人工智能的可解释性问题探析——基于因果关系视角

Analysis of the Interpretability of Artificial Intelligence——Based on a causal relationship perspective

随着人工智能技术的发展与应用,人工智能的可解释性问题受到了越来越多人的关注。如何解释输入与输出之间的联系,成为许多学科的研究重点。对此,相关关系的主张者与因果关系的主张者提出了各自的观点和理论。相比较而言,从因果关系视角阐释人工智能的可解释性问题是更为合理的一种路径。但要实现这一路径,还需要借助人工智能学者朱迪亚·珀尔对因果关系的三个层次区分,以消解传统因果关系理论彼此间不相容所造成的解释上的局限性,从而实现可解释的人工智能。

With the development and application of artificial intelligence technology,the interpretability of artificial intelligence has attracted more and more attention.How to explain the connection between input and output has become the focus of research in many disciplines.In this regard,proponents of correlation and causation put forward their own viewpoints and theories.Comparatively speaking,it is a more reasonable path to explain the interpretability of artificial intelligence from the perspective of causality.But to achieve this path,artificial intelligence scholar Judea Pearl's three-level distinction of causality is needed on order to eliminate the interpretation limitations caused by the incompatibility of traditional causality theories,thus explainable artificial intelligence can be realized.