Stanley Greenstein

The study of law and information technology comes with an inherent contradiction in that while technology develops rapidly and embraces notions such as internationalization and globalization, traditional law, for the most part, can be slow to react to technological developments and is also predominantly confined to national borders. However, the notion of the rule of law defies the phenomenon of law being bound to national borders and enjoys global recognition. However, a serious threat to the rule of law is looming in the form of an assault by technological developments within artificial intelligence (AI). As large strides are made in the academic discipline of AI, this technology is starting to make its way into digital decision-making systems and is in effect replacing human decision-makers. A prime example of this development is the use of AI to assist judges in making judicial decisions. However, in many circumstances this technology is a ‘black box’ due mainly to its complexity but also because it is protected by law. This lack of transparency and the diminished ability to understand the operation of these systems increasingly being used by the structures of governance is challenging traditional notions underpinning the rule of law. This is especially so in relation to concepts especially associated with the rule of law, such as transparency, fairness and explainability. This article examines the technology of AI in relation to the rule of law, highlighting the rule of law as a mechanism for human flourishing. It investigates the extent to which the rule of law is being diminished as AI is becoming entrenched within society and questions the extent to which it can survive in the technocratic society.

We consider counterfactual explanations for private support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. While privacy preservation is essential when dealing with sensitive data, there is a consequent degradation in the classification accuracy due to the introduced perturbations in the classifier weights. For such classifiers, counterfactual explanations need to be robust against the uncertainties in the SVM weights in order to ensure, with high confidence, that the classification of the data instance to be explained is different than its explanation. We model the uncertainties in the SVM weights through a random vector, and formulate the explanation problem as an optimization problem with probabilistic constraint. Subsequently, we characterize the problem's deterministic equivalent and study its solution. For linear SVMs, the problem is a convex second-order cone program. For non-linear SVMs, the problem is non-convex. Thus, we propose a sub-optimal solution that is based on the bisection method. The results show that, contrary to non-robust explanations, the quality of explanations from the robust solution degrades with increasing privacy in order to guarantee a prespecified confidence level for correct classifications.

The widespread use of digital technologies within society incorporating elements of artificial intelligence (AI) is increasing at a phenomenal rate. This technology promises a multitude of advantages for the development of society. However, it also entails risks. A characteristic of AI technology is that, in addition to using knowledge from computer science, it is increasingly being combined with insights from within the cognitive sciences. This deeper insight into human behavior combined with technology increases the ability of those who control the technology to not only predict but also manipulate human behavior. A function of the law is to protect individuals and society from risks and vulnerabilities, including those associated with technology. The more complex the technologies applied in society, the more difficult it may be to identify the potential harms associated with the technologies. Consequently, it is worthwhile discussing the extent to which the law protects society from the risks associated with technology. In applying the law, the dominant method is the “traditional legal science method.” It incorporates a dogmatic approach and can be described as the default legal problem-solving mechanism utilized by legal students and practitioners. A question that arises is whether it is equipped to meet the new demands placed on law in an increasingly technocratic society. Attempting to frame the modern risks to society using a traditional legal science method alone is bound to provide an unsatisfactory result in that it fails to provide a complete picture of the problem. Simply put, modern societal problems that arise from the use of new digital technologies are multidimensional in nature and require a multidisciplinary solution. In other words, applying a restrictive legal approach may result in solutions that are logical from a purely legal perspective but that are out of step with reality, potentially resulting in unjust solutions. This chapter examines the increased digitalization of society from the legal perspective and also elevates the application of the legal informatics approach as a means of better aligning research within in legal science with other disciplines.