Research On Algorithm Discrimination And Equal Right Protection In The Era Of Artificial Intelligence

This paper is divided into five parts.The author makes an in-depth discussion on the increasingly prominent problem of algorithm discrimination in practice from the following aspects,the overview of algorithm discrimination,the threat of algorithm discrimination to the right of equality,the causes of algorithm discrimination,the extraterritorial regulation of algorithm discrimination,and the exploration of the Chinese path of regulating algorithm discrimination.Chapter one is an overview of algorithm discrimination.In the first section,the author thinks that the two concepts of algorithm and discrimination should be discussed respectively.On this basis,we can further understand the connotation of algorithm discrimination.In the second section,the author analyzes the characteristics of algorithm discrimination.In the third section,the author discusses the types and manifestations of algorithm discrimination.According to whether algorithm discrimination is related to individual's identity attributes,we can divide algorithm discrimination into two categories.Firstly,the algorithm discrimination related to identity attribute mainly includes racial discrimination,gender discrimination and face discrimination.Secondly,there is algorithm discrimination which is not related to identity attributes.This kind of algorithm discrimination mainly includes price discrimination and ranking discrimination.Chapter two,the threat of algorithm discrimination to equal right.Firstly,the author discusses the discrimination recognition crisis caused by algorithm discrimination.Firstly,the author points out that the discrimination caused by algorithm makes it more difficult for people to identify the implicit discrimination.Secondly,the author takes the racial discrimination in the United States as an example to further discuss this issue.Finally,the author believes that in the era of artificial intelligence,people are more and more accustomed to sacrifice their basic rights and interests for gaining convenience,which will lead to a higher and higher discrimination threshold.Second,the mode of equal right protection is facing a crisis.The author believes that as far as the current protection mechanism of equal right is concerned,algorithm discrimination threatens it from three aspects,namely,algorithm discrimination impacts on legislative norms,in addition,it challenges administrative supervision,and finally,judicial relief of equal right is also threatened.Chapter three,the causes of algorithm discrimination.The author thinks that algorithm discrimination mainly comes from training data.Specifically,it is the two factors of data bias and data proxy that cause algorithm discrimination.In the first section,the author's opinion is that the whole life cycle of the algorithm from development to use has a strong data dependency,so the bias in the data must be closely related to algorithm discrimination.In addition,the author introduces"feedback loops" to explain how algorithm discrimination further strengthens social prejudice in the real society,which is innovative.In the second section,the author analyzes how personal sensitive data can play a role by finding a proxy,and takes the popular machine learning as an example to further analyze the proxy mechanism of data.Chapter four,the extraterritorial regulation of algorithm discrimination.In the first section,because algorithm discrimination is a global problem,we can gain useful experience by examining the practices in other countries or regions.As a pioneer in legal regulation of algorithm discrimination,the EU has chosen a governance path centered on data protection.The General Data Protection Regulation,which officially took effect in May 2018,has an impact beyond national boundaries.The regulation mainly regulates algorithm discrimination from the two following dimensions,data sanitization and the right to know algorithm.The United States,on the other hand,has chosen to respect the existing legal framework,arguing that algorithm discrimination also applies to the traditional equal protection clause.On this basis,the USA congress enacted the Algorithmic Accountability Act of 2019,which requires algorithm controllers to conduct "automated decision system impact assessment".The essence of this system is to require algorithm designers to strengthen self-audit.Chapter five explores the Chinese path of governance algorithm discrimination.In the author's opinion,given that the practice of governance algorithm discrimination in China is still in its infancy,the author tries to explore the Chinese path of governance algorithm discrimination and improve the solution of governance algorithm discrimination from the three dimensions of technology,audit and remedy.The author thinks that the traditional tort law should be returned and the algorithm controller should be held accountable after the discrimination produces substantial harmful consequences.We can learn from the"differential impact standard" adopted in American legal practice to identify algorithm discrimination.As far as the principle of liability for algorithm discrimination is concerned,we can consider adopting no-fault liability.At the same time,in order to reduce the liability burden of algorithm controller,the author suggests establishing an algorithm compensation foundation to transfer the risk burden of algorithm controller.