A Study On The Approaches To Regulation Of Algorithms From The Perspective Of Comparative Law

The coming of the era of artificial intelligence means the coming of the era of algorithms.Although algorithms provide great convenience to people’s lives and promote social progress,they have a great impact on individuals,groups and society.For individuals,fair trade rights are affected by information cocoons and price discrimination.Also,algorithm-based public decision-making may harm civil fundamental rights such as the right to life.For groups,the implicit discrimination of algorithms affects their equal social status,and for society,algorithmic power and hegemony may exacerbate social unfairness.These algorithmic biases may be attributed to design biases,training data deviations,social inherent biases,and practical biases.What’s more,attention should be paid to the problem of the algorithm’s black box.The technical black box makes the algorithm’s operation process opaque,which adds complexity to explanation while the legal black box protects algorithms as trade secrets from disclosing.Consequently,black box must be opened to solve the algorithmic biases.The United States adopts the approach to free regulation,promoting industry autonomy with little government intervention,and applying the existing legal rules to regulate the algorithms.The United States focuses on the use of algorithms by public authorities,for example,in the Loomis case,the use of risk prediction system in the sentencing phase may affect the defendant’s procedural due process rights.In this regard,the court issued a written warning,paying attention to the ex ante prevention.As for ex post relief,the proof standard of constitutional amendments is too high to meet,thus making it difficult to resort.In addition,trade secrets often become the exception of disclosure under administrative law.In the area of tort liability,strict liability in product liability may be a probable solution.The EU adopts a strict regulatory approach,based on the protection of individual rights,setting a number of obligations for data controllers at the expense of high penalties.In the General Data Protection Regulations,fully automated decisions that have significant legal or similar effects on individuals are of concern,in which case data subjects have the right to inform,access,explanation,etc.The right to explanation,as a right stipulated in the recital 71,is questioned by some scholars,but according to historical interpretation,literal interpretation and systematic interpretation,the right to explanation does exist.However,it is restricted by trade secrets in some cases.The approaches to regulation of algorithms adopted by the United States and EU have its own advantages and disadvantages,and they have common points in terms of the right to explanation and the limitation of trade secrets.At present,China’s artificial intelligence is in the development stage,the existing legislation and draft laws focus on information security,ignoring the protection of solely automated decision-making and the right to explanation.So it is necessary to learn from the European and American regulatory approaches to establish a rights protection-oriented way supplied with industry autonomy.First,Adhere to the principles of algorithm fairness,algorithm transparency and algorithm accountability.Second,focus on the right to explanation to make sure the data subject understand and have the right to object the automated decision-making.Third,form a regulatory system of ex ante prevention and ex post accountability by stipulating the rights of data subjects in advance and offering relief of algorithm infringement through product liability.Fourth,the independence of the public authorities who apply algorithms to decisions should be emphasized.With these means applied,China’s artificial intelligence industry can develop in a healthy way within the legal framework.