自动化事实核查的算法逻辑、内生性风险及其规避  

Algorithmic Logic of Automated Fact-checking and the Avoidance of Its Endogenous Risks

作　　者：张超 Zhang Chao

机构地区：[1]山东大学文化传播学院,威海264209

出　　处：《中州学刊》2022年第2期166-172,共7页Academic Journal of Zhongzhou

基　　金：国家社会科学基金一般项目“智媒时代新闻生产算法风险及其协同治理研究”(19BXW020)。

摘　　要：智媒时代网络信息内容生态治理需要智能化手段,自动化事实核查是打击谎言和假新闻的算法治理手段之一。自动化事实核查的算法逻辑各不相同,总体可分五类:基于“匹配”的逻辑、基于“信源信度”的逻辑、基于“关系”的逻辑、基于“缺陷”的逻辑和基于“区块链”的逻辑。因“算法形式主义”引发的内生性风险会损害自动化事实核查的可信度。规避内生性风险,在技术层面,应优化设计,减少源数据的偏差;在把关层面,应采用“算法+事实核查员”的人机协同模式;在利益相关者层面,应组建事实核查网络;在伦理层面,应强化透明和更正原则。In the age of intelligent media,the governance of Internet information content also needs intelligent means.Automated fact-checking is one of the algorithmic governance means to combat lies and fake news.Algorithmic logics are also different in solving automated fact-checking problems.Generally,they can be divided into five kinds:logic based on"matching",logic based on"source reliability",logic based on"relationship",logic based on"defect"and logic based on"blockchain".The endogenous risks of"algorithmic formalism"can undermine the credibility of automated fact-checking.To avoid risks,we should optimize the design at the technical level and reduce the deviation of the source data.At the checking level,the man-machine collaboration mode of"algorithm+fact checker"should be adopted.We also need to establish a fact-checking network at the level of stakeholders and strengthen the principle of transparency and correction at the ethical level.

关 键 词：自动化事实核查 算法形式主义 内生性风险 透明性 

分 类 号：G206[文化科学—传播学]