Regulating algorithms: Don't let the algorithm "calculate" people
Li Shigang Bao Ding Yurui
The Provisions on the Recommendation and Administration of Internet Information Service Algorithms (hereinafter referred to as the Provisions), which will come into effect on March 1, "shine a bright sword" on various algorithmic chaos. This is the first departmental regulation on the mainland to use algorithms as a special object of regulation. What exactly is an algorithm? Why regulate it? What are the innovations in the Provisions? While regulating, how to make algorithms better benefit human beings?
Algorithms, everywhere
In the era of Internet big data, algorithms have become ubiquitous, almost all services and applications in computers and mobile phones are using algorithms to operate, and news about algorithms is not uncommon.
The English name for the word algorithm comes from the Persian mathematician al-Khwarizmi, which means "method of operation." In the field of computing, algorithms refer specifically to "computer computing methods", that is, a set of calculation methods designed to solve a class of problems or perform a calculation that relies on mathematical models, has a clear definition, and has limited steps.
Abstracting a real-world problem into one or several mathematical problems falls into the world of algorithms. In Internet services, algorithms have been widely used, bringing great convenience to people's production and life.
For example, search engine search sorting algorithms have greatly improved the efficiency of human beings in finding the information they need in the Internet world; The natural language recognition and processing algorithms of the translation software reduce the difficulty of obtaining information in other languages; The path planning algorithm in map navigation can help humans plan the information with the shortest time, the shortest distance, and the fewest traffic lights; The scheduling algorithm in taxi software and takeaway software can achieve fast and efficient matching between the service supply side and the demand side...
Algorithmic "black box" lurks risks
With the development, evolution and iteration of algorithms, algorithms are widely used in various industries. However, as a "black box", for people other than the owner or designer of the algorithm, people can only observe its input and output, but can not understand its operation process mechanism, which may make the algorithm unconsciously have a negative impact on individuals and society. Taking the most common algorithm recommendation technology as an example, it may have a great impact on information dissemination, user rights and interests, and social stability.
First, the information recommended by the algorithm may bring about the problem of illegal and bad information dissemination. When algorithm service providers carry out activities such as news information push, hot list, and information release, they may have situations of disseminating illegal and negative information, including the dissemination of false information, one-sided pursuit of traffic and speculation of vulgar content.
Secondly, algorithm recommendation may also infringe on the rights and interests of users, bringing about unfair and unfair situations. This includes "big data killing", that is, the implementation of unreasonable differential treatment in trading conditions according to consumer preferences, trading habits and other characteristics; Induce users to indulge in the Internet and over-consume; Ignoring labor rights and interests protection when using scheduling decision-making algorithms for work scheduling; Form unreasonable algorithm discrimination through user portraits, etc.
Finally, algorithm recommendation service providers may also have the possibility of manipulating public opinion. In services such as personalized recommendations, hot lists, and information release, algorithms may frequently recommend content that users are interested in, forming an "information cocoon" with polarized opinions; Algorithms may be used to carry out false traffic fraud such as likes, comments, and forwards, to incite user emotions and influence online public opinion; Algorithms can also be used to mask information, over-recommend, manipulate intervention in the ranking of search results, and implement unfair competition.
The main measure to regulate algorithms
Obviously, the algorithm is a double-edged sword, and it is necessary to regulate and guide the algorithm to a certain extent.
The algorithms regulated by the Provisions are mainly recommended algorithms, including algorithms such as generative synthesis, personalized push, sorting selection, retrieval and filtering, and scheduling and decision-making. These algorithms have been widely used by major platforms and have become the core force of platforms to collect and process data, push information, and allocate resources. The Provisions make it clear that algorithm recommendation service providers should be responsible for algorithms.
The Provisions clarify that users of algorithm recommendation services enjoy the right to know, choose, manage and withdraw from the algorithm; Minors, the elderly, workers, consumers and other entities shall receive special protection; Algorithm recommendation service providers shall allow users and the public to conveniently make appeals, complaints, and reports, clarify the handling process and feedback time limit, and promptly accept, handle, and feedback the results of the processing.
The Provisions innovatively propose a regulatory system that runs through the process of algorithm recommendation services: implement hierarchical and classified management of service providers according to the public opinion attributes or social mobilization capabilities of algorithm recommendation services, user scale, and data sensitivity of processing; An algorithm filing system has been established, and algorithm recommendation service providers with public opinion attributes or social mobilization capabilities shall file information such as their names, service forms, and application fields; An algorithm security assessment mechanism has been established, requiring algorithm service providers to set up algorithm security management organizations, security responsibility systems, and scientific and technological ethics review systems, and to equip them with professionals and technical support appropriate to the scale of algorithm recommendation services; To severely crack down on algorithm recommendation violations of laws and regulations, the national and local internet information departments will be responsible for working with relevant departments to carry out algorithm security assessment and supervision and inspection work, forming a multi-departmental coordination regulatory system.
The Provisions also clearly stipulate strict legal responsibilities and establish an accountability system with clear powers and responsibilities. Algorithmic problems that have occurred frequently in the past, such as algorithm discrimination, big data ripening, inducing addiction, telecommunications network fraud, network "water army", monopoly and unfair competition, are expected to be effectively curbed.
Realize the positive and benign development of the algorithm
In view of the emerging technology of algorithms, regulatory algorithms must consider the benign dynamic balance between technological innovation and user rights protection. Overall, the Provisions embody the overall overall coordinated regulatory mechanism and the governance concept of comprehensive governance and fine governance. The regulatory measures of the Provisions are innovative and practical, and play an important role in ensuring the security of information content, maintaining the order of market competition, and protecting the rights and interests of citizens.
Of course, the law's regulation of algorithms cannot be exhaustive. At the specific landing level of the Provisions, whether allowing users to simply withdraw from the algorithm recommendation can meet the best needs of users, how to clearly meet the user's right to choose and delete a large number of labels, how to manage implicit labels, and how to balance the effectiveness of algorithm filing and the protection of trade secrets need to be further explored. In addition to legal regulation, it is also necessary to actively play the role of "soft law" such as industry standards, ethical rules, and corporate practices.
In the Internet era, the supervision of algorithms should not only be limited to the legal system on "paper", but also improve the comprehensive governance capabilities of algorithms, so that algorithms can not only conform to technical logic but also play a positive role. The law should not only protect social order and individual rights, prevent algorithms from doing evil, and guide algorithms to good, but also play an important role in encouraging and improving the mainland's algorithm innovation ability, promoting the security of algorithm applications, and enhancing the core competitiveness of mainland algorithms. The future era is the era of algorithms, and technical and regulatory departments should communicate and learn from each other to achieve positive and benign development of algorithms. (Author affilications are: Fudan University Law School, Chinese Min University Law School)
(Published in Half Moon Talk, No. 2, 2022)