This article will be published in "People's Forum·Academic Frontier" in May 2022. Author: Ma Changshan, dean of the Digital Rule of Law Research Institute of East China University of Political Science and Law, editor-in-chief of "Journal of East China University of Political Science and Law", professor, and doctoral supervisor.
Content summary: Algorithmic governance is a core issue in the digital age. It reflects the life rules and operating logic of the digital society. It has exceeded the governance mechanism and system framework of the industrial and commercial society, and its value scale has also broken through the scope of justice in physical space. , a new type of digital justice emerges at the historic moment. Therefore, in the process of algorithmic governance, the principles of digital justice should be established in a timely manner, the field of digital justice should be framed, and the path to realize digital justice should be explored, so as to build a technology-oriented, people-oriented algorithmic governance order and promote the construction of digital rule of law in our country.
Driven by the emerging technological revolution, mankind has begun to enter a new era in which "everything is digital and everything is computable". Algorithms have not only become an important organizer of production and life, but also an important constructor of social order. However, the problems and risks it brings have become increasingly prominent. For this reason, major countries in the world are actively exploring algorithmic governance systems to promote science and technology for good. , promote the digital transformation of to and social change and progress. However, algorithmic governance is significantly technical, risky, and iterative, making it difficult to apply traditional governance models and mechanisms. Information technology ethics also breaks through the justice horizon and scale of physical space. In this way, establishing the corresponding digital Justice becomes the direction guide and value support for building algorithmic order.
1. The justice principle of algorithmic governance
In today's digital society, all human activities generate data, exchange information all the time, and ensure the normal operation of digital production and life through data analysis, information control and modeling calculations. It is based on this new digital ecology that the multi-ring information circle begins to spread in any space. "Whether it is human, animal or mechanical, it is an integral part of information theory." Therefore, data information forms an independent and parallel data world (World 3) outside the physical world (World 1) and the spiritual world (World 2). Although this data world is a digital representation of the physical world and the spiritual world, it does not simply map them. Instead, it uses the information center position to realize unbounded links, sharing, reconstruction, and interactive control of the physical world and the spiritual world. Among them, the sharing and utilization of data information is the basis. It is not like bread. If you eat half of it, I will lose half. It is precisely in the constant flow of sharing, copy mining and assignment utilization that more data can be generated and more data can be added. More meanings, creating more value; but at the same time, if this kind of sharing and utilization is not controlled, it will inevitably pose a serious threat to digital personality, digital property, digital transactions, and even the sharing and utilization itself. Compared with sharing and utilization, the modeling and calculation of algorithms is a digital engine. It is not only a data information processing system, but also an alternative decision-making system and a dialysis evaluation system. With the help of algorithms, intelligent and automated production and services can be created, allowing the whole society to It can share the development benefits of digitalization; on the other hand, it will also form multiple penetrations, make hidden things visible, and induce manipulation, resulting in disturbing digital control. From this point of view, behind the data information and algorithms lies the digital life paradox of sharing and control. The core of algorithmic governance is to resolve this paradox and establish and follow the principles of justice in the digital age.
(1) Principle of independent choice
As digital transformation continues to accelerate and the level of intelligence continues to improve, people will inevitably face more and more algorithm recommendations and algorithmic decisions in their daily lives, and then enjoy a smart and convenient digital life. However, algorithms are not boundaryless. Using algorithms to process some unique human activities and implanting too much computer logic "may cause these activities to lose their most fundamental humanity"; and the generalization of algorithms in life will replace and misuse, abuse, and can easily evolve into a “take it or leave it” digital compulsion.Therefore, retaining necessary parallel options outside of algorithmic decision-making and maintaining human autonomy and personal selectivity are undoubtedly fundamental requirements for digital justice. my country's relevant laws, regulations and rules also have corresponding provisions, which mainly include: First, independent reservations in key areas. That is, in the service field involving personal dignity and basic survival, the government or operators should control the scope and functions of algorithmic decision-making, so as to retain the choice space for citizens (consumers) whether to accept automated decision-making and provide a channel for human decision-making (artificial services). ; The second is the convenient rejection of personalized services. That is, when the government or operators make algorithmic decisions based on sensitive personal information or personal characteristics, the consent and right of refusal of citizens (consumers) should be respected, ensure that users have convenient channels to turn off these algorithms, and provide Other options after rejection; third, smooth and reliable relief guarantee. That is, when citizens (consumers) question the results of algorithmic decision-making and cannot accept the results, relief channels for human decision-making should be provided for citizens (consumers). In this way, we can effectively limit the digital control tendencies that may arise from algorithmic decision-making, better leverage the positive effects of algorithmic decision-making, release digital welfare, and share smart lives. Therefore, independent choice is an important value principle that algorithmic governance must follow.
(2) Principle of openness and transparency
The important reason why algorithm recommendation and algorithmic decision-making is highly respected is that it has some advantages over human decision-making, such as scientificity based on big data analysis, objectivity without emotional stance, High efficiency of fast super computing and so on. However, algorithms are not perfect or omnipotent, and they cannot “solve” all problems. More importantly, whether in the field of government services or commercial transactions, the providers and recipients of algorithmic services are in a state of obvious asymmetry in information and technology, and there is no "face-to-face" dialogue and interaction between them. At this time, algorithm recommendation and algorithm decision-making are like a "blind box". "Users cannot see the rules clearly, raise different opinions, and cannot participate in the decision-making process. They can only accept the final result." The well-known algorithmic black box will inevitably contain “instability, unfairness, and unearned factors.” In order to solve this problem, laws and ethical norms around the world require algorithms to be explainable, open, and accountable. Article 7 of my country's " Personal Information Protection Law " and Article 4 and 16 of " Internet Information Service Algorithm Recommendation Management Regulations " also stipulate that: Information processors should disclose personal information processing rules and clearly state the purpose of processing. Method and scope; when providing algorithm recommendation services, the basic principles, purpose intentions, and main operating mechanisms of the algorithm recommendation services should be disclosed in an appropriate manner. In this way, people's convenient sharing and rational supervision of algorithm services can be greatly enhanced, and algorithm misuse and abuse can be effectively suppressed, thereby providing the necessary value benchmark for strengthening algorithm governance and promoting its friendliness and good.
(3) The principle of fairness and reasonableness
In modern times, on the basis of the logic of industrial and commercial society and Enlightenment thoughts , a modern legal system with personality rights, property rights, creditor's rights, etc. as the core has been constructed, forming the sanctity of private rights and freedom of contract. , the value of justice of owning one's own responsibilities, and establishing corresponding principles of fairness and reasonableness in the fields of public law and private law. However, with the advent of the information revolution , networking, digitalization, and intelligent transformation have subverted the usual way of human life. "A series of major decision-making rights, from driving to stock trading to company staffing, have been transferred from human hands to In the hands of algorithms. "The justice values and principles of fairness and reasonableness of modern law are facing severe challenges, resulting in fundamental reshaping and era transformation.
First, maintain the exchange balance. In an era where everything is connected and data speaks, everyone is both an information producer and an information consumer, and data information therefore occupies a central position in production and life.When people face digital services such as online car rental, facial payment, personalized recommendations, smart investment advisory, and smart medical care, they need to provide or share necessary personal information to complete the process. That is to say, online life and algorithm applications provide smart and convenient services, but at the same time, people must also hand over some necessary information to the platform and algorithms, thereby forming a digital ecosystem of mutually beneficial exchange and rational utilization. At this moment, it is difficult to rigidly adhere to personal privacy rights, nor can we simply apply the principles of justice in the "physical age". Instead, we should uphold an open and inclusive spirit, on the one hand, strengthen the protection of personal sensitive privacy information, and on the other hand, strengthen the commercialization of personal general information. Utilization and public utilization, thereby establishing new fair and reasonable principles for data information processing and algorithm regulation based on the reciprocal exchange and balance of interests between algorithm service providers and recipients, realizing digital justice and building a digital legal order.
Secondly, uphold legality and reasonableness. Algorithms are undoubtedly designed by humans. Whether it is AI writing AI code or the complex neural network algorithm , they are all completed under the control and guidance of humans. Even if there is an algorithm black box that programmers and may not be able to explain clearly, the algorithm is not out of control and is still within the control of humans. In fact, from clearly labeling data, drawing knowledge maps, to algorithm modeling, and writing codes, the values of designers and operators are inevitably embedded. They all "absorb the social concepts and social goals of designers, and try to Reproduce them in the future”. Therefore, how to limit the value preferences of algorithms and bring them into a legitimate and reasonable normative framework is undoubtedly the fundamental task and important mission of algorithm governance. my country's "Internet Information Service Algorithm Recommendation Management Regulations" clearly requires that "the provision of algorithm recommendation services should abide by laws and regulations, respect social morality and ethics, abide by business ethics and professional ethics, and follow the principles of fairness, openness, transparency, scientific rationality, and good faith. "Principles." Specifically include: algorithms must adhere to mainstream value orientation, avoid illegal and false information, and must not use algorithms to implement shielding, manipulation, inducement, discrimination, unreasonable restrictions, unfair and unfair treatment, and implementation of monopoly and unfair competition. These are the basic norms and value measures to ensure that science and technology are good, legitimate and reasonable, and need to be determined and confirmed based on different application scenarios, different rights and interests, different categories and levels, and different numerical relationships.
Again, follow the proportion principle . As mentioned earlier, the algorithm is a data information processing system, which is necessarily based on data information and is directly related to the data information rights and interests of users, enterprises, governments, etc.; on the other hand, the algorithm is a decision-making system, and its decision-making results are also It will have an important impact on the rights and interests of users, enterprises and governments. In this way, algorithmic governance includes private rights, the interrelationships between private power and public power, and the boundaries of rights and interests, and should be weighed in accordance with the principle of proportionality. my country's "Personal Information Protection Law" includes "not excessive", "minimum impact", "minimum scope", "reasonable scope" and "necessary scope" in various clauses involving information processing, contract conclusion and performance, etc. and limit", "the shortest necessary time" and other provisions. These provisions are the basic requirements of the principle of proportionality and an important support for algorithmic governance. Different from the previous principle of proportionality, the following measurement benchmarks should be followed: First, based on the sharing and control attributes of data information; Second, adhere to the dominant direction of algorithms serving people and sharing the welfare of digital development; Third, protect digital personality and digital survival development and other basic rights; the fourth is to provide necessary priority protection to recipients of algorithmic services who are in a weak position in asymmetric relationships.
Finally, avoid misuse and abuse. Algorithm is an digital technology application. In many fields and in many aspects, the computing power and processing power exceed that of humans, but "not everything that can be calculated is important, and not everything that is important can be calculated." It will also have its own limitations. Such as algorithm errors, algorithm black boxes, algorithm discrimination, algorithm collusion, algorithm manipulation, etc., these will produce adverse consequences that seriously deviate from fairness and reasonableness.Therefore, avoiding the misuse and abuse of algorithms is the bottom line requirement of the principle of fairness and reasonableness, and is an important guarantee for maintaining the order of algorithm governance.
(4) Principles of Digital Human Rights
In the digital age, personal information has fully covered all private life from the cradle to the grave, "until a 'person' is formed in the computer database." Therefore, in daily life, everyone They are both natural persons and digital persons; under the institutional framework of digital government , they are digital citizens. Among them, regardless of algorithms in commercial transactions or algorithms in automated administration, there is the possibility of creating digital divide , algorithmic discrimination, and algorithmic manipulation. This will bring serious obstacles to people's survival and development in the digital society, resulting in the emergence of digital human rights issues. The application of algorithm technology should be people-oriented and based on improving human digital life capabilities and quality of life, and should not be alienated into automatic means of restricting and controlling people. UNESCO "Recommendation on Ethical Issues in Artificial Intelligence" clearly states that artificial intelligence technologies are extremely intrusive. They may violate human rights and fundamental freedoms and are widely used. Therefore, the use of AI systems for social scoring and mass surveillance should be explicitly prohibited, individuals should also have the right to access or even delete their personal data records, and countries should take action to improve data protection and individuals’ understanding of and control over their own data. Therefore, maintaining and protecting digital human rights has become an indispensable and important criterion for algorithm governance.
2. The justice field of algorithmic governance
In the past justice concepts and theories, they basically revolved around the distribution of identity, status, wealth, opportunities and conditions, and were developed in the form of freedom, equality and rights. However, after entering the digital age, everything will be expressed, linked, interacted and circulated through networked, digital and intelligent methods. Data information and modeling algorithms will assume the important function of relationship construction and order shaping. At this time, the field and scope of fairness and justice are no longer limited to physical resource allocation models and mechanisms, but have formed digital "sharing\control" models and mechanisms. One of its serious consequences is that a considerable part of the connotations of freedom, equality, rights, and justice have been hollowed out and realized by digitalization, leaving only the "shell" remaining. For example, the labor rights of food delivery riders have always existed, but the algorithmic order dispatching mechanism of "competition for the lowest price" has caused them to form serious "involution" that is difficult to control; the contractual rights in commercial transactions remain the same, but big data has matured, Algorithmic discrimination has quietly "removed money from the bottom of the pot"; similarly, the various rights of the elderly have not changed, but the digital divide has caused them to lose the ability to realize their rights in life. This shows that the focus of algorithmic governance is to deal with digital control and achieve digital justice. The digital control here does not directly allocate resources through physical transfer methods and artificial means, but implicitly matches resources and mobilizes behaviors through data information processing and algorithm control. Therefore, it is put on the objective cloak of technology and establishes Automatically implemented execution mechanism. Therefore, the suppression of digital control has become a major area of digital justice.
(1) Control of data information
Control of data information refers to the digital anchoring of the counterparty through the collection, storage, processing, transmission, provision, disclosure, deletion, etc. of data information. It is not only the prerequisite for controlling the target object, but also the basis for algorithm modeling. As we all know, with the iterative transformation from commodity economy to digital economy , data information has become a key production factor, and algorithms have also become an important productivity. To this end, information is "not so much for storage as for circulation", thereby building the digital economy into a sharing economy that "does not seek ownership, but seeks use" and is "available but invisible". Since all social behaviors in the digital age "center around production and information control," when large amounts of data are shared and utilized, the control of data information becomes very critical.Specifically, on the one hand, personal data information in social life will be produced at any time, and it will also be collected, stored, mined and used by Internet platforms, technology companies and government agencies in the form of provision, transmission, crawling, etc., thus forming a There is a clear distinction between the information-advantaged parties and the information-weakened parties. On the other hand, algorithm governance is not limited to the code writing and calculation process itself, but a systematic form of data input-modeling calculation-result output including data information processing. The data and information processing processes such as collection, storage, use, processing, transmission, and provision reflect the different values and interest preferences of the information processors. However, consumers or citizens are often difficult to detect and have no way to deal with this. This is undoubtedly algorithmic governance. basic premise. It can be seen that unfair and unreasonable control of data information is essentially a gap between the ability to control data information and the ability to take digital actions, which leads to more serious and widespread social injustice. This requires implementing the principle of digital justice in algorithmic governance, setting up a legal, reasonable, and secure data information processing framework, and strengthening the data information control capabilities of consumers or citizens, thereby placing sharing and control within the scale of digital justice.
(2) Algorithm-based control
Algorithm-based control mainly refers to a control method that relies on algorithm recommendations or algorithm services to exert overt or covert influence on the other party, thereby intervening and inducing the thoughts and behaviors of the other party. It belongs to a An indirect intervention strategy from the outside to the inside. This control is mainly accomplished through the following steps: First, panoramic view. That is, data profiling is carried out through big data collection and analysis. If in 1993 a dog posted comments anonymously on the Internet, people did not know it was a dog; then in 2013, "the new version of the Internet not only knows that you are a dog, it also knows your breed." I’ll sell you a bowl of fine grained dog food.” At this time, everyone is just a “transparent individual” with a panoramic view in front of big data and algorithms, forming an “electronic cage that no one can escape from.” . Secondly, algorithm feeding. An important feature of the digital age is that "whoever controls and processes data information has the ability to control the behavior and thoughts of others." The former Facebook News designed a feeding algorithm to calculate the user's Personal characteristics and preferences, and then push information to suit their preferences through "filter bubbles". In the U.S. presidential election, platform companies such as Facebook, Twitter, and YouTube analyzed large amounts of user information to carry out data profiling and personalized push, thereby manipulating voters’ consciousness and behavior. “This influence will not trigger users’ Consciousness, but a kind of subconsciousness, makes them become part of your thoughts and makes you feel that they are your own opinions. "Similarly, the British Brexit decision-making process also shows a "new political form revolution" -" In the future, data will control politics, not your brain.” Again, inductive bias. Some researchers point out that automatic learning technology and all artificial intelligence technologies almost reflect a priori bias of their founders. They provide us with a preset and already explained world in their algorithms. In this way, the world we see "exactly is what we expect to see or what we have been trained to see." If these presupposed views or preferences are not restricted, they will "control our lives to a greater extent." It can be seen from the above that control based on algorithms will have a profound impact on personal dignity, free choice, equal participation, democratic rights, etc. This requires following the principle of digital justice to prevent algorithms from inducing preferences, controlling thoughts, and influencing people's behavior. Thereby safeguarding digital human rights and achieving the just goal of algorithmic governance.
(3) Control through algorithm
Control through algorithm is mainly to predict, classify, deploy or determine a certain matter through algorithmic decision-making, and then has an important impact on the rights and interests and life of the other party. It is a kind of control method. Automated direct intervention strategies. This mainly includes: First, prediction and evaluation.In recent years, some domestic and foreign public and private sectors such as finance, investment, and police departments have been developing a large number of algorithm prediction and evaluation systems. For example, from September 2004 to April 2007, the public welfare system in Colorado, USA, developed an automated service system; in January 2013 , the Alhambra City Police Department in Southern California uses the "Predictive Crime" system (PredPol); other countries are also developing financial risk assessment systems, urban security risk assessment systems, etc. However, these ratings and assessments of individuals—employment, loan eligibility, risk level—are primarily intended for targeted screening, which is likely to create a conscious or unconscious algorithmic bias. More importantly, it will lead to a "control revolution" that will be reclassified, evaluated, compared, and even seen through. The principles of equality, freedom, justice, and democracy will all encounter significant challenges and dangers of derogation. The second is algorithm instructions. For example, algorithmic orders for food delivery riders and online ride-hailing, algorithmic pricing of platforms, algorithmic transaction orders, smart contracts, etc., are all automated instructions issued by algorithms to realize the deployment of human and material resources. Here, the algorithmic system has transformed from a simple tool into an important "decision maker" and instruction issuer. Based on this, "people who rely on algorithms to determine the direction of action may be guided to a fixed route, just like an actor who can only follow a script." It’s the same as acting.” Once there is a serious value preference or an error occurs, it may lead to serious adverse consequences. The third is algorithmic ruling. In August 2021, the Russian online payment service company Xsolla fired 147 employees using the algorithm's "disengagement and inefficiency" judgment. Coincidentally, in 2019, the internal AI system of Amazon in the United States tracked the work efficiency of employees in the logistics and warehousing department, and then automatically generated dismissal instructions. These are undoubtedly abuses of algorithms that lack warmth and disregard human nature. In our country, the number of automated judgments, automated administration, and intelligent referees is also increasing, such as automated bidding systems, automated law enforcement systems, automated referee systems (intelligent assisted referees), etc. These systems all use algorithmic decision-making to deal with corresponding matters. A decision will be made and will take effect after being confirmed by the responsible person. Although these algorithmic decisions can surpass humans in terms of efficiency, objectivity, consistency, etc., they also have the possibility of algorithmic errors, algorithmic biases, algorithmic discrimination, algorithmic black boxes, algorithmic hegemony, etc., and may even create a "digital decision-making system". The social risks of “lockdown”. Algorithmic practice in the United States shows that it may also "automate original discrimination and exacerbate inequality." Therefore, it is necessary to effectively constrain and regulate the control through algorithms in accordance with the principles of independent choice, openness and transparency, fairness and reasonableness, and digital human rights, and establish an algorithmic governance order that promotes technology for good.
3. The just path of algorithmic governance
Algorithmic governance is a brand-new topic in the digital age. It is a highly complex system that integrates technology, ethics and institutions. Its governance methods, governance strategies and governance orientations have exceeded the scope of traditional governance. Therefore, it is necessary to build a new governance paradigm, practice the principles of digital justice, base ourselves on the field of digital justice, and explore new paths and new strategies to achieve digital justice.
(1) Establish the regulatory mechanism of agile governance
Agile governance is an emerging governance concept in the digital era. It aims to form a governance model and mechanism that is fast, flexible, inclusive and balanced, jointly participated, and people-oriented, so as to respond to emerging technology challenges and solve innovation and development problems. and promote smart governance systems. In March 2022, the General Office of the CPC Central Committee and the State Council issued the "Opinions on Strengthening Science and Technology Ethics Governance", which clearly requires agile governance and ethics first, "strengthen science and technology ethics risk early warning and tracking research and judgment, timely and dynamic adjustment of governance methods and ethical norms, and quickly and Flexibly respond to the ethical challenges brought about by technological innovation. "To this end, it is imperative to explore algorithmic supervision mechanisms for agile governance.
First of all, uphold the position of digital rule of law.That is to say, get rid of the inertial thinking of traditional rule of law, start from the concept and position of digital rule of law, and formulate regulatory rules, establish regulatory procedures, and implement reasonable regulatory strategies in accordance with the basic attributes of data information, the basic laws of algorithm governance, and the basic logic of digital rights. To "step on the accelerator" for innovation and "step on the brake" for order, the standard is that the algorithm is "people-oriented" and follows digital justice. For example, seven national ministries and commissions require "algorithm selection" to optimize the ordering mechanism for food delivery, which is an important attempt and effort.
Secondly, explore sophisticated and agile governance strategies. Including establishing a refined hierarchical and classified supervision mechanism in accordance with the law based on the relative importance of algorithmic services and algorithmic decision-making, user scale, mobilization capabilities, intervention intensity, etc.; establishing rapid collaboration among government departments such as Internet Information, telecommunications , public security, and market supervision. mechanism, safety assessment and supervision and inspection mechanism; improve the science and technology ethics review and supervision system, establish a timely and efficient tracking, monitoring, and early warning mechanism for science and technology ethics risks; establish a flexible, diverse, inclusive and prudent algorithm compliance mechanism, and implement timely and effective positive Incentive regulation and so on.
Once again, strengthen joint participation and cooperation. Agile governance of algorithms is very technical, complex, systematic, and risky. It is difficult for the government alone to do a good job. Therefore, it is necessary to mobilize broad participation from all social forces, including technology companies, network platforms, industry organizations, and technology companies. People, consumers, news media, etc., jointly safeguard the value of digital justice and create a law-abiding algorithmic governance environment.
(2) Establish a self-regulatory mechanism with clear responsibilities
Legal governance relies heavily on technical regulations and operational control. Therefore, an algorithmic self-regulatory mechanism based on subject responsibility is very critical. At present, countries around the world are actively exploring autonomous models of algorithm governance. Technology companies such as Microsoft, Google, IBM and other technology companies have formulated ethical principles for artificial intelligence development. Under the guidance of policies and norms such as the "New Generation Artificial Intelligence Governance Principles - Developing Responsible Artificial Intelligence" and the "New Generation Artificial Intelligence Ethical Code", Chinese enterprises are also taking active actions to promote the development and application of trustworthy AI. According to the "Internet Information Service Algorithm Recommendation Management Regulations" and the requirements of corresponding ethical norms, my country's self-regulatory mechanism for algorithm governance mainly includes:
First, implement the main responsibility of the algorithm. Algorithm services or algorithm decision-making providers should establish and improve management systems and technical measures such as algorithm mechanism review, data security and personal information protection, and conduct strict self-regulation; they should comprehensively use strategies such as content deduplication and fragmentation intervention to optimize retrieval, Transparency and interpretability of rules such as sorting, selection, push, and display; establishing a complete user choice mechanism and fully respecting user autonomy, right of choice, etc.
The second is to establish a risk assessment and early warning mechanism. Regularly review, evaluate, and verify algorithm mechanisms, models, data, and application results, and promptly rectify and correct problems if they are discovered. At the same time, we can learn from the previous "algorithmic bias bounty" launched by Twitter in the United States in July 2021, and publicly solicit its own "algorithmic bias" issues from the society, in order to create an " crowdsourcing model " for algorithm governance, thereby promoting the algorithm Self-discipline mechanism innovation.
The third is to actively develop ethical tools. Through technological research and development, we will promote the ability and level of technological regulation, provide market demanders with appropriate risk assessment tools and corresponding algorithm repair services, and promote the refinement and professional development of ethical services.
The fourth is to promote AI ethical industry constraints. This is a global trend. UNESCO started writing a global AI ethics recommendation in 2019, and adopted the world's first "Recommendation on Ethical Issues in Artificial Intelligence" at the 41st UNESCO General Conference in November 2021. . In 2019, the European Commission High-level AI Expert Group (AI HLEG) released the "Ethical Guidelines for Trustworthy AI", and major countries are also actively responding to and promoting algorithm governance.In my country, in June 2019, the National New Generation Artificial Intelligence Governance Professional Committee issued the "New Generation Artificial Intelligence Governance Principles - Developing Responsible Artificial Intelligence"; in May 2019, the Artificial Intelligence Ethics and Security Research Center, the New Generation The Artificial Intelligence Industry Technology Innovation Strategic Alliance and other academic institutions and industrial organizations jointly released the "Beijing Consensus on Artificial Intelligence"; in August 2019, the Shenzhen Artificial Intelligence Industry Association and Megvii Technology , iFlytek and dozens of other companies Enterprises jointly released the "New Generation Artificial Intelligence Industry Self-Discipline Convention". In addition, major technology companies such as Baidu , Tencent and other major technology companies have also proposed their own AI ethics guidelines; Megvii Technology has also established an artificial intelligence ethics committee based on the formulation of guidelines to promote the safety, reliability and controllability of artificial intelligence. Credible and sustainable. Through these mechanisms, it is possible to establish and improve industry standards, industry guidelines, self-discipline management systems and service specifications, thereby better promoting algorithm governance and safeguarding digital justice in the new era.
(3) Establish a sound and effective social supervision mechanism
In the digital age, the application of algorithms changes everything in production and life, and will have a real and important impact on everyone. However, algorithms are highly technical, invisible, complex, and automatic, making them difficult for people to approach, understand, and grasp. Therefore, it is urgent to establish a sound and effective social supervision mechanism.
First, open channels for social supervision. The "Internet Information Service Algorithm Recommendation Management Regulations" clearly requires that algorithm recommendation service providers should formulate and improve service specifications, provide services in accordance with the law and accept social supervision, and set up convenient and effective user appeals and public complaints and reporting portals, with clear processing procedures and feedback Time limit, timely acceptance, processing and feedback of processing results. However, a lot of efforts are still needed to implement this provision. Moreover, the supervision channels of government departments, industry organizations, etc. also need to be unblocked and expanded.
Second, activate supervision capabilities. In fact, there are many social organizations, grassroots organizations, news media, etc. in our country, but not all of them can play an effective supervisory role. Therefore, activating their supervisory capabilities is crucial. For example, in 2019 Nanjing Jianye District required sanitation workers to wear smart bracelets to "supervise work", and in 2020 Suzhou launched the "Suzhou Civilized Code", etc., which attracted attention and doubts from the society. The local government finally returned to the algorithm governance boundary with the suggestion It has achieved good results in terms of prudent administrative strategies.
Third, strengthen digital abilities. Each of us has the power and capacity to exercise our rights and fulfill our obligations. But after entering the digital age, everyone must not only have rights and behavioral capabilities, but also need to have the ability to master, use, operate digital products and handle affairs online, so as to adapt to and participate in digital life. Therefore, in the face of social supervision of algorithms, people need to develop good digital literacy and digital capabilities, better understand and experience algorithms, and use digital capabilities to supervise algorithms, thereby safeguarding their legitimate rights and interests. This is undoubtedly a broad force to promote algorithmic governance and achieve digital justice.
(4) Establish a fair and reliable judicial protection mechanism
In a society governed by the rule of law, judicial relief is the final channel for rights protection and fair access, and is also the fundamental support for social governance. Therefore, algorithmic governance is no exception. In other words, in the process of algorithmic governance, judicial relief undoubtedly sets up an important barrier to safeguard digital justice.
Since algorithm recommendation and algorithm decision-making are allocative and controllable, judicial relief in algorithmic governance will also be more complex, and interest relationships will be more important.Its judicial relief methods mainly include: first, the judicial review mechanism of algorithms, that is, once an algorithm lawsuit occurs, the judicial authority should initiate an algorithm audit in judicial procedures, and conduct equity measurement and judicial ruling based on the algorithm audit results; second, the algorithm cases The public interest litigation mechanism includes incorporating public interest litigation into algorithm cases involving public interests and the rights and interests of vulnerable groups for rights protection; the third is to focus on the protection of important areas and important rights and interests, such as the basic rights and interests of consumers or digital citizens in algorithm litigation, and the protection of rights and interests. Monopoly and anti-unfair competition, the basic rights of the elderly and minors, digital human rights related to survival and development, etc., all need to be protected. Only in this way can we embody the good-hearted and people-oriented spirit of science and technology, better realize the principle of digital justice, and build a good algorithmic governance order.
Conclusion
Facts have shown that the application of emerging technologies is both creative and destructive. To this end, UNESCO's "Recommendation on Ethical Issues in Artificial Intelligence" breaks through the dilemma of "development first" or "governance first" and emphasizes the coordinated progress of the two. For our country, we also need to uphold an inclusive and prudent governance strategy and actively build an algorithmic governance system that uses science and technology for good. However, the benchmark for science and technology to be good is no longer "physical" justice centered on material distribution; it is digital justice centered on information sharing and control. This requires effective exploration and reconstruction of digital justice concepts, digital justice principles, digital justice fields, digital justice procedures, etc. based on digital production lifestyles and digital behavioral laws, thereby providing core motivation and laying a solid foundation for the construction of digital rule of law in our country.
The special topic "Digital Rule of Law" is specially contributed by the Institute of Digital Rule of Law of East China University of Political Science and Law. The topic coordinator: Qin Qiansong
