"Looking under the Bonnet": Algorithmic Audit Systems Based on a Socio-Technical Approach

Abstract

As algorithms increasingly assume the role of infrastructural components in societal operations, the risks associated with algorithmic alienation have become more pronounced. Based on socio-technical systems theory, this paper proposes positioning algorithmic auditing as socio-technical systems auditing, comprehensively examining the interaction between technology and society by "looking under the bonnet" of algorithms. The research first clarifies the origin and evolution of algorithmic auditing, analyzing the significance of socio-technical systems theory for algorithmic auditing. Second, it elucidates the fundamental elements of algorithmic auditing from the dimensions of normative purpose and objects of action, emphasizing that algorithmic auditing should encompass multi-dimensional assessment across the entire lifecycle. Finally, it constructs a comprehensive audit accountability framework that includes algorithmic entity compliance responsibilities, auditability construction, and accountability mechanisms. The research finds that while global algorithmic auditing practices differ, they exhibit common trends: transitioning from technical auditing to socio-technical systems auditing, extending from ex-post regulation to full lifecycle governance, and evolving from soft ethical regulation to hard legal regulation. This paper suggests that future institutionalization of algorithmic auditing should strengthen international collaboration, develop diverse auditing entities, optimize auditing techniques and methodologies, and balance the relationship between innovation and regulation, with the aim of entering into an algorithmic era characterized by technological safety, decision-making fairness, and social harmony.