On Algorithmic Explanation Obligation

In the era of digital economy,artificial intelligence technology with data and algorithmic as the core accelerates the integration with public governance and economic fields,which promotes the transformation of public governance mode and business model.Smart governance gradually replaces the traditional public governance mode,offline real economy is gradually replaced by online economy,and algorithmic has become the leading force to lead social development in the era of digital economy.However,with the emergence of deep learning algorithmic,the opacity and autonomy of algorithmic technology are becoming more and more obvious,algorithmic injustice occurs frequently,and the application of algorithmic technology is facing controversies and challenges.Scholars at home and abroad have carried out research on algorithmic governance and believe that algorithmic black box is the culprit of algorithmic injustice.However,there are still disputes about whether to break the black box of algorithms and how to break the black box,whether to implement algorithmic disclosure,whether to implement algorithmic accountability,and whether to give individuals the right to explanation algorithms to fight against algorithmic black box.Firstly,this paper explains the importance of breaking algorithmic black box from the cause of algorithmic black box and the social risk caused by algorithmic black box.Secondly,by analyzing and comparing the legal regulation mode of algorithmic black box in European Union,America and our country,it is concluded that the transparency and interpretability of algorithmic are the first criterion to govern algorithmic black box.algorithmic filing,algorithmic accountability,data protection and other ways to regulate the black box of algorithms are based on the transparency and interpretation of algorithms,but it is difficult to break the black box due to its inherent defects.Thirdly,the way to solve the algorithmic black box is that the algorithmic designer and the algorithmic user should fulfill the obligation of algorithmic description.The obligation of algorithmic explanation is based on the transparency and interpretability of algorithmic,which is not only normative and legitimate in theory,but also operable in practice and some strengthening reasons.Finally,the paper puts forward the realization path of algorithmic explanation obligation.The subject of the obligation of algorithmic explanation shall be the legal person or unincorporated organization providing algorithmic service;The purpose of the obligation of algorithmic explanation is to ensure that individuals are informed,and the object of the obligation includes the public with general algorithmic literacy and the regulatory authorities with professional literacy.Due to the conflict of algorithmic technology,the scope and limit of its obligations are divided into three levels: prior obligation,in-process obligation and post-event obligation: prior obligation needs to be performed including the disclosure of algorithmic impact assessment report,algorithmic operation logic,algorithmic application scenarios,algorithmic inherent defects and other basic explanations;The explanation obligation in the matter includes internal supervision and truthfully accept the dynamic explanation of external supervision;Ex post statements are specific statements with relief as the core.At the same time,in order to make up for the deficiency of algorithmic explanation obligation,this paper proposes that the gatekeeper system and algorithmic accountability system can be supplemented by the platform to supervise algorithmic activities to carry out collaborative governance of algorithms.