Legal Regulation Of Algorithmic Discrimination

Algorithmic decision-making has been applied to various aspects of society gradually,such as employment,medical and financial fields.Algorithms may bring discrimination risks while playing a positive role.If algorithmic discrimination cannot be effectively regulated,in the future we may be eliminated by job application algorithms due to our gender,or we may be denied service by insurance companies because of the color of our skin.However,due to the lack of prevention and the difficulty of detecting differential treatment when we face such hidden risks,the algorithmic discrimination has become more rampant.At present,China has not yet formed a systematic system of algorithmic discrimination regulation,so it is necessary to provide ideas for the construction of algorithmic discrimination regulation system in China through the horizontal comparison of the path of judicial review regulation,the path of individual empowerment regulation and the path of public-private cooperation regulation.On both sides of the Atlantic,there are two different paths of algorithmic discrimination regulation,one is the judicial review regulation represented by the United States,and the other is the individual empowerment regulation represented by the European Union.These two protection methods are the most typical at present.The “disparate impact theory” in the judicial review path can regulate the indirect discrimination implied by algorithms,but under this path,plaintiffs will bear more difficult burden of proof,higher litigation costs and greater risk of losing.As can be seen,there are certain obstacles to the opening of proceedings and the identification of disparate impact.In addition,the timing of legal intervention and the inflexibility of regulatory tools prevent this regulation from effectively identifying and avoiding risks.The individual empowerment regulation path creates the right to anti-automation decision making and the right to ex-post algorithmic interpretation,but both of these rights will have dampening effect on technological development.And it is impractical to use the power of individual against algorithmic systems.On this basis,the two new rights are likely to be set aside and the regulatory effect will be difficult to achieve.The public-private cooperation is a more prudent regulatory approach.This path is compatible with the current situation,and it keeps the same pace with the development plan of artificial intelligence,so there is the possibility of landing in China.The regulatory system can be composed of two levels: norm-setting and monitoring feedback.In terms of norm-setting,procedural norms are formulated in the form of “the government collects information and the private sector relies on government guidance”,and substantive norms are formed by self-regulatory norms and post-legal rules.In terms of monitoring and feedback,it relies on section-level information feedback,correction mechanism and community-based supervision mechanism.At the same time,it is a preventive mode in advance,which has obvious advantages in the regulation of algorithm.This path takes discrimination risk screening as the entry point,and puts legal regulation after private regulation through public-private cooperation,which can not only realize the effectiveness of regulation,but also protect trade secrets.Moreover,the application of soft law can effectively reconcile the rights and interests of all parties and ultimately achieve the moderation of algorithmic discrimination regulation.