? In the first module, I introduced you to the legal dimensions of AI software and AI hardware. In this second module, we are going to look at the way in which AI is used in public law. Public law is the law that governs the relationship between a state and individuals. Specifically, we are going to look at four distinct ways in which AI is used in the public law context. In the first lecture of this module, Lena considers the link between AI and legal responsibility. Lena asks, for example, what should happen when an intelligent, autonomous machine make some mistake that leads to damage? Who should compensate a pedestrian who's knocked down by an autonomous vehicle? Or who should be blamed when a robotic medical doctor makes an incorrect diagnosis of a patient? In the second lecture, Anderson introduces you to the use of AI in criminal law. Specifically, Ulrika considers the use of AI during criminal investigations and highlights the benefits and shortcomings of using CCTV, that is surveillance cameras in public places, in combination with facial recognition to capture suspects. In the third lecture, Jeffery Atik explores the potential of AI to assist with the translation of analog legal norms into digital code. In particular, Jeff introduces you to the idea of modeling law with the help of AI and explains why this process could be beneficial. In the final and fourth lecture of this module, Wilhelm considers how AI can be used in the public sector, by public authorities, and by public administrations to optimize the services they provide to citizens. Despite many benefits, Wilhelm explains, for example, that the use of AI in the public sector comes with numerous challenges concerning data protection and the ability of citizens to understand why decisions were taken in a certain manner. Over all, the second module of the course aims to give you a snapshot of some of the problems that the use of AI raises in the public sector before we then move on to consider the use of AI in the private sector in the third module.??

It goes without saying that our society can benefit from AI in many ways. But what should we do when an intelligent autonomous machine makes mistake that leads to damage? For example, who should compensate a pedestrian that is knocked down by a self-driving car? Or who should be blamed when a robot doctor makes an incorrect diagnosis of a patient? In the law, the possibility of holding someone responsible for causing damage has several functions. To hold someone responsible is often a necessary condition for obtaining compensation, and it's crucial to societies attribution of blame for wrongful conduct, but linking responsibility is important not only retrospectively to handle harm that has already occurred, but also has a prospective preventive function by deterring people from causing damage for which they can be held liable, and the law contains various tools for holding human beings responsible for harm that they cause. For example, a person that hurts another person by negligence, by being careless might be held responsible by the state in the context of criminal law, or when sued by a private party in Tokyo. But at the same time, today's legal toolbox is developed and adapted to fit the world that we live in as we know it today. Specifically, the attribution of legal responsibility is to a large extent justified by ideas of human free will and control. The introduction of intelligent machines that act autonomously creates challenges for this system. Is it meaningful to blame robot? Can we ask a bad robots for compensation? If there is no practically useful way in which we can hold a machine responsible for the damage it causes, who then should be responsible??

? Well, perhaps we could hold the developer of the intelligent machine responsible. When normal, unintelligent machines like a hair dryer cause damage, we tend to look for the company that made the machine and attribute responsibility to them. But intelligent machines differ from traditional hardwares and other unintelligent machines in ways that makes it challenging to place the responsibility with the developer. This is specifically true for systems that use different kinds of machine learning techniques like CCTV cameras that use advanced facial recognition technology. In short, machine learning means that the system learns from and adapts to its environment, that it is dynamic and changes over time. It is very difficult for the developer to predict or control how the system develops and how it will modify itself, it depends on the environment it interacts with and what it learns from the environment. Many machine learning systems are also high-low part, which means that it can be hard or even impossible for human app server including the developer, to understand why it behaves the way that it does. This raises the question, of course, whether it is reasonable to hold the developer or perhaps the user liable for damage caused by an autonomous machine in situations where they took all reasonable care but something went wrong anyway. In some legal domains, liabilities strict in the sense which of course creates a large incentive for controlling that the product is actually safe. But we could argue that it would be unfair to impose strict liability for damage caused by devices that by definition cannot be fully controlled, and would anyone even dare to develop or use these products under such conditions? Moreover, what happens if the developer of the system is not here anymore but the system that they created is still here and continuing to learn and change? Who should be responsible then? Attribution of legal responsibility for damage will certainly have some role to play in the regulation of AI but it is unlikely to suffice, to compensate for and prevent damage that machines cause, so we need alternative approaches. The tech industry has begun to answer to this need by developing its own standards for responsible AI. Should we perhaps conclude that the regulation of AI must be left to the industry and to voluntary measures? Not necessarily, many believe that industries voluntary approaches must be complemented by legal regulation and argue that the development of intelligent autonomous machines forces us to seek alternative legal ways to serve the functions that traditional retrospective attribution oblique responsibility has so far served. Indeed, the legal toolbox contains various instruments for compensation and prevention not only retrospective responsibility. For example, law can require developers to buy insurances that would compensate people who are harmed by AI when there is no other person to hold legally responsible for the harm. Mandatory insurance is like this, already applying some domains in healthcare, for example, or the law could require developers or users to take precautionary measures to prevent potential risks from materializing in the first place. This kind of proactive preventive approach is common in modern environmental law, for example and who knows? Maybe one day the law will come up with a meaningful way to acknowledge electronic persons so that in the future, we can actually hold machines responsible for what they do. To conclude then, the rise of intelligent autonomous machines and the legal challenges that they create, does not necessarily make legal regulation less relevant in this domain, but it does call for some legal engineering and for a great deal of legal ingenuity.?