Hu Minjie
Zhejiang University Guanghua Law School Professor
With the widespread use of automated administration, flawed administrative instructions are brought about by data sources, algorithms and system factors. These will have an impact on the rights of the counterparty and cause problems and difficulties in subsequent relief. Based on the different mechanisms of defect formation in automated administrative instructions, identify the essence of defective instructions in automated administration, and understand that the algorithm black box is the basis for defects in automated instructions and is the basis of legal normative considerations. Based on this, traditional due process protection, complaints, and reconsideration and litigation face new challenges. The defective instructions actually correspond to several possibilities of relief, namely relief for administrative procedures , administrative activities and factual behaviors. For flawed instructions, we need to explore relief content that is more in line with the characteristics of automation administration.
Introduction
Automated administration is increasingly widely used in the administrative field. It has the advantages of making administrative decisions more accurate and consistent, saving costs and helping cost benefits, reducing the arbitrary use of power by administrative officials, but it may also bring many problems, such as destroying due procedures and generating algorithm black boxes. For example, in 2018, a citizen in Hangzhou participated in the lottery for passenger car indicators and obtained the indicator in June 2019, but because he was not notified, it was not discovered until the indicators had expired in 2020. According to Article 46 of the "Hangzhou City Small Passenger Car Total Volume Regulation and Management Regulations", if it is not used within the deadline, it shall be deemed to be a waiver of the indicator, and no incremental indicators shall be applied for within two years from the date of expiration of the validity period. This situation is actually due to the failure of the system to send, which caused the citizen to not receive the relevant text message. Faced with this situation, how can the administrative counterparty obtain corresponding relief? For example, the 2014 Guidance Case No. 26 "Li Jianxiong v. Guangdong Provincial Department of Transportation Government Information Disclosure Case". On June 1, 2011, the plaintiff submitted an application for government information disclosure to the defendant Guangdong Provincial Department of Transportation through the Guangdong Provincial People's Government network system. According to the provisions of the "Regulations on the Disclosure of Government Information of the People's Republic of China", the defendant should reply to the plaintiff before the 23rd of the month, but the defendant did not reply within the statutory period and provide the applied government information. Therefore, the plaintiff requested the court to make a judgment to confirm that the defendant's failure to reply within the statutory period was illegal; and the defendant Guangdong Provincial Department of Transportation argued that it took a certain time for the external network to enter the intranet data processing, the defendant failed to receive it, and thus failed to accept the application in a timely manner. In this case, citizens submitted government information disclosure applications through the government public network as required, but how do they define the date of receipt of government information disclosure applications? For example, during the prevention and control of the new crown pneumonia epidemic, due to some technical failure, untimely update of personal information, or when the high-speed rail passes through medium- and high-risk areas but the passengers do not get off the train, the health code may still be displayed as unqualified. How should the counterparty seek relief? It can be foreseen that such flawed instructions will surface more in the future when automated administrative instructions are frequently used.
Referring to the theory of administrative behavior defects, the automated administrative defect instructions referred to in this article mainly refer to defects caused by data source errors, algorithm errors, and system errors in automated administration. In serious cases, illegal automation instructions can also be included, such as discrimination, violation of fairness and other defective instructions.
Since automated administration is generated "one-time" in many cases, defective instructions generated in automated administration need to be distinguished and relief and guarantees are provided according to different paths. The issues discussed in this article are focusing on this, that is, how the administrative counterpart obtains corresponding relief for the consequences caused by defective instructions arising from automated administration.
1. Source of defective instructions in automated administration
Flawed instructions in automated administration originate from many possibilities, and we cannot judge legal liability or remedies based solely on their appearance. The mechanisms formed behind this appearance, including the subsequent corresponding relief methods, vary.
Data source defects
Any kind of automated administration requires collecting data.If there is a problem with the data source, the conclusion drawn from this is fundamentally wrong. For example, during the generation of health codes, the red code cannot be corrected on the spot because an individual accidentally filled in the wrong data; for example, the identification error was obtained when the information was obtained. In 2018, the "illegal" information of Dong Mingzhu collected by the traffic police in through the automated face recognition system was actually the photos on the corresponding body advertisements, and no real illegal incidents occurred; and, due to its inherent defects, electronic monitoring equipment may not be able to identify the wrong judgment caused by specific behaviors at a certain moment. Article 22 of the " Road Traffic Safety Violations Handling Procedures " revised by the Ministry of Public Security in 2020 stipulates that "If traffic technology monitoring equipment records or enters the illegal behavior information of the road traffic violation information management system, if any of the following circumstances and is verified, it should be eliminated: police cars, fire trucks, ambulances, and engineering rescue vehicles perform emergency tasks; during the stolen and robbery of motor vehicles; if there is evidence to prove that rescue difficulties or emergency avoidance; the scene has been processed by the traffic police; due to inconsistent traffic signal indications;... other situations that should be eliminated." Reverse thinking based on the elimination situation stipulated in this article, it can also be believed that the illegal behavior information of the road traffic violation information management system may have "error" situations that do not conform to the actual situation.
For example, in the "Guo Ruili and the Public Security Traffic Penalty Case of the Tenth Traffic Police Brigade of Zhengzhou Public Security Bureau", during the annual review, the plaintiff encountered the defendant requesting a fine on the grounds that the plaintiff's vehicle had a violation record of " electronic eye ", otherwise the vehicle would not be reviewed. It was not until December 11, 2013 that the plaintiff was forced to pay a violation fine of "electronic eye" before the vehicle was examined. The plaintiff in the case believes that the electronic eyes adopted by the defendant did not apply for installation verification and mandatory annual inspection to the government measurement supervision and management department stipulated in the metrology method in accordance with the law. The information of the "electronic eyes" is illegal as the basis for punishment. Of course, the court finally held that the relevant provisions of the " of the Measuring Law of the People's Republic of China" are different from those of the administrative subject for handling traffic violations because the administrative subject for the annual inspection of motor vehicles are different from those for handling traffic violations. The former is the quality inspection department and the latter is the traffic management department; the annual inspection of motor vehicles is not the statutory responsibility of the traffic management department as the defendant, and the plaintiff's claim that the defendant abuses his power has no basis. However, this does not indicate that the "electronic eye" does not have legality even if it has not been reviewed annually. It is just that the court did not determine the subject in this case because the subject is not qualified. For example, the credit evaluation punishment mechanism is currently used in many areas of public governance. Credit evaluation is actually based on various data sources. If there are problems with these data themselves, the same is true. Article 37 of the " Shanghai Medical and Health Personnel Protection Measures " stipulates that public security organs and others shall collect relevant information from the municipal public credit information platform in a timely manner in accordance with the law; if the relevant illegal acts are minor or actively eliminate or reduce the harmful consequences of illegal acts, they shall not be collected; the municipal health and health authorities shall summarize the breach of trust information stipulated in the preceding paragraph and promptly push them to medical and health institutions and medical insurance institutions.
Automated administration reduces the discretion space of civil servants and determines the rules in an electronic form, which to a certain extent helps administrative entities handle cases more rationally. However, it is precisely because of its very fixed and mechanized that administrative instructions rely too much on data and are restricted by them, thus breeding many problems caused by data errors.
algorithm defects
computer is not only a digital calculator, but also a digital and symbol processor. It needs to be calculated according to certain rules, or is good at calculating and similar tasks, but if some instructions are wrong or defective, it will eventually give an incorrect result. For example, after the implementation of the health code, many cases of abnormal code change caused by backend information collection errors have occurred in various places. The number of people applying for review because the health code display does not match the actual situation has exceeded 30,000 within four days of the launch. There are obvious differences in specific information collection procedures in various places, and there are also problems such as unified norms.What data comparison and analysis is based on the technical background? How to assign values to various types of data? How to deal with data conflicts? How to specifically determine the evaluation criteria for risk levels and what are the specific basis for dynamic adjustments? These standards set by administrative agencies have not been disclosed.
Another example is that there may be problems in the design of an internal system instruction because even the best programmer will have shortcomings. Due to the lack of a context with case characteristics and the need for some traditional police domain knowledge in traffic violation speed measurement, the algorithm will have certain flaws, including the lack of appropriate sensor data or the inability to process advanced cognitive functions in the software.
System defects
In the current intelligent era in which humans are living, the intelligent system based on automated administration is inevitably flawed and cannot make independent administrative decisions . Typically, such as the electronic traffic monitoring equipment set up on various roads. Article 15 of the "Procedures for Handling Procedures for Road Traffic Safety Illegal Acts" stipulates that the traffic management department of the public security organs may use traffic technical monitoring equipment to collect and fix evidence of illegal acts; traffic technical monitoring equipment shall comply with national standards or industry standards, and shall be used to collect evidence of illegal acts after being recognized and certified by relevant national departments; traffic technical monitoring equipment shall be maintained, maintained and tested regularly to keep its functions intact. In the "Second Instance of Public Security Administrative Management and Road Traffic Management between Fang Yi and Jinhua Public Security Bureau Traffic Police Detachment", the appellant pointed out that if the traffic technical monitoring equipment does not comply with the legal provisions, it means that it has been infringing on the huge legitimate economic and safety interests of the public for so many years... Since the law stipulates that traffic technology monitoring equipment needs to be regularly maintained, calibrated and tested, traffic police must of course implement it in accordance with the law. Although the appellant's claim is relatively simple, it actually vividly demonstrates the problems that machine failures may arise. Of course, the court ultimately believes that if the evidence for determining the illegal act is indeed sufficient, whether the technical specification requirements and whether the equipment maintenance evidence is provided will not affect the determination of the legality of the administrative penalty involved in the case.
Another example, in the guidance case No. 26 of the Supreme People's Court at the beginning of the writing, the defendant Guangdong Provincial Department of Transportation argued that the plaintiff applied for government information disclosure through the public network system of the Guangdong Provincial People's Government, namely the provincial government's government affairs external network (hereinafter referred to as the provincial foreign network), rather than the defendant's internal local area network (hereinafter referred to as the department's internal network). Since the defendant's intra-department network is physically isolated from the Internet and the provincial network, the Internet and provincial network data cannot be directly accessed to the intra-department network for processing. It is necessary to access the intra-department network through the gate through the data "ferry" method. Therefore, the defendant staff failed to immediately discover the plaintiff's application submitted in the public network system of the Guangdong Provincial People's Government, resulting in the defendant's failure to accept the application in a timely manner. From this we can see that system failures will affect the time point of the information disclosure application and will also have an impact on the determination of subsequent facts.
2. Legal and normative basis for remedy for defective instructions in automated administration
Automated administration instructions are based on data, rely on various computer systems, and use algorithms to form certain rules and commands. From the perspective of administrative agencies, it has replaced the investigation and evidence collection of various administrative personnel, and has even developed to the present, administrative decisions can be made automatically and directly based on the automatic collection of evidence. For example, in the "Tian Zhipeng vs. Administrative Penalty Dispute Case of the Station Front Branch of Xi'an Public Security Bureau in Shaanxi Province", the Traffic Police Detachment of Xi'an Public Security Bureau established a self-service payment system for road traffic violations. Tian Zhipeng failed to issue a simplified procedure "Penal Decision" when handling violations in the traffic police hall, which violated the provisions on the punishment procedures in the " Road Traffic Safety Law " and requested the revocation of the administrative penalty decision in accordance with the law. Does this automated administration still need to follow the requirements of traditional administrative procedures? What should be the corresponding follow-up relief measures? This is the legal normative basis that should be considered for the defective instructions brought about by automated administration.
The essence of defective instructions in automated administration
Compared with traditional administration, the most intuitive feature of automated administration is that automated administration instructions are generally electronic, while our traditional administrative activities and decisions often emphasize written and formal requirements, namely the so-called key administrative behavior ; for some emergencies or relatively simple cases, the non-key administrative behavior method may be adopted. This is a difference in the form sense. From a substantial comparison, automated administrative instructions are prone to flaws, and there are at least two aspects:
1. Reducing discretion
Automatic administration reduces the discretion space of administrative staff. Relying on instructions, this discretion has become a mechanized rule-making process. The relevant factors that must be considered in the discretion can be input through the system in advance, and the relevant factors that do not need to be considered will be eliminated in advance. The machine will calculate millions of connected microinstructions , which can achieve the strategic objectives of legal norms. This complex micro-instruction directory can be applied to larger behavioral action networks for a wider goal to achieve wider goals. Of course, legal philosophers are quite controversial about this, because in their appearance that the law must be vague and uncertain, the legal reasoning process needs to understand and support the principles of analogical reasoning. For example, laws often require the realization of fairness, so can this fairness be quantified? First, fairness is a vague concept, and legal fairness may be difficult to translate into algorithmic fairness; second, fairness being quantified and algorithmized may bring about discrimination. "Big Data: Inclusive Tools or Exclusive Tools" released by the Federal Trade Commission (FTCh)? 》 (Big Data: a Tool for Inclusion or Exclusion?) report, focusing on the issue of discrimination and prejudice in big data: Are the data sets representative? Will the data model used lead to bias? How accurate is prediction based on big data? Will relying on big data lead to moral or fairness issues? These problems often arise with reducing discretion. Although we have been trapped by abuse of discretion, its existence is necessary based on administrative efficiency considerations and case decision requirements. When we move from an administrative decision that requires personalized judgment to a regular process, factors such as the overall interests of society, human emotions, local customs, etc. may be ignored.
2. Reduce the program
Automatic administration reduces the procedural links of administrative actions. For example, in the "Tian Zhipeng vs. Administrative Penalty Dispute Case of the Station Branch of Xi'an Public Security Bureau of Shaanxi Province", Tian Zhipeng believed that when the traffic police hall handled the violation, the administrative entity did not issue a simplified procedure "Penal Decision", which violated the procedural provisions of the "Road Traffic Safety Law". During the trial, the court held that Tian Zhipeng's illegal behavior was clear and that the defendant's application of simplified procedures was in compliance with the regulations. As for the failure to issue the "Penal Decision", it was because the self-service payment system itself could print it on its own and the defendant publicized the illegal handling process in the processing hall, and posted the "Notice on the Process of Replenishing Legal Documents" and related operating instructions; Tian Zhipeng could print the "Penal Decision" in the self-service system of the processing hall based on the operating instructions, but the plaintiff did not print it according to the above notification content, so he claimed that the defendant did not explicitly pay the fees and did not issue the "Penal Decision" and other opinions to him. The court refused to adopt it, and his lawsuit request was rejected. In terms of form, the procedure was reduced, but it was essentially transferred to another form, namely self-service printing, and at this time the plaintiff had paid the fine, which can be regarded as giving up the corresponding rights on his own. But here naturally raises relevant questions. Are the procedures that automate administration depends on simple procedures? Or can the necessary programs be defaulted? At this time, can the obligations of the administrative subject be transformed into the rights of the administrative counterparty? An automated law enforcement system threatens procedural rights, i.e. the opportunity for counterparty to be notified and defend. In order to relieve the legitimate rights and interests of the counterparty, any automated law enforcement system should have a large amount of failure protection, redundant backup of and related mechanisms to ensure that appropriate notification is completed and that individuals can effectively appeal the automatically issued punishment.
Understand the core of automated administration: algorithm
algorithm is the core content of automated administration. Only by understanding the algorithm can we understand why automated administration can cause flaws and cause many problems. "Algorithms are a series of instructions that tell the computer what to do." "The core of algorithm is a set of instructions that run according to the set program in order to obtain ideal results." All algorithms include the following common features: input, output, clarity, finiteness, and effectiveness. The trend of automated administration development is learning algorithms, which will increase the judgments that many humans have, are intuitive and original, and can weigh possible futures based on probability. Although many of the problems we discussed above cannot be called this type of learning algorithm, similarly, whether it is technical monitoring or automated fine payment system, there are technical blind spots, namely the so-called algorithm black box. We cannot know what the factors considered behind the system are. The non-disclosure and opaque algorithms make automated administration prone to endless nakedness.
Therefore, automated administration will create core propositions related to human rights protection, including privacy protection and personal information protection. Based on the existing relief path for administrative behavior in my country, according to Article 16, Paragraph 12 of the " Administrative Litigation Law", if an administrative agency infringes on other legitimate rights and interests such as personal rights, property rights, etc., it may file a lawsuit. However, the problem is that "legal rights and interests" are often submerged in the algorithm black box in automated administration. The black box formed by the algorithm has been further expanded to become networked. Many of us live in a networked society at all times, and we often have no way to know when, how and whether our information has been violated. In other words, automated administration based on algorithms in the black box cannot rely on a decision to promptly know its own rights protection situation like traditional administrative actions, which increases the difficulty of subsequent relief for defective instructions.
The traditional administrative law foundation of defective instruction relief
In addition to the above-mentioned rights basis, it is also necessary to further clarify what scope of the relevant relief of its instructions can be included in the traditional administrative law perspective. According to the three main types of data source defects, algorithm defects, and system defects, they actually correspond to several possibilities of relief, namely relief for administrative procedures, administrative activities and factual behavior. As shown in Figure 1.
Figure 1 The foundation of traditional administrative law for automated administrative defect instructions and relief
1. Data source defect
Since data is used as the basis for basic data collection and analysis, it will not directly constitute a certain decision, so in a sense, it is similar to an administrative procedure. For example, in the first instance administrative judgment of the "Gao Ruili and the Public Security Traffic Penalty Case of the Tenth Traffic Police Brigade of the Zhengzhou Public Security Bureau", the relevant information collected by the "electronic eye" becomes the basis for subsequent administrative penalties, and is an evidence material in the Administrative Litigation Law. When this is a program, decisions made on such programs alone can indeed have problems with legitimacy. In this case, the plaintiff shall file an illegal information record of the vehicle in the jurisdiction of Zhengzhou Airport in . If the plaintiff has any objection, he shall submit it to the traffic management department of the section of the road. The plaintiff took the initiative to go to the defendant to accept illegal information processing in other places, and did not express any objection to the contents of the "Penal Decision" such as "If there is any traffic violation in other places, sign and confirm after no objection". The plaintiff signed the name in the signature of the person being punished and actively fulfilled the obligation to pay the fine. This is a punishment for his own rights and should be deemed to have no objection to the facts of the violation. In other words, the counterparty is equivalent to waive the corresponding procedural rights such as statements and defenses. In this case, the information collected by the electronic eye did not constitute the final decision, but was because the actual behavior of the person being punished to sign and pay the fine recognized the validity of the administrative penalty. Of course, there are many kinds of defects in the data source. If the error in filling in the information by yourself, it should be attributed to your own reasons and requires sufficient reasons to explain it before it can be corrected.
2. Algorithm defects
Since in most cases, the algorithm still depends on the technicians who set the algorithm, and the content of this automated administrative decision made using the machine is also determined by civil servants and implemented by inputting programs. Therefore, this decision relying on the algorithm is an administrative activity. Correspondingly, problems caused by algorithm errors should also be borne by the administrative authorities and their staff. However, based on factors such as the algorithm black box, if it is a purely fully automated administrative decision, how to review the algorithm still needs further exploration. In other words, algorithms are also the essence of a "law". The review of algorithm defects should become an independent proposition after the development of automated administration and do not belong to the existing content in the current administrative legal system.
3. System defects
At this time, it is closer to a factual behavior and will also cause later compensation issues. In theory, the problem with the automation system may be due to natural factors of force majeure or errors caused by human factors. If it is the former, there is no legality issue, but the damage caused by this will also cause compensation; if it is the latter, due to the close relationship between the system and the administrative agency, the administrative agency and its staff should also bear the main responsibility.
To sum up, based on the process of automated administration, namely data sources, algorithms (black boxes) and then the system formed, such a data flow, some related issues can be partially adjusted in the existing legal system. However, more relief needs to rely on more relief methods that target automated administrative characteristics, and will be further discussed in the future.
3. Possible path to relieve defective instructions
Based on the above basis, we can exploratoryly give a possible path to relieve defective instructions in automated administration.
One of the relief paths for defective instructions: legitimate administrative procedures
From a broad perspective, administrative procedures are also one of the relief paths, such as notifications, hearings, etc. As discussed above, automation administration will reduce procedures, so where is the boundary of such reduction procedures? Will there be subsequent relief issues arising from the reduction procedures? These two issues can actually be classified into one, that is, automated administrative instructions have black box attributes, which makes many programs actually reduce, so that the rights and obligations of administrative counterparts are still affected accordingly. A summary of several automated fine payment cases discussed above shows that: Generally speaking, for automated administrative instructions, it is more suitable for cases that can be handled by simple procedures. Of course, with the development of technology, if artificial intelligence can handle complex coding and case scenarios in the future, it is not ruled out that it will fundamentally challenge our existing administrative procedures and regulations. For illegal situations where the amount of punishment is not large and the circumstances are easy to determine, automated law enforcement can be introduced. At the same time, in order to protect the counterparty's procedural rights and reduce or avoid subsequent relief issues, the parties should be required to know, confirm and recognize it, and they cannot simply rely on automated law enforcement machines, otherwise it will be easy to cause the counterparty to panic " The Scene of Ignorance ". The notification process itself should also be in accordance with the requirements of the Administrative Procedure Law, and administrative agencies should adopt electronic methods such as text messages and the Internet to supplement them with written notifications. In addition, there is another point that depends on the decisions made by automated administration, in terms of technical means, it is necessary to ensure the recognizability of the administrative counterpart, that is, the object of the instruction imposition. For example, in facial recognition adopted in many places and fields, giving priority to the recognition of administrative subjects is a basic requirement. Even in road traffic law enforcement using electronic monitoring equipment, the identification of basic vehicle information by administrative entities needs to be given priority.
In addition to the above simple programs, is there any feasible adoption of some relief programs in complex programs? Such as a hearing. The author believes that in most cases, automated administration is more suitable for occasions where simple procedures can be used, which actually ruled out the application of hearings.If automated administrations arise in the subsequent process of obtaining information, such as intrusion of privacy rights, procedural protection will still be required, especially due process protection. For example, during the prevention and control of the new crown pneumonia epidemic, some places require public officials to "not fabricate, spread, or spread rumors through the Internet, social software, etc., keep work secrets, and do not publish, disseminate or disclose non-publicly published epidemic prevention and control work information without authorization." On February 4, 2020, the "Notice on Doing a Good Job in the Protection and Utilization of Big Data to Support Joint Prevention and Control Work" issued by the Office of the Central Cyber Security and Information Technology clearly states that personal information collected for epidemic prevention and control and disease prevention and control shall not be used for other purposes; no unit or individual shall disclose personal information such as name, age, ID number, telephone number, home address, etc. without the consent of the person being collected, except for those who have been desensitized due to the needs of joint prevention and control work. Articles 43-45 of the Cybersecurity Law also stipulate the corresponding personal information protection related content.
Technical due process requires that necessary procedures such as notification and hearing apply, especially when automated administrative decisions involve citizens' personal rights and freedom rights; at the same time, non-discrimination data mining technology must be used technically to avoid discrimination in individuals; in addition, administrative agencies are also required to make certain reasons. For example, Article 14 of the 2018 " Data Protection Act" stipulates that public administration departments using artificial intelligence must notify the subjects who are performing the automated decision as soon as possible on the content of automated decisions within a reasonable period of time. The data subjects can require public organizations to reconsider the automated decisions made on themselves within one month after receiving the notice, or a natural person will reassess the decisions.
The second relief path for defective instructions: Appeal
As a traditional method of relief, appeal is a common method. Can it still be applied in automated administration? In fact, similar to procedures, under specific circumstances, these means still have the necessity to be applicable. However, based on the characteristics of automated administration, we need to carry out some degree of reform of the existing complaint and petition systems to adapt to the information age. For example, professional committees such as Machine Intelligence Committee are established to identify and determine some professional issues involving automated administration. For example, the European Parliament’s Legal Affairs Committee advocates the establishment of European robotics and artificial intelligence institutions; some scholars advocate the establishment of transnational, independent, multi-stakeholder AI and data committees. They believe that artificial intelligence is not just a utility, but needs to be regulated once it matures. It is also a powerful force to reshape life, interaction and the environment, and is a part of the profound transformation of human habitat into the information circle.
In fact, there are many similar institutions in China, such as Chinese Automation Association Pattern Recognition and Machine Intelligence (PRMI). The Professional Committee was established in 1981 and is a member organization of International Pattern Recognition Association (IAPR). In addition, there are the Machine Learning Professional Committee, the Natural Language Understanding Professional Committee, the Discrete Intelligent Computing Professional Committee, the Intelligent Education Technology Professional Committee, the Intelligent Control and Intelligent Management Professional Committee, etc. These committees emphasize more professionalism and are often prone to construction paths of technological deepening, and their role in improving citizen participation is still limited, and most ordinary citizens cannot even know their existence. The Artificial Intelligence Professional Committee should not be just a think tank for administrative agencies, but should also build a bridge between administrative agencies and ordinary citizens, and play an explanatory and explanatory role in automated administrative decision-making. In the process of handling complaints and petitions, we can learn from the goal of setting up professional committees, that is, to add professional considerations, and use third-party entrustment and assistance to better improve the level of complaints and petition handling and the ability of administrative agencies to handle issues related to automated administration. For example, since the implementation of the health code, various places have established review systems, and networked review is also a way of appeal in the information age.
Next step, the field of judicial practice can learn from the experience of British law and establish a Data Protection Specialist (Data Protection Officer) to provide data suggestions to data controllers and handle complaints. In fact, some places in my country have similar practices. For example, Article 43 of the " Shanghai Public Data and One-Stop Service Management Measures " stipulates that public management and service institutions should set up or determine special security management agencies and determine security management responsible persons. The role of the security management responsible person is somewhat similar to that of the UK's data protection specialist, but the difference is that the security management responsible person is mostly staff within the administrative agency, and is often mainly engaged in technical responsibilities and does not extend to the handling of subsequent responsibilities such as appeals. It is worth noting that since the "immediateness" of automated administration is stronger, the traditional aging system regulations cannot or do not have to be applied in the case of corresponding systems. The time limit regulations that cause the loss of administrative counterparts' rights and interests should also be considered for adjustments, especially some relief measures that can be determined in a timely manner can be dealt with as soon as possible, such as review of health code information, and once problems are found, they must be restored in a timely manner.
The third path to relief for defective instructions: Administrative reconsideration and administrative litigation
Administrative procedures, administrative activities or factual behaviors are more likely to play a role in the reconsideration and litigation process in response to the different consequences caused by defective instructions. Overall, in more cases, defects in data sources are more likely to form a source of evidence in reconsideration and litigation in automated administrative decisions. In theory, algorithm defects may constitute the direct object of review of reconsideration and litigation. That is to say, with the strengthening of the court's ability to review algorithms, the ability to review algorithms should also be available in the future. System defects are more classified into factual behavior, and even if it affects the rights and obligations of the counterparty, there is a possibility of reconsideration and litigation. Under the existing legal system, the subject of review is still only based on the corresponding decisions made by automated administration, rather than automated administration itself.
In the reconsideration and litigation process, several points need to be considered in cases involving automated administration: First, from the perspective of system settings and algorithm design, automated administration now plays more the functions of "assistant" and "assistant", even in the case of making decisions directly. Therefore, in most cases, the corresponding responsibilities should also be borne by the administrative organs and their staff, unless there is evidence to the contrary that their responsibilities can be excluded; secondly, the allocation of burden of proof or the explanation of related issues should be carried out from the perspective of benefiting the administrative counterpart. For example, in the above-mentioned guiding case No. 26 "Li Jianxiong v. Guangdong Provincial Department of Transportation Government Information Disclosure Case", the first instance court held that due to the integration and authority that the Guangdong Provincial People's Government "Government Information Online Disclosure System upon Application" as a government information application disclosure platform, if no exception is made, a successful application submitted from the platform should be deemed to be deemed that the relevant administrative authority has received the government information disclosure application submitted by the plaintiff through the Internet. As for the transfer of the application between the external network, the internal network, and the superior and lower administrative agencies, the transfer of the application is an internal management matter of the administrative agency and cannot be a reason for the administrative agency to postpone the processing. Judging from the reason why the Supreme People's Court chose this case as a guiding case, it is also because this case has correctly determined the effectiveness of online electronic data evidence, fully considering the balance of the parties' ability to provide evidence, and it has certain significance for standardizing administrative agencies to accept government information disclosure applications online and protecting the public's right to apply for government information disclosure in accordance with the law. This case has a positive effect on guiding the trial of similar cases. Of course, in many cases, electronic documents involved in automated administration are actually converted into evidence in administrative litigation or reconsideration, and administrative litigation evidence rules can be used to determine them. In fact, with the development of automated administration, it is more necessary to strengthen control over pre- and in-processes. As the final relief entity, judicial organs will face more challenges in algorithm review in the future.
Conclusion
Automated administration weaves a dense network for us, making the administrative legal content that already involves "from cradle to grave" more closely related to civic life. In the face of digitalization, administrative agencies can not only use automated administration to improve administrative efficiency, but also have an impact on due process, privacy protection, autonomy of will, etc. due to automated administration. This is also a new type of technical risk, and administrative agencies can prevent and respond to rules through machine coding. Therefore, how to determine these encodings, the determination algorithm will still depend on the efforts of legal and information technology experts themselves.
Li Jianliang once pointed out that if AI development requires good humanistic conditions and thinking environment corresponding to it, the thinking elements that overlap and pass on epistemology and normology must enter the public law order. When humans face overwhelming and boundless digital invasions, whether the legal system and order need the normative mechanism and practical concept of "human nature reservation" needs to be carefully considered. In other words, as humans who may often accompany automated administration, when the system is created, human values such as dignity, equality, and fairness should be considered, that is, to avoid subsequent problems as much as possible. The handling of automated administrative defective instructions requires us to fully consider the characteristics of automated administrative, and if necessary, it still needs to assist in manual review and other means. If automated administration is regarded as an administrative process, we need to clarify various legal relationships and behavioral attributes from the complex and complex networked appearances to determine the corresponding path for subsequent relief.
However, in fact, based on its uniqueness, there are still many issues that need further discussion. It can be said that the technological revolution will affect the structure of legal directives and the relationship between administrative entities representing public power and counterparts. Some scholars even believe that in the future, more emphasis will be placed on how legislators choose goals rather than how to implement the law, and academic interest in topics such as judicial behavior will gradually disappear. We don't know about this. But from riders trapped in algorithms to drivers who were punished for being disconnected from Beidou to suicide, the challenges brought by automated administration and technological revolution cannot be avoided. In the era of algorithms, the "change" and "unchange" of administrative law still need further exploration.
Original link
Hu Minjie‖On the flaws in automated administration and their relief
"Journal of Beijing Administrative College" Introduction
"Article is the great cause of the country, and journals are the garden of academics." Journal of Beijing Administrative College was founded in 1999. It is a comprehensive academic and theoretical journal of social sciences sponsored by Beijing Administrative College (Party School of the Beijing Municipal Committee of the Communist Party of China). It is a journal from the source of Chinese Social Science Citation Index (CSSCI), a core journal of Chinese national Chinese, a core journal of Chinese humanities and social sciences, a top 100 social science newspapers in the country, a core journal of Chinese humanities and social sciences, and an important source of reprinting from "copied newspaper materials". The main columns of the Journal of "Journal of Science include "Local Government and Governance", "Politics·Administrative", "Marxism and Contemporary", "Capital Research", "Economic·Management", "Legal·Society", "Philosophy·Humanities", "Matthew Ricci and Chinese-Western Cultural Exchange", "Capital Research", "Overseas Academic", etc., and special academic seminars are conducted from time to time.
