Human Participation as a Legal Factor in Dealing with AI Risks

It is a common feeling nowadays to find oneself amazed by the speedy development of technology. Every day there seems to be a new breakthrough, a new accomplishment, without any final points in the foreseeable future. Artificial Intelligence (‘AI’) is one of the main technological achievements of the recent years, in which highly complex cognitive abilities – always regarded as exclusive to humans in the biological world – are not owned by humans anymore. On the opposite: it is not rare to regard machines as much more capable than humans in many aspects, achieving in seconds that which would take as much more time. However, AI bring challenges and it is not a tool free of controversies. As new as it is, it can also bring considerable challenges. Certainly, it is not expected it to be excluded from usage in our daily lives, especially considering the ever-growing amounts of data available in the current world, but society should not be simply taken away by technology’s speed and adopt AI without thinking about its consequences. We may have advanced, but this does not imply in the abandonment of already existing legal obligations, such as human rights, to mention one. The present research aims to analyse how human participation has a place in guaranteeing that AI systems are lawfully and ethically implemented. This is so that we can understand at what stages the human should be considered a legal factor in order to control and oversee AI implementation and employment, and which duties it gives rise to. This will be done via de analysis of the existing literature on the matter, as well as the analysis of international and national regulation on the subject of AI. Through the comparison of the conclusions taken from the doctrinal writings, both in the legal field and in the AI field, it will be possible to assess whether the existing regulation has the same concerns as displayed in the academic field, pointing out to the same duties. Therefore, the first chapter will deal with the core idea of the present paper. After stating what is our understanding of AI and which are the main adverse effects that can arise from it, there will be a focus on why the human factor is necessary to control AI implementation. This is so that the human-centric vision of AI, vastly dealt with in the international and national regulation is ascertained and guaranteed. This analysis will be followed by the detailed study of three cases in which human participation is the solution for AI adverse effects. Chapter II deals with information bias, and how the unexpected and undesired patterns of discrimination can be avoided through constant human monitoring; Chapter III deals with the lack of information, and how the possibility and encouragement of public participation in AI development and regulation can vastly improve the accessibility and intelligibility of AI systems; finally, Chapter IV deals with the limits of autonomous decision-making, in which there are some instances that only humans can guarantee outcomes that will comply with the necessary human-centric vision of AI. All of these chapters will be followed by an analysis of duties arising from the findings of the research.