|
| 文章编号：1672-3104(2026)01-0032-16 |
| |
|
迈向算法正义：算法歧视的社会建构及其治理策略 |
| |
毛俊响1，郭敏2 |
| |
|
（1. 中南大学法学院，湖南长沙，410083；
2. 中南大学人文学院，湖南长沙，410083） |
| |
| 摘 要： 算法作为一种社会存在，与社会结构相互作用。算法歧视是社会建构的结果，社会歧视的历史承继、算法运行的修正障碍、价值偏好的隐性渗透、社会生态的利益导向，共同作用于算法歧视。算法歧视并非算法内生技术性问题衍生的新形式不平等，而是历史与现实问题在算法时代的映射。算法歧视从财富、权力、声望三个方面阻碍合理化社会流动，加剧了社会结构失衡。为此，需要建构算法正义来实现对算法歧视的纠正。建构算法正义，需要回应分配正义与关系正义的双重要求，将受保护特征纳入算法决策，尊重群体多元、避免多样性“武器化”，正视基于群体差异的特殊优待并提升少数群体的算法决策话语权。实现算法正义，应采用技术、法律、伦理三元协同治理模式，重点是设计以算法区分为中心的法律规制模式，布局以“科技向善”为核心的人工智能全周期治理机制。 |
| |
| 关键词： 算法歧视；算法正义；算法区分；算法治理 |
| |
|
| |
Towards algorithmic justice: The social construction of algorithmic discrimination and its governance strategies |
| |
MAO Junxiang1, GUO Min2 |
| |
（1. School of Law, Central South University, Changsha 410083, China;
2. School of Humanities, Central South University, Changsha 410083, China） |
| |
| Abstract: Algorithm, as a socially embedded entity, interacts dynamically with the social structure. Algorithmic discrimination emerges as a socially constructed phenomenon shaped by the historical legacy of societal discrimination, the structural barriers to algorithmic correction, the subtle infiltration of value biases, and the interest-driven orientation of the social ecosystem. This discrimination does not represent a novel form of inequality inherent to algorithmic technology itself, but rather a reflection of enduring historical and contemporary issues in the algorithmic age. Algorithmic discrimination impedes equitable social mobility across dimensions of wealth, power, and prestige, thereby exacerbating imbalances within social structures. Addressing this necessitates constructing “algorithmic justice” to counter algorithmic discrimination. Establishing algorithmic justice requires a dual commitment to distributive and relational justice, integrating protected characteristics into algorithmic decisions, respecting group diversity without weaponizing it, acknowledging the need for special accommodations based on group disparities, and amplifying the decision-making voice of minority groups in algorithmic contexts. Realizing algorithmic justice should adopt a tripartite governance model which encompasses technology, law, and ethics, with a focus on designing legal frameworks centered on algorithmic distinctions and implementing a comprehensive AI governance mechanism grounded in “technology for good.” |
| |
Key words: algorithmic discrimination; algorithmic justice; algorithmic distinctions; algorithmic governance |
| |
|
|
