From March 25-26, 2023, Peking University School of Transnational Law (STL) held a Workshop on Algorithmic Recommendation and Law, co-initiated by Professor Gilad Abiri from STL and Professor Zheng Ge from Koguan School of Law at Shanghai Jiao Tong University. The workshop invited experts and scholars from top universities and institutions such as Peking University, Fudan University, China University of Political Science and Law, Shanghai Jiao Tong University, Beihang University, University of International Business and Economics, and Tencent Research Institute. The conference mainly discussed the opportunities and challenges brought by the rapid development of recommendation algorithm technology, analyzed, and proposed suggestions for accompanying legal regulatory issues.
STL Dean Philip J. McConnaughay delivered the opening speech and welcomed guests from all over China. Professor Gilad Abiri gave a welcome speech, thanking the experts, scholars, and organizers for their support and contributions to the workshop. The workshop was divided into six sessions, with two speakers in each session introducing their research projects.
During the first session, Associate Professor Shen Weiwei, from the School of Law at China University of Political Science and Law, discussed the current misunderstandings about algorithm discrimination, starting from the legislative changes related to the 2022 Recommendation Algorithm Management Standards. He categorized algorithm discrimination into computational algorithmic discrimination and data-driven algorithmic discrimination. He pointed out that future research and supervision should focus on the latter rather than the former and that remedies should shift from judicial relief to administrative department management. The academic community should pay more attention to anti-algorithm discrimination. Dr. Cai Peiru from Fudan University School of Law introduced how recommendation algorithms have become controversial issues, and then analyzed how existing laws respond to these issues by listing relevant provisions in various laws and regulations. She discussed the boundaries of individual rights protection and the tasks that recommendation algorithms should undertake in practice.
In session two, Professor Ding Xiaodong from Renmin University of China Law School started with the four types of controversies regarding recommendation algorithms, exploring the attribution of the right to interpret algorithm systems, the scope and boundaries of this right, the timing requirements for exercising this right, and the requirements for the exercise of power. Professor Ding argued that this interpretive power should not be considered an absolute fixed scope and that companies should be required to continuously self-regulate and provide a prior general explanation to reduce distrust between individuals and automated decision-makers. Associate Professor Hu Ling from Peking University Law School pointed out that in the digital economy, the more common regulatory approaches for algorithmic recommendations are pre-approval, information disclosure, and post-infringement relief. However, what is less discussed is the standardization of recommendation behavior, which requires stability and predictability. From the perspectives of power distribution, economic motivation, evaluation criteria, and enforcement capabilities, the speaker proposed a theoretical framework ranging from administrative subcontracting to platform subcontracting and analyzed the application of this theoretical framework.
Session three featured speeches by Professor Dai Xin, Deputy Dean of Peking University Law School, and Cao Jianfeng, Senior Researcher at Tencent Research Institute. Professor Dai Xin discussed the pathological classification of recommendation algorithms, pointing out that social skepticism about recommendation algorithms varies in different contexts, and proposed a new typological framework based on biases and noise, building upon the research of Kahneman et al. (2021). Researcher Cao Jianfeng, starting from industry practice, proposed the need to establish a three-dimensional governance structure of ethical governance, legal governance, and technical governance to ensure trustworthy, responsible, and human-centered algorithm applications.
In Session four, Associate Professor Liu Xiaochun from China University of Political Science and Law, combining recent judicial cases, explored the logic and impact of recommendation algorithm applications on platform liability in copyright infringement, analyzing how recommendation algorithms affect platforms' responsibility in aiding infringement. Associate Professor Danny Friedmann from STL introduced the downgrade of IP-infringing content in search to an algorithmic gray zone - between displaying search results and canceling search results, highlighting the opacity of this zone for content uploaders and users, and the absence of relevant remedial mechanisms that may stifle the freedom to share information.
In session five, Associate Professor Zhang Xin from University of International Business and Economics Law School systematically sorted out global algorithm governance trends, analyzed the structure, institutions, and future trends of China's algorithm governance using the "Internet Information Service Algorithm Recommendation Management Regulations" as an example, and explored the consensus on algorithm legislation in recent years. Associate Professor Zhao Jingwu from Beihang University pointed out that the implementation of algorithmic transparency obligations, as a core scheme for regulating algorithms, is usually considered a prerequisite for algorithmic accountability mechanisms, but there is a gap in the interpretation of the "causal" relationship between algorithmic transparency and civil liability. In fact, the fulfillment of algorithmic transparency obligations should focus on the"form of notification"and the"content of notification", abandoning binary thinking when judging whether algorithm recommendation service providers are at fault and changing the judgment method for the impact on users' substantive rights and interests.
Session six featured speeches by Professor Zheng Ge from Shanghai Jiao Tong University KoGuan Law School and Assistant Professor Gilad Abiri from STL. Professor Zheng Ge pointed out that some recommendation algorithms should be regarded as threats to human autonomy, while others are only mild promoters, having the effect of pushing uncertain thoughts in one direction or another. Based on this, he proposed a new legal basis for related regulations and their standards, combining the fundamental differences between Kantian and Marxist theories on freedom and autonomy. Professor Gilad Abiri analyzed and elaborated on the following two issues brought about by labor dispatch and algorithm management and their relevance: the possibility of replacing employee will with employer will, i.e., "algorithmic personhood"; and the issue of poverty. He believes that regulating labor dispatch algorithms can alleviate the first issue to some extent but cannot significantly improve the latter.
At the closing ceremony, Professor Zheng Ge thanked the attendees and expressed anticipation for the research achievements presented at the conference.
Peking University School of Transnational Law, established in 2008, is the only law school in the world that offers both an English-taught Juris Doctor (J.D.) and a Chinese-taught Juris Master (J.M.) degree education. STL has a highly internationalized faculty and its curriculum and talent cultivation model are highly forward-looking, focusing on practical needs, and committed to cultivating high-end foreign-related legal talents with an international perspective and global competitiveness to meet the challenges of the new world.