Legal Regulation Of Algorithmic Bias In The Era Of Big Data

With the advent of the era of big data,the party and the state attach great importance to the important role of big data in social economy and raise big data into one of our national strategies.The core value of big data is prediction,and the implementation of its prediction function is mainly done by big data algorithm.The ubiquitous use of big data has also created a series of problems.The traditional data management-cutting regulatory path has been unable to meet the needs of regulating big data problems in the face of big data powerful data processing and prediction functions.This paper believes that to effectively regulate the series of problems generated by big data,it is necessary to start from the core of big data algorithm-big data algorithm.In this paper,the author selects the increasingly serious problem of algorithmic bias as an entry point,in order to find an effective path to effectively regulate big data and big data algorithm problems.The first chapter of this paper gives a basic explanation of the discrimination of big data algorithms.The author believes that the big data algorithm gives big data life,but because the algorithm designer may discriminate and the algorithm data itself may be discriminatory,the big data algorithm itself may have discrimination.The problem of algorithmic bias in various fields of the real world confirms this possible existence.The author believes that big data algorithm discrimination is an unfair treatment behavior of a certain tag group in the operation of big data algorithm.It has five characteristics of discriminatory,concealed,complex,systematic and irreversible.The main sources of algorithmic bias are algorithm design and data input,including: data sample size difference,data itself with bias,algorithm design flaws,machine autonomous learning acquisition,and acquisition in human-computer interaction.The second chapter of this paper emphasizes the necessity of algorithmic bias regulation.The authors start with the huge harm of the big data algorithm itself with discrimination against genes and big data algorithms.The author believes that the big data algorithm itself has discriminatory genes.This kind of discrimination gene is mainly caused by three factors:First,the inherent discrimination in human society.The discrimination of big data algorithm is to some extent the digital expression of the inherent discrimination in human society in the digital society.Second,the algorithm itself lacks human ethics,and it is difficult to acquire human ethics under the specific purpose of the algorithm,so it is inevitable to producediscriminatory behavior.Third,the sensitive attributes of human beings in the era of big data have nowhere to hide,even after data.Processing,big data algorithms can still find this sensitive attribute from other dimensions and combine it with the specific purpose of the algorithm,which is extremely likely to lead to discrimination.The author believes that big data algorithm discrimination is an "invisible injustice" and has great harm to society: first,it damages fairness and justice,and because of the hidden and complex characteristics of algorithmic bias,this kind of harm is not easy to be people.The second is to cure and expand discrimination.On the one hand,algorithmic bias continues the tradition of discrimination in human society,on the other hand,it constantly expands the “digital divide” of existing society;thirdly,it leads to the materialization and commercialization of human beings,and humanity The object that can be calculated,predictable,and controllable in the algorithm code;the fourth is the deprivation of self-determination,the human being is more and more living in the rule of the algorithm with discrimination;the fifth is to eat away the residual value of consumers,the big data algorithm discriminates “Primary price discrimination” in economics has become possible,and consumer surplus value has been exploited.The third chapter of this paper sorts out several dilemmas in the current algorithmic bias regulation.First,the boundary problem of algorithm decision-making,which areas can't use algorithms to make decisions,which areas should limit algorithm decisions,and which areas can open algorithm decisions? How to design the system to ensure that these boundaries are not exceeded? The second is the responsibility of algorithmic power.The algorithm has become a power that affects people and society.Who is responsible for the corresponding responsibility? The third is the algorithm and trade secrets.How to balance the “black box”and the protection of corporate trade secrets? Fourth,the algorithm is transparent and censored.How can we make it transparent and transparent in the face of such a complex big data algorithm? How to review the algorithm?The fourth chapter of this paper combs the response to algorithmic bias in the domestic and foreign legal systems.From the domestic point of view,China does not have clear legal regulation on the issue of algorithmic bias.From some sporadic laws and regulations,it shows a relatively tolerant attitude towards the big data algorithm “measures while walking”.From the perspective of data,it is still from the perspective of data.Regulate.Foreign countries have mainly formed a regulatory path based on the EU “risk prevention model” and the US“risk control model”.The EU focuses on the prevention of big data algorithms from theperspective of personal information protection and data management,while the United States attaches great importance to the innovative value of big data,and pays more attention to setting up a unified regulatory body to control the harm of big data algorithms.The fifth chapter of this paper proposes the legal regulation path of algorithmic discrimination.The first is to strengthen the legislation of artificial intelligence,improve the framework design of discriminative regulation of big data algorithms,and clarify the access and boundary problems of algorithm decision-making.The second is to establish a unified government regulatory agency and strengthen industry self-discipline,establish an algorithmic review and monitoring mechanism,through pre-examination review and assessment,assess risks,strengthen monitoring in the event,supervise the whole process,and control the risk of discrimination caused by big data algorithms.The third is to improve the algorithm responsibility distribution mechanism,establish a dual-track responsibility system for platform responsibility and technical responsibility,and strengthen the social responsibility of big data enterprises.The fourth is to establish a damage relief mechanism,through the protection of user rights,the establishment of industry insurance protection mechanisms,damage compensation mechanisms,effective relief for users suffering from algorithmic decision-making damage.