By Ernest Lim

My review article in the Oxford Journal of Legal Studies offers a critical analysis of Horst Eidenmüller and Gerhard Wagner’s Law by Algorithm by focusing on four major sets of issues that are covered in this important work: (i) separate legal personality for artificial intelligence (AI) systems; (ii) the exploitation and protection of consumers; (iii) liability; and (iv) online dispute resolution.

On separate legal personality, it is shown that neither unbundled products nor difficulties in proving that the AI systems resulted in damages or losses necessarily justify giving legal personality to these systems. By unbundled products, Eidenmüller and Wagner mean that the software and hardware would not be marketed in a bundle. Instead, users can combine different kinds of software with hardware, or users can be given permission to modify the software. According to the authors, it may be difficult to ascertain whether and the extent to which the manufacturers and suppliers of the component parts as well as the users are liable. But it is unnecessary to confer separate legal personality on AI for liability purposes. It is argued that the principle of presumption of causality introduced in the proposed EU AI Liability Directive and the proposed EU revised Product Liability Directive can address this problem.

As for the dual problem of businesses systematically exploiting widespread behavioural biases in consumers and altering consumers’ preferences using micro-targeted advertisements, it is suggested that the UK Consumer Protection from Unfair Trading Regulations 2008, which prohibit unfair commercial practices and aggressive commercial practices, can be amended in two aspects to address the problem. First, the enforcement authority can request courts to order the defendant to disclose information on the algorithms or how the AI system in question has been developed. Second, in order to show aggressive commercial practice, the Regulations can be amended to remove the requirement on the part of the enforcement authority to show that the commercial practice is likely to cause the consumer to take a transactional decision that she would not have otherwise taken.

With respect to the problem of liability, my review article considers the issues of causation that jurisdictions that have not or will not adopt the EU proposed presumption of causality need to grapple with. These issues are: first, there could be a lack of evidence demonstrating causation due to the acts of commission by human beings (such as the destruction of evidence) or acts of omission (such as the failure to disclose or record the relevant information); and second, it may be impossible to establish causation through no fault of anyone (such as where there is neither destruction of evidence nor failure to disclose or record information or because of the black box nature of the AI system). To address these issues, the article distinguishes at least two different stages of AI design and development: data management; and second, system development and implementation.

Finally, regarding online dispute resolution, two sets of distinctions are drawn. The first is between smart contracts and self-driving contracts, and the second between AI-assisted and AI-substitutionary adjudication. On the first set of distinctions, one concern is whether one party to the contract can trust the other to develop the smart contracts or algorithms for the former. Another concern is whether smart and self-driving contracts can accurately and adequately capture the parties’ intentions. The article then analyses the problems with AI-assisted adjudication: undue reliance on AI software and excessive standardisation of decisions. This is followed by an examination of the problems of AI-substitutionary adjudication such as unfeasibility and legitimacy deficit.

Keywords:  Artificial Intelligence (AI), Algorithms, Liability, EU AI Liability Directive, EU Product Liability Directive

AUTHOR INFORMATION

Professor Ernest Lim is the Vice-Dean (Faculty Development) and Professor of Law at NUS Law.
Email: ernestlim@nus.edu.sg