摘要
算法从来就不是客观和中立的,通常算法由谁来制定、遵循何种规则和如何被执行反映的是设计者的价值选择与社会分配正义问题。就此而言,算法即规则,而规则即统治,算法已然成为人工智能时代社会变革的关键驱动力量和新秩序的塑造者。然而,作为一种新式的权利和财富分配工具,算法在赋权一部分人的同时,算法滥用、算法误用以及算法垄断也在排斥和边缘化另外一部分人。算法歧视主要源自以下三种形成机制:原生性偏见、学习性偏见和外源性偏见。算法偏见与一般偏见不同,一旦形成不仅难以消除,而且极易产生社会放大效应。治理和规约算法偏见需要遵循以下四原则:授权同意原则、透明性原则、问责求偿原则以及人的价值和尊严优先原则。进而,在算法开发者、应用者、销售者、传播者以及管理者之间建立沟通桥梁并设立管控机制,以便将算法伤害和算法偏见收缩到最小,形成一个公平、公正的算法秩序。
The algorithm has become the key driving force of social change in the era of artificial intelligence and the maker of the new order.However,as a new type of right and wealth distribution tool,the abuse,misuse,and monopoly of algorithm are also repelling and marginalizing others while empowering someone.Algorithmic bias mainly comes from the following three formation mechanisms:original bias,learning bias and exogenous bias.Being different from general bias,once the algorithmic bias is formed,it is difficult to eliminate and easy to produce social amplification effects.The governance and rule of algorithm biases need to follow the four principles:authorization and consent,transparency,accountability and compensation,and prioritizing human value and dignity.Furthermore,establishing communication bridges and control mechanisms between algorithm developers,users,sellers,communicators,and managers,so as to minimize the algorithm damage and bias,and form an algorithm order with fairness and justice.
作者
董青岭
朱玥
Dong Qingling;Zhu Yue
出处
《探索与争鸣》
CSSCI
北大核心
2021年第3期82-86,178,共6页
Exploration and Free Views
基金
国家社科基金重大项目“基于大数据的国家安全态势感知原理及冲突数据库建设”(19ZDA131)
关键词
算法政治
算法歧视
算法偏见
算法正义
算法秩序
algorithmic politics
algorithmic discrimination
algorithmic bias
algorithmic justice
algorithmic order
