渗透方式和固化机理,探索在自然语言处理技术广泛应用的背景下,预防不可控信息的传播,巩固主流价值体系的方法路径。[研究方法]以描述性研究为基础,从ChatGPT驱动下价值观渗透的四种方式入手,研判可能出现的舆情传播态势。结合价值观的本身特点,综合科技哲学、伦理学等相关学科知识,整合提出目前存在的价值观渗透、固化风险,并提出相应的对策。[研究结论]通过与ChatGPT的结合,价值观能以数据投喂、自主示弱、自我优化、语言景观等方式强化其渗透过程中的隐蔽性,在渗透之后,将会通过隐性教化、知识内部分野和去实践化等方式进行固化,从而扩大影响。研究结合多领域知识,提出应对此种情况的三项基本原则:辩证扬弃、针锋相对和复合规制,力求促进情报搜集和舆情控制工作健康平稳发展。

[Research purpose]Taking ChatGPT as an example,this paper explores the ways in which values are permeated and solidified in natural language processing,and explores the methods and paths to prevent the spread of uncontrollable information and consolidate the mainstream value system in the context of the widespread application of natural language processing technology.[Research method]Based on descriptive research,starting from four ways of value infiltration driven by ChatGPT,this study analyzes the possible trend of public opinion dissemination.Combined with the characteristics of values themselves,integrating knowledge from related disciplines such as philosophy of science and technology,ethics,etc.it summarizes the existing risks of value infiltration and solidification,and proposes corresponding countermeasures.[Research conclusion]By combining with ChatGPT,values can strengthen their concealment in the process of penetration through methods such as data feeding,self weakening,self optimization,and language landscape.After penetration,they will be solidified through implicit education,internal knowledge differentiation,and de-practicality,thereby expand their influence.By combining knowledge from multiple fields,this study proposes three basic principles for addressing this situation:dialectical development,tit for tat,and composite regulation,aiming to promote the healthy and stable development of intelligence collection and public opinion control work.